3 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 * Server support for kernels:
5 2.6.32-358.18.1.el6 (RHEL6)
6 * Client support for unpatched kernels:
7 2.6.32-358.18.1.el6 (RHEL6)
9 2.6.18-348.3.1.el5 (RHEL5)
10 2.6.16.60-0.69.1 (SLES10)
12 * Recommended e2fsprogs version: 1.42.7.wc1
13 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
14 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
15 stack overflow. For more information, please refer to bugzilla 17630.
16 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
17 the re-exporting nodes to be mounted with "32bitapi" mount option
19 --------------------------------------------------------------------------------
21 05-30-2013 Intel Corporation
23 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
24 * Server support for kernels:
25 2.6.32-358.6.1.el6 (RHEL6)
26 * Client support for unpatched kernels:
27 2.6.32-358.6.1.el6 (RHEL6)
29 2.6.18-348.3.1.el5 (RHEL5)
30 2.6.16.60-0.69.1 (SLES10)
32 * Recommended e2fsprogs version: 1.42.7.wc1
33 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
34 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
35 stack overflow. For more information, please refer to bugzilla 17630.
36 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
37 the re-exporting nodes to be mounted with "32bitapi" mount option
39 --------------------------------------------------------------------------------
41 9-30-2012 Whamcloud, Inc.
43 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
44 * Server support for kernels:
45 2.6.32-279.5.1.el6 (RHEL6)
46 * Client support for unpatched kernels:
47 2.6.32-279.5.1.el6 (RHEL6)
49 2.6.18-238.19.1.el5 (RHEL5)
50 2.6.16.60-0.69.1 (SLES10)
51 * Recommended e2fsprogs version: 1.41.90.wc4
52 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
53 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
54 stack overflow. For more information, please refer to bugzilla 17630.
55 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
56 the re-exporting nodes to be mounted with "32bitapi" mount option
58 Severity : enhancement
60 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
62 Severity : enhancement
64 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
66 Severity : enhancement
68 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
70 Severity : enhancement
72 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
74 Severity : enhancement
76 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
78 --------------------------------------------------------------------------------
80 04-30-2012 Whamcloud, Inc.
82 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
83 * Server support for kernels:
84 2.6.32-220.4.2.el6 (RHEL6)
85 * Client support for unpatched kernels:
86 2.6.32-220.4.2.el6 (RHEL6)
88 2.6.18-238.19.1.el5 (RHEL5)
89 2.6.16.60-0.69.1 (SLES10)
90 * Recommended e2fsprogs version: 1.41.90.wc4
91 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
92 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
93 stack overflow. For more information, please refer to bugzilla 17630.
94 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
95 the re-exporting nodes to be mounted with "32bitapi" mount option
97 Severity : enhancement
99 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
101 Severity : enhancement
103 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
105 Severity : enhancement
107 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
109 Severity : enhancement
111 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
113 --------------------------------------------------------------------------------
115 09-30-2011 Whamcloud, Inc.
117 * Server support for kernels:
118 2.6.18-238.19.1.el5 (RHEL5)
119 2.6.32-131.6.1.el6 (RHEL6)
120 * Client support for unpatched kernels:
121 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
122 2.6.18-238.19.1.el5 (RHEL5)
123 2.6.32-131.6.1.el6 (RHEL6)
124 2.6.29.4-167.fc11 (FC11)
125 2.6.16.60-0.69.1 (SLES10)
127 * Recommended e2fsprogs version: 1.41.90.wc3
128 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
129 Lustre filesystem with 4K stack may cause a stack overflow. For more
130 information, please refer to bugzilla 17630.
131 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
132 mounted with "32bitapi" mount option
134 Severity : enhancement
135 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
137 Severity : enhancement
138 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
140 Severity : enhancement
141 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
144 Frequency : only with SLES10
146 Description: Use OFED "KMP" provided by Novell
147 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
148 Lustre is now built against this package. That means you need to
149 install the ofed-kmp package from Novell for the patchless client.
150 Note that the ofed-kmp that Novell ships may not exactly match the
151 kernel version but should still be compatible.
153 Severity : enhancement
155 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
156 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
158 --------------------------------------------------------------------------------
160 2010-07-15 Oracle, Inc.
162 * Server support for kernels:
163 2.6.18-164.11.1.el5 (RHEL 5)
164 2.6.18-164.11.1.0.1.el5 (OEL 5)
165 * Client support for unpatched kernels:
166 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
167 2.6.18-164.11.1.el5 (RHEL 5),
168 2.6.18-164.11.1.0.1.el5 (OEL 5)
169 2.6.16.60-0.42.8 (SLES 10),
171 2.6.29.4-167.fc11 (FC11)
172 * Recommended e2fsprogs version: 1.41.10-sun2
174 Severity : enhancement
176 Description: Distribute lfs_migrate script with lustre
177 Details : The lfs_migrate mentioned in the manual is now included in the
178 lustre tree in lustre/scripts and will distributed in lustre RPMs.
180 Severity : enhancement
182 Description: Failover nid list added to import proc file.
184 Severity : enhancement
186 Description: "som_preview" mount option required to use SOM
187 Details : The clients will need to use som_preview mount option
191 Description: Lustre does not do 1MB IOs to HW RAID
192 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
194 Severity : enhancement
196 Description: lfs quota output is not very convenient for awk/sed-parsing
197 Details : Some positions in lfs quota output table could be empty or
198 non-empty which made it hard to parse it with scripts, now
199 a dash is put instead of space where there is not supposed
204 Description: Update to OFED 1.5.1
205 Details : OFED 1.5.1 will be used in environments that don't use
206 vendor supplied OFED.
210 Description: Remove lnet. prefix from lctl params display
211 Details : Also renamed lnet.memused to lnet_memused to
216 Description: Use conf_param -d to remove permanant settings.
220 Description: Add list_param -R
221 Details : lctl list_param now supports recursive mode.
224 Bugzilla : 21829, 22312
225 Description: Fix llobdstat.
226 Details : 1. Now updating the write_bytes counter
227 2. Port script fixes from 1.8.
231 Description: decrease the usage of memory on clients.
232 Details : 1. On clients, recycle dentries and inodes unused.
233 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
238 Description: llobdstat fix and enhancement
239 Details : add a counter to set a limit to how many samples will be returned
240 fix a wildcard in the path to limit to obdfilter stats only
244 Description: lfs check only allowed for root
245 Details : Code cleanups and added a regresion test of non-root use of
250 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
251 Details : Fix hr->hr_index overload.
255 Description: Fix FIEMAP functionality.
259 Description: FC11 patchless client support added
263 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
264 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
268 Description: New acc-sm script and various test-framework enhancments.
272 Description: Previously landed patch was reverted.
276 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
280 Description: Add ldiskfs maxdirsize mount option.
284 Description: Fix mkfs.lustre support for 16TB LUNs.
288 Description: Print mmp_check_interval
289 Details : print mmp_check_interval and make it possible to abort
290 mount operation in case it takes too long.
294 Description: Add cascading_rw to lustre mpi tests.
298 Description: Store SOM data in EA.
299 Details : Also several SOM related cleanups.
303 Description: Add refcount to md_open_data
307 Description: Use refcount on lov_request_set to prevent memory corruption.
311 Description: Statahead: use parent dir's i_mutex for synchronization.
315 Description: loadgen improvements.
319 Description: Enable peer health detection in ksocklnd and ko2iblnd.
323 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
327 Description: Fix race on obd_devcie:md_stats
331 Description: Fix SLES 11 client builds.
332 Details : Resolved uninitalized pointer gcc warnings.
336 Description: Allow non-root access for "lfs check".
337 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
341 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
343 Severity : enhancement
345 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
346 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
350 Description: Sanity test_27p fails.
351 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
352 Sanity tests 27 were cleaned up and improved.
356 Description: Unify req format on client/servers
357 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
358 buffers and swabbers approach..
363 Description: drop join file stuff.
365 Severity : enhancement
367 Description: Add 2.6.27 and SLES11 (patchless client) support
371 Description: Fix in ptlrpc_expire_one_request() to print the signed time
374 Severity : enhancement
376 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
379 Severity : enhancement
381 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
383 Severity : enhancement
385 Description: correctly handle big reply message.
386 Details : send LNet event if reply is bigger then buffer and adjust this buffer
391 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
393 Severity : enhancement
395 Description: Add support for OEL5.
399 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
401 Severity : enhancement
403 Description: "weak-modules" support
404 Details : Implement "weak-modules" support which enables kernel modules
405 to be used with any kernel that implements the same kABI. In
406 order to achieve this modules are now installed in
407 /lib/modules/$(uname -r)/updates/kernel on all distributions.
411 Description: client could not reconnect to OST because of an active request.
412 Details : The client now retries to reconnect to the same server, if a
413 connect request failed with EBUSY or -EAGAIN. Test case 35b
414 conf-sanity.sh has been created.
418 Description: Don't allow make backward step on assignin osc next id.
419 Details : race between allocation next id and ll_sync thread can be cause
420 of set wrong osc next id and can be kill valid ost objects.
424 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
428 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
432 Description: Error handling in osc_statfs_interpret() has been improved.
433 Details : Check in osc_statfs_interpret() for EBADR.
437 Description: Do not send statfs() requests to OSTs disabled by administrator.
438 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
440 Severity : enhancement
442 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
444 Severity : enhancement
446 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
448 Severity : enhancement
450 Description: Enhancement for lfs(1) command to use numeric uid/gid.
454 Description: Improved cleanup procedure for conf-sanity test_46a.
458 Description: Do not start bulk transfer, if client tried to reconnect.
459 Details : Do not start bulk transfer, if client tried to reconnect,
460 but failed due to active rpcs. Check for exp_abort_active_req
461 has been implemented.
465 Description: Prevent inconsistences between linux and lustre mount structures.
466 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
467 Make the sleep interruptible.
469 Severity : enhancement
471 Description: provide server to client comms path
472 Details : server to client communications path via new LDLM_SET_INFO rpc
476 Description: truncate starts GFP_FS allocation under transaction causing deadlock
477 Details : ldiskfs_truncate calls grab_cache_page which may start page
478 allocation under an open transaction. This may lead to
479 calling prune_icache with consequent lustre reentrance.
483 Description: send correctly lsm on open replay
484 Details : MDS is trust to LSM size on replay open, but client can set wrong size
489 Description: enable client interface failover
490 Details : When a child reconnects from another NID, properly update export
491 nid hash position and ldlm reverse import.
493 Severity : enhancement
495 Description: Communicate OST degraded/readonly state via statfs to MDS
496 Details : Flags in the statfs returned from OSTs indicate whether the
497 OST is in a degraded RAID state, or if the filesystem has
498 turned read-only after a filesystem error is detected.
503 Description: don't panic if EPROTO was hit when reading symlink
504 Details : correctly handling request reference in error cases.
506 Severity : enhancement
508 Description: Add LustreNetLink, a kernel-userspace communcation path.
510 Severity : enhancement
512 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
517 Description: access to llog context before init.
518 Details : move handling CATALOGS file at osc layer and forbid access to llog
522 Frequency : rare, only when client is writing to OST on local node
524 Description: Avoid deadlock for local client writes
525 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
526 memory freeing context. This allows OST threads to set the
527 PF_MEMALLOC flag on task structures in order to allocate memory
528 from reserved pools and complete IO.
529 Use GFP_HIGHUSER for OST allocations for non-local client writes,
530 so that the OST threads generate memory pressure and allow
531 inactive pages to be reclaimed.
533 Severity : enhancement
535 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
540 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
541 Details : move ldlm namespace creation in setup phase to avoid grab
542 _lprocfs_lock with cli_sem held.
546 Description: Temporarily disable grant shrink.
547 Details : Disable the feature for debugging.
550 Frequency : only when formatting test filesystems
552 Description: Unable to run several mkfs.lustre on loop devices at the same time
553 Details : mkfs.lustre returns error 256 on the concurrent loop devices
554 formatting. The solution is to proper handle the error.
556 Severity : enhancement
558 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
560 Severity : enhancement
562 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
567 Description: resolve race between obd_disconnect and class_disconnect_exports
568 Details : if obd_disconnect will be called to already disconnected export he
569 forget release one reference and osc module can't unloaded.
571 Severity : enhancement
573 Description: Allow tuning service thread via /proc
574 Details : For each service a new
575 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
576 created that can be used to set min/max thread counts, and get the
577 current number of running threads.
582 Descriptoin: don't return error if have partially created objects for file.
583 Details : lov_update_create_set uses set->set_success as index for created
584 objects, so if some requests will be failed, they will have hole
585 at end of array and qos_shrink_lsm for allocate correct lsm.
587 Severity : enhancement
589 Description: Update OFED support to 1.4.1
594 Description: Reduce small size read RPC
595 Details : Set read-ahead limit for every file and only do read-ahead when
596 available read-ahead pages are bigger than 1M to avoid small size
599 Severity : enhancement
601 Description: move AT tunable parameters for more consistent usage
602 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
604 Severity : enhancement
606 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
607 Details : allow skip disconnected ost for send statfs request and hide error
613 Description: fix racy locking of mballoc block bitmaps causing BUG
614 Details : The locking of the mballoc buddy bitmap and the in-memory
615 block bitmap was using two different spin locks in some
616 cases. This made it possible to incorrectly access the
617 mballoc bitmap while another process was modifying it,
618 causing a sanity assertion to fail. While no on-disk corruption
619 was reported, there was some risk of this happening.
622 Frequency : rare, on llog test 6
624 Descriptoin: don't allow connect to already connected import
625 Details : allowing connect to already connected import is hide connecting problem.
630 Descriptoin: allow kill process which wait statahead result
631 Details : for some reasons 'ls' can stick in waiting result from statahead,
632 in this case need way for kill this process.
634 Severity : enhancement
636 Description: Add state history info file, enhance import info file
637 Details : Track import connection state changes in a new osc/mdc proc file;
638 add overview-type data to the osc/mdc import proc file.
640 Severity : enhancement
642 Description: MDS create should not wait for statfs RPC while holding DLM lock.
644 Severity : enhancement
646 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
651 Descriptoin: OST grant too much space to client even there are not enough space.
652 Details : Client will shrink its grant cache to OST if there are no write
653 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
654 this grant cache if there are already not enough avaible space
655 (left_space < total_clients * 32M).
658 Frequency : start MDS on uncleanly shutdowned MDS device
660 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
661 Details : stay in waiting mds<>ost recovery finished produce random bugs
662 due race between two ll_sync thread for one lov target. send
663 ACTIVATE event only if connect realy finished and import have
667 Frequency : rare, connect and disconnect target at same time
669 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
670 Details : don't call obd_disconnect under lov_lock. this long time
671 operation and can block ptlrpcd which answer to connect request.
676 Descriptoin: don't lose wakeup for imp_recovery_waitq
677 Details : recover_import_no_retry or invalidate_import and import_close can
678 both sleep on imp_recovery_waitq, but we was send only one wakeup
682 Frequency : always with long access acl
684 Descriptoin: mds can't pack reply with long acl.
685 Details : mds don't control size of acl but they limited by reint/getattr
688 Severity : enhancement
690 Description: Update to SLES10 kernel-2.6.16.60-0.33.
692 Severity : enhancement
694 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
697 Frequency : start MDS on uncleanly shutdowned MDS device
699 Descriptoin: aborting recovery hang on MDS
700 Details : don't throttle destroy RPCs for the MDT.
703 Frequency : on remount
705 Description: external journal device not working after the remount
706 Details : clear dev_rdonly flag for external journal devices in
712 Description: shutdown vs evict race
713 Details : client_disconnect_export vs connect request race.
714 if client will evicted at this time - we start invalidate
715 thread without referece to import and import can be freed
719 Frequency : rare, need acl's on inode.
721 Description: client can't handle ost additional correctly
722 Details : if ost was added after client connected to mds client can have
723 hit lnet_try_match_md ... to big messages to wide striped files.
724 in this case need teach client to handle config events about add
725 lov target and update client max ea size at that event.
727 Severity : enhancement
729 Description: Changelogs
730 Details : Changelogs are a lightweight mechanism to track filesystem
731 metadata and namespace changes. The changelog is recorded
732 permanently on the MDTs, and is periodically "consumed" / purged
733 when records are no longer needed.
735 Severity : enhancement
737 Description: compact fld format with extents
738 Details : Store range of seq rather than every seq in FLD. Seq
739 controller update FLD rather than clients. In Case of CMD, mdt0
740 has FLD, all other metadata server act as non persistent proxy
741 for FLD queries and cache fld entries in fld cache.
746 Description: don't skip ost target if they assigned to file
747 Details : Drop slow OSCs if we can, but not for requested start idx.
748 This means "if OSC is slow and it is not the requested
749 start OST, then it can be skipped, otherwise skip it only
750 if it is inactive/recovering/out-of-space.
754 Description: more cleanup in mds_lov
755 Details : not send LOV EA under replay, we can't know about they size at this
756 time. Don't allow client connect to mds before any ost connected,
757 for avoid problems with LOV EA size and returning EIO to client.
759 Severity : enhancement
761 Description: Interoperability at server side (Disk interoperability)
763 Severity : enhancement
765 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
767 Severity : enhancement
769 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
771 Severity : enhancement
773 Description: New client IO stack (CLIO).
775 Severity : enhancement
777 Description: Commit on sharing. Eliminate inter-client dependencies between
778 uncommitted transactions by doing transaction commits.
779 Thereby clients may recovery independently.
782 Frequency : Create a symlink file with a very long name
784 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
785 Details : If there is no extra space in the request for early cancels,
786 ldlm_req_handles_avail() returns 0 instead of a negative value.
788 Severity : enhancement
790 Description: Add /proc entry for import status
791 Details : The mdc, osc, and mgc import directories now have
792 an import directory that contains useful import data for debugging
795 Severity : enhancement
797 Description: Re-disable certain /proc logging
798 Details : Enable and disable client's offset_stats, extents_stats and
799 extents_stats_per_process stats logging on the fly.
802 Frequency : Only on FC kernels 2.6.22+
804 Description: oops in statahead
805 Details : Do not drop reference count for the dentry from VFS when lookup,
806 VFS will do that by itself.
808 Severity : enhancement
810 Description: Generic /proc file permissions
811 Details : Set /Proc file permissions in a more generic way to enable non-
812 root users operate on some /proc files.
816 Description: Hitting mdc_commit_close() ASSERTION
817 Details : Properly handle request reference release in
818 ll_release_openhandle().
822 Description: quota recovery deadlock during mds failover
823 Details : This patch includes att18982, att18236, att18237 in bz14840.
825 1. fix osts hang when mds does failover with quotaon
826 2. prevent watchdog storm when osts threads wait for the
831 Frequency : only patchless client
832 Description: add workaround for race between add/remove dentry from hash
834 Severity : enhancement
836 Description: Allow OST glimpses to return PW locks
840 Description: LBUG when llog conf file is full
841 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
845 Description: Prevent import from entering FULL state when server in recovery
849 Description: service mount cannot take device name with ":"
850 Details : Only when device name contains ":/" will mount treat it as
856 Description: replace ptlrpcd with the statahead thread to interpret the async
857 statahead RPC callback
861 Frequency : on recovery
862 Description: I/O failures after umount during fail back
863 Details : if client reconnected to restarted server we need join to recovery
864 instead of find server handler is changed and process self
865 eviction with cancel all locks.
867 Severity : enhancement
869 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
871 Severity : enhancement
873 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
875 Severity : enhancement
877 Description: Upcall on Lustre log has been dumped
878 Details : Allow for a user mode script to be called once a Lustre log has
879 been dumped. It passes the filename of the dumped log to the
880 script, the location of the script can be specified via
881 /proc/sys/lnet/debug_log_upcall.
886 Description: avoid idr_remove called for id which is not allocated.
887 Details : Move assigment s_dev for clustered nfs to end of initialization,
888 to avoid problem with error handling.
893 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
894 Details : When connection is reused this not moved from CONN_UNUSED_HASH
895 into CONN_USED_HASH and this prodice warning when put connection
896 again in unused hash.
899 Severity : enhancement
901 Description: File striping can now be set to use an arbitrary pool of OSTs.
903 Severity : enhancement
905 Description: Export bytes_read/bytes_write count on OSC/OST.
909 Description: Early reply size mismatch, MGC loses connection
910 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
911 the connect flags are properly negotiated.
916 Description: quotas are not honored with O_DIRECT
917 Details : all writes with the flag O_DIRECT will use grants which leads to
918 this problem. Now using OBD_BRW_SYNC to guard this.
922 Description: add quota statistics
923 Details : 1. sort out quota proc entries and proc code.
924 2. add quota statistics
926 Severity : enhancement
928 Description: enable quota support for HEAD.
932 Description: Properly propagate oinfo flags from lov to osc for statfs
933 Details : restore missing copy oi_flags to lov requests.
935 Severity : enhancement
937 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
938 : l_getgroups(8), lst(8), routerstat(8)
939 Details : included man pages for llobdstat(8), llstat(8),
940 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
942 Severity : enhancement
944 Description: configure's --enable-quota should check the
945 : kernel .config for CONFIG_QUOTA
946 Details : configure is terminated if --enable-quota is passed but
947 : no quota support is in kernel
951 Description: Remove portals compatibility
952 Details : Remove portals compatibility, not interoperable with releases
957 Description: Resolve device initialization race
958 Details : Prevent proc handler from accessing devices added to the
959 obd_devs array but yet be intialized.
961 Severity : enhancement
963 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
965 Severity : enhancement
967 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
972 Description: Using wrong pointer in osc_brw_prep_request
973 Details : Access to array[-1] can produce panic if kernel compiled with
974 CONFIG_PAGE_ALLOC enabled
978 Description: Client runs out of low memory
979 Details : Consider only lowmem when counting initial number of llap pages
983 Description: *optional* service tags registration
984 Details : if the "service tags" package is installed on a Lustre node
985 When the filesystem is mounted, a local-node service tag will
986 be created. See http://inventory.sun.com/ for more information
987 about the Service Tags asset management system.
991 Description: Kernel BUG tries to release flock
992 Details : Lustre does not destroy flock lock before last reference goes
993 away. So always drop flock locks when client is evicted and
994 perform unlock regardless of successfulness of speaking to MDS.
998 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1002 Description: sanity test 65a fails if stripecount of -1 is set
1003 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1008 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1009 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1012 Severity : enhancement
1014 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1018 Description: do not process already freed flock
1019 Details : flock can possibly be freed by another thread before it reaches
1020 to ldlm_flock_completion_ast.
1024 Description: LBUG during stress test
1025 Details : Need properly lock accesses the flock deadlock detection list.
1029 Description: oops in page fault handler
1030 Details : kernel page fault handler can return two special 'pages' in error
1031 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1035 Description: timeout with invalidate import.
1036 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1037 should be handled by ptlrpcd. This produce long age waiting and
1038 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1040 Severity : enhancement
1042 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1046 Description: enable MGS and MDT services start separately
1047 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1048 a co-located MGS without starting the MGS, which is a complement
1049 to 'nosvc' mount option.
1053 Frequency : after recovery
1054 Description: precreate to many object's after del orphan.
1055 Details : del orphan st in oscc last_id == next_id and this triger growing
1056 count of precreated objects. Set flag LOW to skip increase count
1057 of precreated objects.
1061 Frequency : rare, on clear nid stats
1062 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1063 Details : when clean nid stats sometimes try destroy live entry,
1064 and this produce panic in free.
1068 Description: Stack overflow during MDS log replay
1069 ease stack pressure by using a thread dealing llog_process.
1073 Description: wait until IO finished before start new when do lock cancel.
1074 Details : VM protocol want old IO finished before start new, in this case
1075 need wait until PG_writeback is cleared until check dirty flag and
1076 call writepages in lock cancel callback.
1078 Severity : enhancement
1080 Description: using special macro for print time and cleanup in includes.
1084 Description: mds_mfd_close() ASSERTION(rc == 0)
1085 Details : In mds_mfd_close(), we need protect inode's writecount change
1086 within its orphan write semaphore to prevent possible races.
1090 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1094 Frequency : rare, on shutdown ost
1095 Description: don't hit live lock with umount ost.
1096 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1097 use shrink_dcache_sb instead.
1101 Description: don't panic with use echo client
1102 Details : echo client pass NULL as client nid pointer and this produce null
1103 pointer dereference.
1107 Description: fix build on ppc32
1108 Details : compile code with -m64 flag produce wrong object file for ppc32.
1112 Description: add message levels for liblustreapi
1116 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1117 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1118 will result in an -EAGAIN and in the exit call path we call
1119 alloc_rr() which will with increasing aggressiveness attempt to
1120 aquire precreated objects on the minimum number of required OSCs.
1124 Description: Use old size assignment to avoid deadlock
1125 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1126 the scheduling while holding a spinlock. We do not need locking
1127 for size in ll_update_inode() because size is only updated from
1128 the MDS for directories or files without objects, so there is no
1129 other place to do the update, and concurrent access to such inodes
1130 are protected by the inode lock.
1134 Description: resolve "_IOWR redefined" build error on SLES10
1138 Description: dump the memory debugging after all modules are unloaded to
1139 suppress false negative in conf_sanity test 39
1141 Severity : enhancement
1143 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1148 Description: fix SLES kernel versioning
1149 Details : the kernel version for our SLES 10 kernel did not include a "-"
1150 before the "smp" at the end. while this was not a problem in
1151 general, it did mean that software trying to use the kernel
1152 version to try to detect a vendor specific kernel would fail.
1153 this was most evident by the OFED build scripts.
1157 Description: Don't update lov_desc members until making sure they are valid
1158 Details : When updating lov_desc members via proc fs, need fix their
1159 validities before doing the real update.
1163 Description: don't put request into delay list while invalidate in flight.
1164 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1165 import in flight. this produce timeout for invalidate and sometimes
1166 can cause stale data.
1170 Frequency : on ppc only
1171 Description: not convert ost objects for directory because it's not exist.
1172 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1176 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1178 Severity : enhancement
1180 Description: Update to support 2.6.22.14 vanilla kernel.
1184 Frequency : rare, on recovery
1185 Description: read procfs can produce deadlock in some situation
1186 Details : Holding lprocfs lock which send rpc can produce block for destroy
1187 obd objects and this also block reconnect with -EALREADY. This
1188 isn't fix all lprocfs bugs - but make it rare.
1190 Severity : enhancement
1192 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1195 Frequency : frequent on X2 node
1197 Description: mdc_set_open_replay_data LBUG
1198 Details : Set replay data for requests that are eligible for replay.
1202 Description: lustre_mgs: operation 101 on unconnected MGS
1203 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1204 MGC, and late on MGC cannot successfully connect to MGS and a lot
1205 of the error messages complaining that MGS is not connected.
1209 Frequency : on network error
1210 Description: panic with double free request if network error
1211 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1212 but it's true only for synchronus enqueue, for async enqueue
1213 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1216 Severity : enhancement
1218 Description: client-side metadata stat-ahead during readdir(directory readahead)
1219 Details : perform client-side metadata stat-ahead when the client detects
1220 readdir and sequential stat of dir entries therein
1223 Frequency : on start mds
1225 Description: Implement get_info(last_id) in obdfilter.
1228 Frequency : occasional
1230 Description: Correctly check stale fid, not start epoch if ost not support SOM
1231 Details : open with flag O_CREATE need set old fid in op_fid3 because
1232 op_fid2 was overwritten with new generated fid, but mds can answer
1233 with one of these two fids and both is not stale. Setattr
1234 incorrectly started an epoch and assume will be called
1235 done_writing, but without SOM done_writing ever being called.
1238 Frequency : rare, depends on device drivers and load
1240 Description: MDS or OSS nodes crash due to stack overflow
1241 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1242 In some cases, in conjunction with device drivers that use a lot
1243 of stack the MDS (or possibly OSS) service threads could overflow
1244 the stack. One change which was identified to consume additional
1245 stack has been reworked to avoid the extra stack usage.
1248 Frequency : occasional
1250 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1251 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1252 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1253 oscc_has_objects and see OSC has no objests and call
1254 oscc_internal_create to resend create request.
1256 Severity : enhancement
1258 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1260 Severity : enhancement
1262 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1265 Frequency : very rare
1267 Description: Fix replay if there is an un-replied request and open
1268 Details : In some cases, older replay request will revert the
1269 mcd->mcd_last_xid on MDS which is used to record the client's
1270 latest sent request.
1272 Severity : enhancement
1274 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1276 Severity : enhancement
1278 Description: Add rhel5 support to HEAD.
1280 Serverity : enhancement
1282 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1287 Description: Don't allow skipping OSTs if index has been specified.
1288 Details : Don't allow skipping OSTs if index has been specified, make
1289 locking in internal create lots better.
1293 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1294 Details : access to bitfield in structure is always rounded to long
1295 and this produce problem with not atomic change any bit.
1299 Description: Lustre make rpms failed.
1300 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1301 builds Lustre rpms from tarball.
1304 Frequency : rare on shutdown ost
1306 Description: If llog cancel was not send before clean_exports phase, this can
1307 produce deadlock in llog code.
1308 Details : If llog thread has last reference to obd and call class_import_put
1309 this produce deadlock because llog_cleanup_commit_master wait when
1310 last llog_commit_thread exited, but this never success because was
1311 called from llog_commit_thread.
1315 Description: allow userland application know is lost one of stripes.
1316 Details : fill lvb_blocks with error code on ost and return it to
1317 application if error flag found.
1321 Description: NULL lov_tgts causing MDS oops
1322 Details : more safe checks for NULL lov_tgts for avoid oops.
1324 Severity : enhancement
1326 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1330 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1331 Details : If an OST is down the MDS will hang indefinitely in
1332 obd_statfs_rqset() waiting for the statfs data. While for
1333 MDS QOS usage of statfs, it should not stuck in waiting.
1335 Severity : enhancement
1337 Description: remote_acl support
1338 Details : Support ACL-based permission check for remote user.
1339 Support setfacl/getfacl for remote user with the utils
1340 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1341 the system "{s,g}etfacl" utils.
1343 Severity : enhancement
1345 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1347 Severity : enhancement
1349 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1353 Description: make mgs_setparam() handle fsname containing dash
1354 Details : fsname containing a dash does not work with lctl conf_param
1356 Severity : enhancement
1358 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1360 Severity : enhancement
1362 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1366 Frequency : rare, at shutdown
1367 Description: access already free / zero obd_namespace.
1368 Details : if client_disconnect_export was called without force flag set,
1369 and exist connect request in flight, this can produce access to
1370 NULL pointer (or already free pointer) when connect_interpret
1371 store ocd flags in obd_namespace.
1375 Frequency : only at startup
1376 Description: not alloc memory with spinlock held.
1377 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1378 if any spinlock held.
1380 Severity : enhancement
1382 Description: make lustre randomly fail allocating memory
1383 Details : Make lustre randomly failed allocating memory for testing purpose.
1385 Severity : enhancement
1387 Description: lost problems with lov objid file
1388 Details : Fixes some scability and access to not inited memory problems
1389 in work with lov objdid file.
1394 Description: lfs find does not continue on file error
1395 Details : Continue other files processing when a file/dir is absent.
1399 Description: Inconsistent usage of lustre_pack_reply()
1400 Details : Standardize the usage of lustre_pack_reply() such that it
1401 always generate a CERROR on failure.
1404 Frequency : occasional
1406 Description: MDS hang or stay in waiting lock
1407 Details : If client receive lock with CBPENDING flag ldlm need send lock
1408 cancel as separate rpc, to avoid situation when cancel request
1409 can't processed due all i/o threads stay in wait lock.
1413 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1417 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1421 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1425 Description: Update to RHEL4 latest kernel.
1427 Severity : enhancement
1429 Description: Build SLES10 patchless client fails
1430 Details : The configure was broken by run ./configure with
1431 --with-linux-obj=.... argument for patchless client. When the
1432 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1433 can't search header adequately. Use absolute path such as
1434 -I($LINUX)/include instead.
1438 Description: interrupt oig_wait produce painc on resend.
1439 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1440 private set, and this produce situation when rq_ptlrpcd_data not
1441 copyed to new allocated request and triggered LBUG on assert
1442 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1443 ptlrpcd set if request is changed and can be safety changed to use
1448 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1449 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1450 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1451 these macros and replace them with their difinitions instead.
1455 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1456 Details : Update which_patch & target file for SLES10 latest kernel.
1459 Description: Remove client patches from SLES 10 kernel.
1460 Details : This causes SLES 10 clients to behave as patchless clients
1461 even on a Lustre-patched (server) kernel.
1463 Severity : enhancement
1465 Description: self-adjustable client's lru lists
1466 Details : use adaptive algorithm for managing client cached locks lru
1467 lists according to current server load, other client's work
1468 pattern, memory activities, etc. Both, server and client
1469 side namespaces provide number of proc tunables for controlling
1472 Severity : enhancement
1474 Description: light-weight GSS support
1475 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1476 and reduce performance overhead.
1478 Severity : enhancement
1480 Description: Linux keyring support
1481 Details : Support using service of Linux keyring for Lustre GSS internal
1482 context refresh/cache mechanism.
1486 Description: Fix errors in lfs documentation
1487 Details : Fixes man pages
1491 Description: don't use GFP_* in generic Lustre code.
1492 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1496 Description: obdclass is limited by single OBD_ALLOC(idarray)
1497 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1501 Description: warnings with build patchless client with vanila 2.6.19 and up
1502 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1503 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1507 Frequency : Only for SLES
1508 Description: sanity_quota fail test_1
1509 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1510 file, which makes TSTID[2] inunique.
1514 Frequency : Only for SLES9
1515 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1516 Details : Update target/ChangeLog/which_patch .
1520 Frequency : Only for test_10 in sanity-quota.sh
1521 Description: a bug in quota test code
1522 Details : checking a test flag in a wrong place
1526 Frequency : Only for test_13 in sanity-quota.sh
1527 Description: a bug in quota test script
1528 Details : shouldn't check overfull allocated quota
1532 Description: jbd statistics
1533 Details : Port older jbd statistics patch for sles10
1535 Severity : enhancement
1537 Description: Add printing inode info into message about error in writepage.
1542 Description: reply_lock_interpret crash due to race with it and lock cancel.
1543 Details : Do not replay locks that are being cancelled. Do not reference
1544 locks by their address during replay, just by their handle.
1548 Frequency : When flocks are used.
1549 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1550 Details : Flock locks might destroy just granted lock if it could be merged
1551 with another existing flock, this is done in completion handler,
1552 so teach ldlm_cli_enquque_fini that this is a valid case for
1558 Description: Oops in read and write path when failing to allocate lock.
1559 Details : Check if lock allocation failed and return error back.
1563 Description: lstripe command fails for valid OST index
1564 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1565 instead of lov->desc.ld_active_tgt_count.
1569 Description: sanity.sh failed test 103
1570 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1575 Description: df doesn't work properly if diskfs blocksize != 4K
1576 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1580 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1581 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1583 Severity : enhancement
1585 Description: Adaptive timeouts
1586 Details : RPC timeouts adapt to changing server load and network
1587 conditions to reduce resend attempts and improve recovery time.
1591 Description: llapi_file_create() does not allow some changes
1592 Details : add llapi_file_open() that allows specifying the mode and
1593 open flags, and also returns an open file handle.
1597 Description: merge and cleanup kernel patches.
1598 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1602 Description: Add journal checksum support.(Kernel part)
1603 Details : The journal checksum feature adds two new flags i.e
1604 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1605 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1606 indicates that the commit block contains the checksum for
1607 the blocks described by the descriptor blocks. Now commit
1608 record can be sent to disk without waiting for descriptor
1609 blocks to be written to disk. This behavior is controlled
1610 using JBD2_FEATURE_ASYNC_COMMIT flag.
1614 Description: OSS needs mutliple precreate threads
1615 Details : Add ability to start more than one create thread per OSS.
1619 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1620 Details : execute lfs setstripe on client
1624 Description: mds_obd_create error creating tmp object
1625 Details : When the user sets quota on root, llog will be affected and can't
1626 create files and write files.
1629 Frequency : Always on ia64 patchless client, and possibly others.
1631 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1632 Details : This allows the patchless client to be loaded on architectures
1633 without this export.
1637 Description: RedHat Update kernel for RHEL5
1638 Details : Add the kernel config file for RHEL5.
1642 Description: RedHat Update kernel for RHEL5
1643 Details : Modify the kernel config file more closer RHEL5.
1647 Description: Build failure against Centos5 (RHEL5)
1648 Details : Define PAGE_SIZE when it isn't present.
1652 Description: test_7 on sanity-quota.sh can't be used on separate machines
1653 Details : add facet to handle it
1657 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1658 Details : with NFS, the anon dentry's parent was set to itself in
1659 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1660 corresponding dentry other than use rec->ur_name.
1662 Severity : enhancement
1664 Description: lfs setstripe enhancement
1665 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1669 Description: enable data checksumming by default
1670 Details : enable checksum by default, allow --disable-checksum
1671 configure option and "-o nochecksum" mount option. Checksums
1672 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1676 Description: lustre support for RHEL5
1677 Details : Add support for RHEL5.
1681 Description: Client eviction due to failover config
1682 Details : after a connection loss, the lustre client should attempt to
1683 reconnect to the last active server first before trying the
1684 other potential connections.
1688 Description: when mds and osts use different quota unit(32bit and 64bit),
1689 quota will be released repeatly.
1690 Details : void sending multiple quota reqs to mds, which will keep the
1691 status between the reqs.
1695 Description: osts not allocated evenly to files
1696 Details : change the condition to increase offset_idx
1699 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1701 Description: Data corruption for OSTs that are formatted larger than 2TB
1703 Details : When generating the bio request for lustre file writes the
1704 sector number would overflow a temporary variable before being
1705 used for the IO. The data reads correctly from Lustre (which
1706 will overflow in a similar manner) but other file data or
1707 filesystem metadata may be corrupted in some cases.
1711 Description: Tune the kernel for good SCSI performance.
1712 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1713 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1718 Description: rewrite ext2-derived code in obdclass/uuid.c
1719 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1720 from scratch preserving functionality.
1724 Description: "lfs find -obd UUID" prints directories
1725 Details : "lfs find -obd UUID" will return all directory names instead
1726 of just file names. It is incorrect because the directories
1727 do not reside on the OSTs.
1731 Description: use i_size_read and i_size_write in 2.6 port
1732 Details : replace inode->i_size access with i_size_read/write()
1735 Frequency : when removing large files
1737 Description: scheduling issue during removal of large Lustre files
1738 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1739 It causes scheduling issues when removing large files (17TB in the
1743 Frequency : only with liblustre clients on XT3
1745 Description: evictions taking too long
1746 Details : allow llrd to evict clients directly on OSTs
1749 Frequency : only on ppc
1751 Description: /proc/fs/lustre/devices broken on ppc
1752 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1753 We should make sure the type of 'index' is loff_t and then cast
1754 explicitly as needed below. Do not assign an explicitly cast
1758 Frequency : only for rhel5
1760 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1761 Details : Modify the target file & which_kernel.
1763 Severity : enhancement
1765 Description: omit set fsid for export NFS
1766 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1767 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1771 Frequency : Always, for kernels after 2.6.16
1772 Description: Fix warning idr_remove called for id=.. which is not allocated.
1773 Details : Last kernels save old s_dev before kill super and not allow
1774 to restore from callback - restore it before call kill_anon_super.
1778 Description: buffer overruns could theoretically occur
1779 Details : llapi_semantic_traverse() modifies the "path" argument by
1780 appending values to the end of the origin string, and a
1781 overrun may occur. Adding buffer overrun check in liblustreapi.
1785 Description: Fix error on 'ls .' at the top of the Lustre mount.
1786 Details : Don't revalidate dentry if it is a root dentry.
1790 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1791 Details : Modify vm-tunables-rhel4.patch.
1795 Description: Kernel config for 2.6.18-vanilla.
1796 Details : Modify targets/2.6-vanilla.target.in.
1797 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1798 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1799 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1800 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1804 Description: improve handling recoverable errors
1805 Details : if request processig with error which can be recoverable on server
1806 request should be resend, otherwise page released from cache and
1811 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1812 Details : Modify target file & which_patch.
1813 A flaw was found in the IA32 system call emulation provided
1814 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1815 value could be stored in the %RAX register, which could trigger an
1816 out-of-bounds system call table access. An untrusted local user
1817 could exploit this flaw to run code in the kernel
1818 (ie a root privilege escalation). (CVE-2007-4573).
1822 Description: O_DIRECT bypasses client statistics.
1823 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1824 read_ahead_stats were updated. copied the stats section from
1825 osc_send_oap_rpc() into async_internal().
1829 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1833 Description: change order of libsysio includes
1834 Details : '#include sysio.h' should always come before '#include xtio.h'
1838 Description: handle "serious error: objid * already exists" more gracefully
1839 Details : If LAST_ID value on disk is smaller than the objects existing in
1840 the O/0/d* directories, it indicates disk corruption and causes an
1841 LBUG(). If the object is 0-length, then we should use the existing
1842 object. This will help to avoid a full fsck in most cases.
1844 Severity : enhancement
1846 Description: adapt the lustre_config script to support the upgrade case
1847 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1848 server targets to 1.6 in parallel.
1852 Description: To avoid grant space > avaible space when the disk is almost
1853 full. Without this patch you might see the error "grant XXXX >
1854 available" or some LBUG about grant, when the disk is almost
1856 Details : In filter_check_grant, for non_grant cache write, we should
1857 check the left space by if (*left > ungranted + bytes), instead
1858 of (*left > ungranted), because only we are sure the left space
1859 is enough for another "bytes", then the ungrant space should be
1860 increase. In client, we should update cl_avail_grant only there
1861 is OBD_MD_FLGRANT in the reply.
1865 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1866 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1867 described in CVE-2007-4573. This problem could allow untrusted
1868 local users to gain root access.
1871 Frequency : when using O_DIRECT and quotas
1873 Description: Incorrect file ownership on O_DIRECT output files
1874 Details : block usage reported by 'lfs quota' does not take into account
1875 files that have been written with O_DIRECT.
1880 Description: touch file failed when fs is not full
1881 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1882 otherwise we can get ENOSP while fs is not full.
1886 Description: parallel lock callbacks
1887 Details : Instead of sending blocking and completion callbacks as separated
1888 requests, adding them to a set and sending in parallel.
1891 Frequency : only for Cray XT3
1892 Bugzilla : 12829/13455
1893 Description: Changing primary group doesn't change the group lustre assigns to
1895 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1896 overridden with the primary group provided by the group upcall,
1897 whereas the supplied fsgid can be trusted if it is in the list of
1898 supplementary groups returned by the group upcall.
1900 Severity : enhancement
1902 Description: Allow masking D_WARNING, D_ERROR messages from console
1903 Details : Console messages can now be disabled via lnet.printk.
1907 Description: User code with malformed file open parameter crashes client node
1908 Details : Before packing join_file req, all the related reference should be
1909 checked carefully in case some malformed flags cause fake
1910 join_file req on client.
1914 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1915 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1916 mode is equal to the granted mode. However, it should also check
1917 that LDLM_FL_CP_REQD is not set.
1921 Description: Heavy nfs access might result in deadlocks
1922 Details : After ELC code landed, it is now improper to enqueue any mds
1923 locks under och_sem, because enqueue might want to decide to
1924 cancel open locks for same inode we are holding och_sem for.
1928 Description: Client eviction while running blogbench
1929 Details : A lot of unlink operations with concurrent I/O can lead to a
1930 deadlock causing evictions. To address the problem, the number of
1931 oustanding OST_DESTROY requests is now throttled to
1932 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1933 callbacks are processed in priority.
1937 Description: enable ACLs on MDS by default
1938 Details : ACLs must be enabled on MDS by default.
1941 Frequency : PPC/PPC64 only
1943 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1944 Details : fix duplicated definitions between asm-ppc64/types.h and
1945 lustre_types.h on PPC.
1948 Frequency : PPC/PPC64 only
1950 Description: asm-ppc/segment.h does not exist
1951 Details : fix compile issue on PPC.
1955 Description: better handle error messages in extents code
1958 Frequency : RHEL4 only
1960 Description: mkfs is very slow on IA64/RHEL4
1961 Details : A performance regression has been discovered in the MPT Fusion
1962 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1963 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1964 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1966 Severity : enhancement
1968 Description: SNMP support enhancement
1969 Details : Adding total number of sampled request for an MDS node in snmp
1972 Severity : enhancement
1974 Description: Optimize ldlm waiting list processing for PR extent locks
1975 Details : When processing waiting list for read extent lock and meeting read
1976 lock that is same or wider to it that is not contended, skip
1977 processing rest of the list and immediatelly return current
1978 status of conflictness, since we are guaranteed there are no
1979 conflicting locks in the rest of the list.
1983 Description: Time out and refuse to reconnect
1984 Details : When the failover node is the primary node, it is possible
1985 to have two identical connections in imp_conn_list. We must
1986 compare not conn's pointers but NIDs, otherwise we can defeat
1987 connection throttling.
1991 Description: port llog fixes from b1_6 into HEAD
1992 Details : Port llog reference couting and some llog cleanups from b1_6
1993 (bug 10800) into HEAD, for protect from panic and access to already
1994 free llog structures.
1998 Description: Detect stride IO mode in read-ahead
1999 Details : When a client does stride read, read-ahead should detect that and
2000 read-ahead pages according to the detected stride pattern.
2004 Description: data checksumming impacts single node performance
2005 Details : add support for several checksum algorithm. Currently, only CRC32
2006 and Adler-32 are supported. The checksum type can be changed on
2007 the fly via /proc/fs/lustre/osc/*/checksum_type.
2011 Description: use adler32 for page checksums
2012 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2017 Description: build for x2 fails
2018 Details : fix compile issue on Cray systems.
2022 Description: Properly match for duplicate locks
2023 Details : Due to different lock order from skiplists code, we need to
2024 traverse entire list for now
2027 Frequency : only on PPC/SLES10
2029 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2030 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2031 be defined. Add a hack in mkfs_lustre.c to work around this compile
2036 Description: LASSERT on MDS when client holding flock lock dies
2037 Details : ldlm pool logic depends on number of granted locks equal to
2038 number of released locks which is not true for flock locks, so
2039 just exclude such locks from consideration.
2043 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2044 Details : Use fsfilt_sync() for both the whole filesystem sync and
2045 individual file sync to eliminate dangerous inode locking
2046 with I_LOCK that can lead to a deadlock.
2050 Description: performance in 1.6.3
2051 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2052 sure that 1MB requests are not fragmented by the block layer.
2054 Severity : enhancement
2056 Description: organize the server-side client stats on per-nid basis
2057 Details : Change the structure of stats under obdfilter and mds to
2065 The "uuid"s file would list the uuids of _active_ exports.
2066 And the clear entry is to clear all stats and stale nids.
2068 Severity : enhancement
2070 Description: eliminate client locks in face of contention
2071 Details : file contention detection and lockless i/o implementation
2072 for contended files.
2076 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2079 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2081 Description: MDS may lose file striping (and hence file data) in some cases
2082 Details : If there are additional extended attributes stored on the MDS,
2083 in particular ACLs, SELinux, or user attributes (if user_xattr
2084 is specified for the client mount options) then there is a risk
2085 of attribute loss. Additionally, the Lustre file striping
2086 needs to be larger than default (e.g. striped over all OSTs),
2087 and an additional attribute must be stored initially in the
2088 inode and then increase in size enough to be moved to the
2089 external attribute block (e.g. ACL growing in size) for file
2094 Description: skiplist implementation simplification
2095 Details : skiplists are used to group compatible locks on granted list
2096 that was implemented as tracking first and last lock of each
2097 lock group the patch changes that to using doubly linked lists
2101 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2102 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2103 because LASSERT on both the data supplied by a client, and the
2104 data on disk is dangerous and incorrect.
2106 Severity : enhancement
2108 Description: Slow truncate/writes to huge files at high offsets.
2109 Details : Directly associate cached pages to lock that protect those pages,
2110 this allows us to quickly find what pages to write and remove
2111 once lock callback is received.
2115 Description: more ldlm soft lockups
2116 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2117 starve other threads from the resource lock for a long time in
2118 case of long waiting queue, so change the debug level from
2119 D_OTHER to the less frequently used D_INFO.
2121 Severity : enhancement
2123 Description: add -gid, -group, -uid, -user options to lfs find
2127 Description: Hung threads in invalidate_inode_pages2_range
2128 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2129 once one is completed. As a result, some RPCs are stuck in the
2130 queue and are never sent.
2134 Description: filter threads hungs on waiting journal commit
2135 Details : Cleanup filter group llog code, then only filter group llog will
2136 be only created in the MDS/OST syncing process.
2140 Description: Procfs and llog threads access destoryed import sometimes.
2141 Details : Sync the import destoryed process with procfs and llog threads by
2142 the import refcount and semaphore.
2144 Severity : enhancement
2146 Description: openlock cache of b1_6 port to HEAD
2151 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2152 Details : If insertion of an extent fails, then discard the inode
2153 preallocation and free data blocks else it can lead to duplicate
2158 Description: don't always update ctime in ext3_xattr_set_handle()
2159 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2160 In some cases the ctime should not be updated, for example for
2161 2.0->1.8 compatibility it is necessary to delete an xattr and it
2162 should not update the ctime.
2166 Bugzilla : 15713/16362
2167 Description: Assertion in iopen_connect_dentry in 1.6.3
2168 Details : looking up an inode via iopen with the wrong generation number can
2169 populate the dcache with a disconneced dentry while the inode
2170 number is in the process of being reallocated. This causes an
2171 assertion failure in iopen since the inode's dentry list contains
2172 both a connected and disconnected dentry.
2176 Description: assertion failure in ldlm_handle2lock()
2177 Details : fix a race between class_handle_unhash() and class_handle2object()
2178 introduced in lustre 1.6.5 by bug 13622.
2183 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2184 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2189 Description: oops in sd_iostats_seq_show()
2190 Details : unloading/reloading the scsi low level driver triggers a kernel
2191 bug when trying to access the sd iostat file.
2196 Description: Kernel panics during QLogic driver reload
2197 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2198 patch, causing memory corruption.
2203 Description: journal_dev option does not work in b1_6
2204 Details : pass mount option during pre-mount.
2206 Severity : enhancement
2208 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2209 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2210 extent information of a file. It can be used to map logical blocks
2211 in a file to physical blocks in the block device.
2215 Description: LDLM soft lockups - improvement
2216 Details : It is be possible to send the lock handle along with each read
2217 or write request because the client is already doing a lock match
2218 itself so there isn't any reason the OST should have to re-do that
2222 Frequency : only on Cray X2
2224 Description: X2 build failures
2225 Details : fix build failures on Cray X2.
2229 Description: xid & resent requests
2230 Details : Initialize RPC XID from clock at startup (randomly if clock is
2233 Severity : enhancement
2235 Description: Add lustre_start utility to start or stop multiple Lustre servers
2240 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2241 Details : In case of memory pressure, list_del() can be called twice on
2242 req->rq_history_list, causing a kernel oops.
2246 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2247 Details : fix stack overflow in the distributed lock manager by defering
2248 export eviction after a failed AST to the elt thread instead of
2249 handling it in the dlm interpret routine.
2253 Description: Convert some comments to new format.
2254 Details : Update documenting comments to match doxygen conventions.
2258 Description: Grammar fixes.
2259 Details : A couple of trivial spelling fixes.
2263 Description: OSD_COUNTERS-mandatory
2264 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2265 to simplify the code.
2269 Description: simplify cmm_device freeing logic.
2270 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2274 Description: Add lockdep support to dt_object_operations locking interface.
2275 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2276 parameter indicating lock ordering. Update mdd code to use new
2281 Description: Introduce failloc constants for lockless IO tests.
2282 Details : Add two new failloc constants to test lockless IO. Only one of
2283 them in implemented---another is checked in yet to be landed
2288 Description: Add lockdep support for inode mutex.
2289 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2290 in the situations where Lustre has to lock more than one inode
2295 Description: Add optional invariants checking support.
2296 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2297 configure switch. This macro is to be used for consistency and
2298 sanity checks that are too expensive to be left in `production'
2303 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2304 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2309 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2310 Details : Use spin_lock_nested() in (the only) situation where more than
2311 one ldlm_lock is locked simultaneously. Also, fix possible
2312 dead-lock in ldlm_lock_change_resource() by enforcing particular
2317 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2318 Details : Instead of specifying each ldlm_lock call-back through separate
2319 parameter, wrap them into struct ldlm_callback_suite.
2323 Description: Kill join_lru obd method and its callers.
2324 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2325 in memory---a requirement implemented through ->o_join_lru() obd
2326 method in HEAD. Remove this method and its users.
2330 Description: Add asynchronous ldlm ENQUEUE completion handler.
2331 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2332 case that stock ldlm_completion_ast() cannot handle as it waits
2333 until lock is granted. Introduce new ldlm_completion_ast_async()
2334 for this. Also comment ldlm_completion_ast().
2338 Description: ldlm_error <-> errno conversion.
2339 Details : Add functions to map (rather arbitrary) between LDLM error codes
2340 and standard errno values. CLIO needs this to prevent LDLM specific
2341 constants from escaping ldlm and osc.
2345 Description: Kill unused ldlm_handle2lock_ns() function.
2346 Details : Kill unused ldlm_handle2lock_ns() function.
2350 Description: Add lu_ref support to ldlm_lock
2351 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2352 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2353 and ldlm_resource. LDLM interface has to be changed, because code that
2354 releases a reference on a lock, has to "know" what reference this is.
2355 In the most frequent case
2357 lock = ldlm_handle2lock(handle);
2359 LDLM_LOCK_PUT(lock);
2361 no changes are required. When any other reference (received _not_ from
2362 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2363 instead of LDLM_LOCK_PUT().
2365 Arguably, changes are pervasive, and interface requires some discipline
2366 for proper use. On the other hand, it was very instrumental in finding
2367 a few leaked lock references.
2371 Description: Add ldlm_lock_addref_try().
2372 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2373 attempts to addref a lock that might be being canceled
2378 Description: Add ldlm_weigh_callback().
2379 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2380 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2381 hard-coded `number of pages' logic.
2385 Description: Add lockdep annotations to llog code.
2386 Details : Use appropriately tagged _nested() locking calls in the places
2387 where llog takes more than one ->lgh_lock lock.
2391 Description: Add loi_kms_set().
2392 Details : Wrap kms updates into a helper function.
2396 Description: Constify instances of struct lsm_operations.
2397 Details : Constify instances of struct lsm_operations.
2401 Description: lu_conf support.
2402 Details : On a server, a file system object is uniquely identified
2403 by a fid, which is sufficient to locate and load all object
2404 state (inode). On a client, on the other hand, more data are
2405 necessary instantiate an object. Change lu_object_find() and
2406 friends to take additional `lu_conf' argument describing object.
2407 Typically this includes layout information.
2411 Description: lu_context fixes.
2412 Details : Introduce new lu_context functions that are needed on the client
2413 side, where some system threads (ptlrpcd) are shared by multiple
2414 modules, and so cannot be stopped during module shutdown.
2418 Description: Add start and stop methods to lu_device_type_operations.
2419 Details : Introduce two new methods in lu_device_type_operations, that are
2420 invoked when first instance of a given type is created and last one
2421 is destroyed respectively. This is need by CLIO.
2425 Description: Add lu_ref support to struct lu_device.
2426 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2427 track leaked references.
2431 Description: Introduce lu_kmem_descr.
2432 Details : lu_kmem_descr and its companion interface allow to create
2433 and destroy a number of kmem caches at once.
2437 Description: Fix lu_object finalization race.
2438 Details : Fix a race between lu_object_find() finding an object and its
2439 concurrent finalization. This race is (most likely) not possible
2440 on the server, but might happen on the client.
2444 Description: Introduce lu_ref interface.
2445 Details : lu_ref is a debugging module allowing to track references to
2446 a given object. It is quite cpu expensive, and has to be
2447 explicitly enabled with --enable-lu_ref. See usage description
2452 Description: Factor lu_site procfs stats into a separate function.
2453 Details : Separate lu_site stats printing code into a separate function
2454 that can be reused on a client.
2458 Description: Constify instances of struct {lu,dt,md}_device_operations.
2459 Details : Constify instances of struct {lu,dt,md}_device_operations.
2463 Description: Introduce struct md_site and move meta-data specific parts of
2464 struct lu_site here.
2465 Details : Move md-specific fields out of struct lu_site into special struct
2466 md_site, so that lu_site can be used on a client.
2470 Description: Kill mdd_lov_destroy().
2471 Details : Remove unused mdd code.
2475 Description: Add st_block checking to multistat.c.
2476 Details : Add st_block checking to multistat.c.
2480 Description: Add lu_ref support to struct obd_device.
2481 Details : Add lu_ref tracking to obd_device.
2485 Description: Kill obd_set_fail_loc().
2486 Details : Remove unused code.
2490 Description: Add special type for ptlrpc_request interpret functions.
2491 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2492 there. Actual usage will be in CLIO.
2496 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2497 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2498 doesn't work with lockdep.
2502 Description: Add rwv.c test program.
2503 Details : New testing program exercising readv(2) and writev(2) (Qian).
2507 Description: Add sendfile.c test program.
2508 Details : New testing program exercising sendfile(2) (Jay).
2512 Description: Ratelimit a message that can be very frequent.
2513 Details : Ratelimit a memory allocation failure message that can
2518 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2519 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2523 Description: Liblustre build fixes.
2524 Details : Liblustre build fixes.
2528 Description: libcfs: add cfs_{need,cond}_resched() interface.
2529 Details : libcfs: add cfs_{need,cond}_resched() definition and
2530 implementations for Linux, NT, and liblustre.
2532 Severity : enhancement
2534 Description: More exported tunables for mballoc
2535 Details : Add support for tunable preallocation window and new tunables for
2536 large/small requests
2540 Description: Detect corruption of block bitmap and checking for preallocations
2541 Details : Checks validity of on-disk block bitmap. Also it does better
2542 checking of number of applied preallocations. When corruption is
2543 found, it turns filesystem readonly to prevent further corruptions.
2547 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2548 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2549 window to 0. If it does hit cache miss, and read-pattern is still
2550 stride-io mode, does not reset the stride window, but also does not
2551 increase the stride window length in this case.
2555 Frequency : only for big-endian servers
2556 Description: Check if system is big-endian while mounting fs with extents feature
2557 Details : Mounting a filesystem with extents feature will fail on big-endian
2558 systems since ext3-based ldiskfs is not supported on big-endian
2559 systems. This can be over-riden with "bigendian_extents" mount option.
2561 Severity : enhancement
2563 Description: The root squash functionality
2564 Details : A security feature, which is to prevent users from being able
2565 to mount lustre on their desktop, run as root, and delete
2566 all of the files in the filesystem. The goal is accomplished by
2567 remapping user id (UID) and group id (GID) of the root user to
2568 a UID and GID specified by the system administartor via Lustre
2569 configuration management server (MGS). The functionality also
2570 allows to specify sets of clients for which the remapping does
2575 Description: Excessive recovery window
2576 Details : With AT enabled, the recovery window can be excessively long (6000+
2577 seconds). To address this problem, we no longer use
2578 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2579 timeout no longer depends on the service time, it is set to
2580 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2581 time via pb_service_time.
2585 Description: Watchdog triggered on MDS failover
2586 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2587 the OSTs know that the MDS "UUID" can be reused for the same export
2588 from a different NID, so we do not need to wait for the export to be
2592 Frequency : rare, only if using MMP with Linux RAID
2594 Description: MMP doesn't work with Linux RAID
2595 Details : While using HA for Lustre servers with Linux RAID, it is possible
2596 that MMP will not detect multiple mounts. To make this work we
2597 need to unplug the device queue in RAID when the MMP block is being
2598 written. Also while reading the MMP block, we should read it from
2599 disk and not the cached one.
2601 Severity : enhancement
2603 Description: open file using fid
2604 Details : A file can be opened using just its fid, like
2605 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2608 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2611 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2612 Details : add proper locking for creation of jbd_4k slab cache
2616 Description: MMP check in ext3_remount() fails without displaying any error
2617 Details : When multiple mount protection fails during remount, proper error
2620 Severity : enhancement
2622 Description: Allow stripe size to be up to 4G-64k
2623 Details : Fix math logic to allow large stripe sizes.
2627 Description: add check for >8TB ldiskfs filesystems
2628 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2629 Don't allow >8TB ldiskfs filesystems to be mounted without
2630 force_over_8tb mount option
2635 Descriptoin: OSS complains "too many blocks" and crashes
2636 Details : In filter_commitrw_write():
2637 1. lquota_chkquota();
2638 2. obdo_from_inode();
2639 3. lquota_pending_commit();
2640 "2." has changed uid and gid in this case, then caused
2641 lquota_chkquota() and lquota_pending_commit() to use
2642 different uid and gid. That is the root of the bug.
2646 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2647 Details : Detect and print out the 'stripe count = -1' case properly.
2651 Description: Out or order replies might be lost on replay
2652 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2653 request with tid smaller then one currently being added, add it
2654 to the start, not end of the list.
2656 --------------------------------------------------------------------------------
2658 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2660 * Support for kernels:
2661 2.6.5-7.283 (SLES 9),
2662 2.6.9-55.EL (RHEL 4),
2663 2.6.16.46-0.14 (SLES 10),
2664 2.6.18.8 vanilla (kernel.org)
2665 * Client support for unpatched kernels:
2666 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2667 2.6.16 - 2.6.22 vanilla (kernel.org)
2668 * Due to recently discovered recovery problems, we do not recommend
2669 using patchless RHEL 4 clients with this or any earlier release.
2670 * Recommended e2fsprogs version: 1.39.cfs8
2671 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2672 * Starting with this release, the ldiskfs backing filesystem required
2673 by Lustre is now in its own package, lustre-ldiskfs. This package
2674 should be installed. It is versioned separately from Lustre and
2675 may be released separately in future.
2679 Description: block reactivating mgc import until all deactivates complete
2680 Details : Fix race when failing back MDT/MGS to itself (testing)
2683 Frequency : at statup only
2685 Description: mds_lov_synchronize race leads to various problems
2686 Details : simultaneous MDT->OST connections at startup can cause the
2687 sync to abort, leaving the OSC in a bad state.
2689 Severity : enhancement
2691 Description: add optional extra BUILD_VERSION info
2692 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2693 to override the lustre version.
2696 Frequency : 2.6.18 servers only
2698 Description: ll_kern_mount() doesn't release the module reference
2699 Details : The ldiskfs module reference count never drops down to 0
2700 because ll_kern_mount() doesn't release the module reference.
2705 Description: server LBUG when using old ost_num_threads parameter
2706 Details : Accept the old ost_num_threads parameter but warn that it
2707 is deprecated, and fix an off-by-one error that caused an LBUG.
2712 Description: Transient SCSI error results in persistent IO issue
2713 Details : iobuf->dr_error is not reinitialized to 0 between two
2717 Frequency : sometimes when underlying device returns I/O errors
2719 Description: OSTs not going read-only during write failures
2720 Details : OSTs are not remounted read-only when the journal commit threads
2721 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2722 instead of the ext3 wrappers.
2726 Description: poor connect scaling with increasing client count
2727 Details : Don't run filter_grant_sanity_check for more than 100 exports
2728 to improve scaling for large numbers of clients.
2731 Frequency : SLES10 only
2733 Description: sanity-quota.sh quotacheck failed: rc = -22
2734 Details : Quotas cannot be enabled on SLES10.
2737 Frequency : liblustre clients only
2739 Description: getdirentries does not give error when run on compute nodes
2740 Details : getdirentries does not fail when the size specified as an argument
2741 is too small to contain at least one entry
2743 Severity : enhancement
2745 Description: Add LNET router traceability for debug purposes
2746 Details : If a checksum failure occurs with a router as part of the
2747 IO path, the NID of the last router that forwarded the bulk data
2748 is printed so it can be identified.
2753 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2754 Details : Due to a race condition, liblustre clients were occasionally
2755 evicted incorrectly.
2757 Severity : enhancement
2759 Description: lfs setstripe use optional parameters instead of postional
2762 Severity : enhancement
2764 Description: Nanosecond timestamp support for ldiskfs
2765 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2766 resolution timestamps. There is not yet support for this at
2767 the Lustre filesystem level.
2770 Frequency : during server recovery
2772 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2773 Details : request with rq_no_resend flag not awake l_wait_event if they get
2777 Frequency : nfs export on patchless client
2779 Description: connectathon hang when test nfs export over patchless client
2780 Details : Disconnected dentry cannot be found with lookup, so we do not need
2781 to unhash it or make it invalid
2784 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2785 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2786 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2791 Description: open req refcounting wrong on reconnect
2792 Details : If reconnect happened between getting open reply from server and
2793 call to mdc_set_replay_data in ll_file_open, we will schedule
2794 replay for unreferenced request that we are about to free.
2795 Subsequent close will crash in variety of ways.
2796 Check that request is still eligible for replay in
2797 mdc_set_replay_data().
2802 Description: disable writes to filesystem when reading health_check file
2803 Details : the default for reading the health_check proc file has changed
2804 to NOT do a journal transaction and write to disk, because this
2805 can cause reads of the /proc file to hang and block HA state
2806 checking on a healthy but otherwise heavily loaded system. It
2807 is possible to return to the previous behaviour during configure
2808 with --enable-health-write.
2810 Severity : enhancement
2812 Description: 64-bit inode version
2813 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2814 to the inode. This will be required for version-based recovery.
2819 Description: MDS fails to start if a duplicate client export is detected
2820 Details : in some rare cases it was possible for a client to connect to
2821 an MDS multiple times. Upon recovery the MDS would detect this
2822 and fail during startup. Handle this more gracefully.
2824 Severity : enhancement
2826 Description: Add -o localflock option to simulate old noflock
2828 Details : This will achieve local-only flock/fcntl locks
2834 Description: log_commit_thread vs filter_destroy race leads to crash
2835 Details : Take import reference before releasing llog record semaphore
2840 Description: Wrong request locking in request set processing
2841 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2842 imp_delayed_list, in this place should be used imp_lock.
2845 Frequency : when reconnection
2847 Description: Grant Leak when osc reconnect to OST
2848 Details : When osc reconnect ost, OST(filter) should check whether it
2849 should grant more space to client by comparing fed_grant and
2850 cl_avail_grant, and return the granted space to client instead
2851 of "new granted" space, because client will call osc_init_grant
2852 to update the client grant space info.
2855 Frequency : when client reconnect to OST
2857 Description: Grant Leak when osc do resend and replay bulk write
2858 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2859 bulk write request, because the grant info while be sync between
2860 OSC and OST when reconnect, and we should ignore the grant info
2861 these of resend/replay write req.
2866 Description: Grant space more than avaiable left space sometimes.
2867 Details : When then OST is about to be full, if two bulk writing from
2868 different clients came to OST. Accord the avaliable space of the
2869 OST, the first req should be permitted, and the second one
2870 should be denied by ENOSPC. But if the seconde arrived before
2871 the first one is commited. The OST might wrongly permit second
2872 writing, which will cause grant space > avaiable space.
2875 Frequency : when client is evicted
2877 Description: Grant might be wrongly erased when osc is evicted by OST
2878 Details : when the import is evicted by server, it will fork another
2879 thread ptlrpc_invalidate_import_thread to invalidate the
2880 import, where the grant will be set to 0. While the original
2881 thread will update the grant it got when connecting. So if
2882 the former happened latter, the grant will be wrongly errased
2883 because of this race.
2888 Description: Checking Stale with correct fid
2889 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2890 to check whether it is stale, because sometimes, we want the
2891 enqueue happened anyway, and op_data.fid2 will not be initialized.
2893 Severity : enhancement
2895 Description: update patchless client
2896 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2899 Frequency : only with 2.4 kernel
2901 Description: random memory corruption
2902 Details : size of struct ll_inode_info is to big for union inode.u and this
2903 can be cause of random memory corruption.
2908 Description: Memory leak in recovery
2909 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2910 It should also check obd_fail before fsfilt_start, otherwise if
2911 fsfilt_start return -EROFS,(failover mds during mds recovery).
2912 then the req will return with repmsg->transno = 0 and rc = EROFS.
2913 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2914 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2915 be freed no matter whether fsfilt_commit success or not.
2918 Frequency : only with huge count clients
2920 Description: Prevents from taking the superblock lock in llap_from_page for
2922 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2923 for llap_from_page call in ll_removepage prevents from taking the
2924 superblock lock for a soon died page.
2929 Description: Not check open intent error before release open handle
2930 Details : in some rare cases, the open intent error is not checked before
2931 release open handle, which may cause
2932 ASSERTION(open_req->rq_transno != 0), because it tries to release
2933 the failed open handle.
2938 Description: Set cat log bitmap only after create log success.
2939 Details : in some rare cases, the cat log bitmap is set too early. and it
2940 should be set only after create log success.
2944 Description: Accessing a block bevice can re-enable I/O when Lustre is
2945 tearing down a device.
2946 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2951 Description: service threads may hog cpus when there are a lot of requests
2952 Details : Insert cond_resched to give other threads a chance to use some CPU
2957 Description: the cat log was not initialized in recovery
2958 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2959 unlink log on mds will not be initialized until mds post
2960 recovery. And also in mds post recovery, the unlink log will
2961 initialization will be done asynchronausly, so there will be race
2962 between add unlink log and unlink log initialization.
2966 Description: brw_stats were being printed incorrectly
2967 Details : brw_stats were being printed as log2 but all of them were not
2968 recorded as log2. Also remove some code duplication arising from
2969 filter_tally_{read,write}.
2973 Frequency : rare, only in recovery.
2974 Description: ASSERTION(req->rq_type != LI_POISON) failed
2975 Details : imp_lock should be held while iterating over imp_sending_list for
2976 prevent destroy request after get timeout in ptlrpc_queue_wait.
2980 Description: replay-single.sh test 52 fails
2981 Details : A lock's skiplist need to be cleanup when it being unlinked
2982 from its resource list.
2986 Description: Short directio read returns full requested size rather than
2988 Details : Direct I/O operations should return actual amount of bytes
2989 transferred rather than requested size.
2991 Severity : enhancement
2993 Description: metadata RPC reduction (e.g. for rm performance)
2994 Details : decrease the amount of synchronous RPC between clients and servers
2995 by canceling conflicing lock before the operation on the client
2996 and packing thier handles into the main operation RPC to server.
2998 Severity : enhancement
3000 Description: Async OSC create to avoid the blocking unnecessarily.
3001 Details : If a OST has no remain object, system will block on the creating
3002 when need to create a new object on this OST. Now, ways use
3003 pre-created objects when available, instead of blocking on an
3004 empty osc while others are not empty. If we must block, we block
3005 for the shortest possible period of time.
3009 Description: improve handling recoverable errors
3010 Details : if request processig with error which can be recoverable on server
3011 request should be resend, otherwise page released from cache and
3014 Severity : enhancement
3016 Description: refine locking for avoid write wrong info into lov_objid file
3017 Details : fix possible races with add new target and write/update data in
3020 --------------------------------------------------------------------------------
3022 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3027 Frequency : on some architectures
3029 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3030 Details : Uninitialized flags sometimes cause configuration commands to
3034 Frequency : patchless clients only
3036 Description: missing __iget() symbol export
3037 Details : The __iget() symbol export is missing. To avoid the need for
3038 this on patchless clients the deathrow inode reaper is turned
3039 off, and we depend on the VM to clean up old inodes. This
3040 dependency was during via the fix for bug 12181.
3042 --------------------------------------------------------------------------------
3044 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3046 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3047 INTEROPERATE with older versions automatically. In many cases a
3048 special upgrade step is needed. Please read the
3049 user documentation before upgrading any part of a live system.
3050 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3051 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3052 * WARNING: Lustre configuration and startup changes are required with
3053 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3057 Severity : enhancement
3059 Description: MountConf
3060 Details : Lustre configuration is now managed via mkfs and mount
3061 commands instead of lmc and lconf. New obd types (MGS, MGC)
3062 are added for dynamic configuration management. See
3063 https://mail.clusterfs.com/wikis/lustre/MountConf for
3066 Severity : enhancement
3068 Description: dynamic OST addition
3069 Details : OSTs can now be added to a live filesystem
3071 Severity : enhancement
3073 Description: startup order invariance
3074 Details : MDTs and OSTs can be started in any order. Clients only
3075 require the MDT to complete startup.
3077 Severity : enhancement
3079 Description: parallel, asynchronous orphan cleanup
3080 Details : orphan cleanup is now performed in separate threads for each
3081 OST, allowing parallel non-blocking operation.
3083 Severity : enhancement
3085 Description: optimized stripe assignment
3086 Details : stripe assignments are now made based on ost space available,
3087 ost previous usage, and OSS previous usage, in order to try
3088 to optimize storage space and networking resources.
3090 Severity : enhancement
3092 Description: Permanently set tunables
3093 Details : All writable /proc/fs/lustre tunables can now be permanently
3094 set on a per-server basis, at mkfs time or on a live system.
3096 Severity : enhancement
3098 Description: Lustre message v2
3099 Details : Add lustre message format v2.
3101 Severity : enhancement
3103 Description: client OST exclusion list
3104 Details : Clients can be started with a list of OSTs that should be
3105 declared "inactive" for known non-responsive OSTs.
3109 Description: SPEC SFS validation failure on NFS v2 over lustre.
3110 Details : Changes the blocksize for regular files to be 2x RPC size,
3111 and not depend on stripe size.
3113 Severity : enhancement
3115 Description: Multiple MD RPCs in flight.
3116 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3117 To avoid overly-overloading MDS, introduce a limit on number of
3118 MDS RPCs in flight for a single client and add /proc controls
3119 to adjust this limit.
3121 Severity : enhancement
3123 Description: client read/write statistics
3124 Details : Add client read/write call usage stats for performance
3125 analysis of user processes.
3126 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3127 file access. extents_stats shows chunk size distribution.
3128 extents_stats_per_process show chunk size distribution per
3131 Severity : enhancement
3133 Description: mds statistics
3134 Details : Add detailed mds operations statistics in
3135 /proc/fs/lustre/mds/*/stats.
3139 Description: Failure of copying files with lustre special EAs.
3140 Details : Client side always return success for setxattr call for lustre
3141 special xattr (currently only "trusted.lov").
3146 Description: Refcount LNET uuids
3147 Details : The global LNET uuid list grew linearly with every startup;
3148 refcount repeated list entries instead of always adding to
3151 Severity : enhancement
3153 Description: Dynamic service threads
3154 Details : Within a small range, start extra service threads
3155 automatically when the request queue builds up.
3158 Frequency : mixed-endian client/server environments
3160 Description: mixed-endian crashes
3161 Details : The new msg_v2 system had some failures in mixed-endian
3164 Severity : enhancement
3166 Description: Easy OST removal
3167 Details : OSTs can be permanently deactivated with e.g. 'lctl
3168 conf_param lustre-OST0001.osc.active=0'
3170 Severity : enhancement
3172 Description: MGS proc entries
3173 Details : Added basic proc entries for the MGS showing what filesystems
3176 Severity : enhancement
3178 Description: provide MGS failover
3179 Details : Added config lock reacquisition after MGS server failover.
3181 Severity : enhancement
3183 Description: add Linux 2.4 support
3184 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3188 Description: a large application tries to do I/O to the same resource and dies
3189 in the middle of it.
3190 Details : Check the req->rq_arrival time after the call to
3191 ost_brw_lock_get(), but before we do anything about
3192 processing it & sending the BULK transfer request. This
3193 should help move old stale pending locks off the queue as
3194 quickly as obd_timeout.
3197 Frequency : when an incorrect nid is specified during startup
3199 Description: ptlrpc connect to non-existant node causes kernel crash
3200 Details : LNET can't be re-entered from an event callback, which
3201 happened when we expire a message after the export has been
3202 cleaned up. Instead, hand the zombie cleanup off to another
3205 Severity : enhancement
3207 Description: plain/inodebits lock performance improvement
3208 Details : Grouping plain/inodebits in granted list by their request modes
3209 and bits policy, thus improving the performance of search through
3213 Frequency : only if OST filesystem is corrupted
3215 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3216 Details : for a short time RPCs with bulk IO are in the replay list,
3217 but replay of bulk IOs is unimplemented. If the OST filesystem
3218 is corrupted due to disk cache incoherency and then replay is
3219 started it is possible to trip an assertion. Avoid putting
3220 committed RPCs into the replay list at all to avoid this issue.
3223 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3225 Description: System hang on startup
3226 Details : This bug allowed the liblustre (e.g. catamount) client to
3227 return to the app before handling all startup RPCs. This
3228 could leave the node unresponsive to lustre network traffic
3229 and manifested as a server ptllnd timeout.
3231 Severity : enhancement
3233 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3234 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3235 boolean to control whether to print peer debug info when a
3236 client's RPC times out.
3239 Frequency : only for kernels with patches from Lustre below 1.4.3
3241 Description: Remove old rdonly API
3242 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3245 Frequency : only for devices with external journals
3247 Description: Set external device read-only also
3248 Details : During a commanded failover stop, we set the disk device
3249 read-only while the server shuts down. We now also set any
3250 external journal device read-only at the same time.
3253 Frequency : when upgrading from 1.4 while trying to change parameters
3255 Description: The wrong (new) MDC name was used when setting parameters for
3256 upgraded MDT's. Also allows changing of OSC (and MDC)
3257 parameters if --writeconf is specified at tunefs upgrade time.
3260 Frequency : when setting specific ost indicies
3262 Description: QOS code breaks on skipped indicies
3263 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3264 created with --index need not be sequential.
3266 Severity : enhancement
3268 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3269 Details : The uninit_groups feature works in conjunction with the kernel
3270 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3271 up the pass1 processing of e2fsck. This is a read-only feature
3272 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3273 mount filesystems that have had this feature enabled.
3275 Severity : enhancement
3277 Description: Improve multi-block allocation algorithm to avoid fragmentation
3278 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3279 allocation locality and avoid filesystem fragmentation.
3281 ------------------------------------------------------------------------------
3283 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3285 * Support for kernels:
3286 2.4.21-47.0.1.EL (RHEL 3)
3287 2.6.5-7.283 (SLES 9)
3288 2.6.9-42.0.10.EL (RHEL 4)
3289 2.6.12.6 vanilla (kernel.org)
3290 2.6.16.27-0.9 (SLES 10)
3291 * Recommended e2fsprogs version: 1.39.cfs5
3293 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3297 Frequency : occasional, depends on client load and configuration
3298 Bugzilla : 12181, 12203
3299 Description: data loss for recently-modified files
3301 Details : In some cases it is possible that recently written or created
3302 files may not be written to disk in a timely manner (this should
3303 normally be within 30s unless client IO load is very high).
3304 The problem appears as zero-length files or files that are a
3305 multiple of 1MB in size after a client crash or client eviction
3306 that are missing data at the end of the file.
3308 This problem is more likely to be hit on clients where files are
3309 repeatedly created and unlinked in the same directory, clients
3310 have a large amount of RAM, have many CPUs, the filesystem has
3311 many OSTs, the clients are rebooted frequently, and/or the files
3312 are not accessed by other nodes after being written.
3314 The presence of the problem can be detected by looking at
3315 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3316 smaller than the second (nr_unused) then dirty files will not
3317 be flushed automatically to disk. "sync; sleep 10" should be
3318 run several times on the node before unmounting it to update
3319 Lustre (this is also safe to run on nodes without this problem).
3321 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3322 that can cause this same problem, so customers using that kernel
3323 also need to update the kernel in addition to Lustre. In order
3324 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3326 It is normal that files written just before a client crash (less
3327 than 30s) may not yet have been flushed to disk, even for local
3331 Frequency : frequent on thin XT3 nodes
3333 Description: UUID collision on thin XT3 Linux nodes
3334 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3335 randomly, since we relied on an insufficiently-seeded PRNG.
3340 Description: OSS hangs after "All ost request buffers busy"
3341 Details : A deadlock between quota and journal operations caused OSS
3342 hangs after printing "All ost request buffers busy."
3345 Frequency : always on liblustre builds
3347 Description: Cleanup compiler warnings on liblustre
3350 Frequency : always on liblustre builds on XT3
3352 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3356 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3357 > 32000 subdirectories
3358 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3359 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3360 subdirectory count crosses 32000. This will aid e2fsck to
3361 correctly handle more than 32000 subdirectories.
3364 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3366 Description: System hang on startup
3367 Details : This bug allowed the liblustre (e.g. catamount) client to
3368 return to the app before handling all startup RPCs. This
3369 could leave the node unresponsive to lustre network traffic
3370 and manifested as a server ptllnd timeout.
3372 Severity : enhancement
3374 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3375 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3376 boolean to control whether to print peer debug info when a
3377 client's RPC times out.
3382 Description: make O_SYNC working on 2.6 kernels
3383 Details : 2.6 kernels use different method for mark pages for write,
3384 so need add a code to lustre for O_SYNC work.
3389 Description: Failure to close file and release space on NFS
3390 Details : Put inode details into lock acquired in ll_intent_file_open.
3391 Use mdc_intent_lock in ll_intent_open to properly
3392 detect all kind of errors unhandled by mdc_enqueue.
3397 Description: proc file read during shutdown sometimes raced obd removal,
3399 Details : Add lock to prevent obd access after proc file removal.
3402 Frequency : Only for files larger than 4GB on 32-bit clients.
3404 Description: improperly doing page alignment of locks
3405 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3406 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3411 Description: RPCs being resent when they shouldn't be
3412 Details : Some RPCs that should not be resent are being resent. This
3413 can cause inconsistencies in the RPC state machine. Do not
3414 resend such requests.
3417 Frequency : rare, only with NFS export
3419 Description: Crash on NFS re-export node
3420 Details : under very unusual load conditions an assertion is hit in
3421 ll_intent_file_open()
3424 Frequency : only if OST filesystem is corrupted
3426 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3427 Details : for a short time RPCs with bulk IO are in the replay list,
3428 but replay of bulk IOs is unimplemented. If the OST filesystem
3429 is corrupted due to disk cache incoherency and then replay is
3430 started it is possible to trip an assertion. Avoid putting
3431 committed RPCs into the replay list at all to avoid this issue.
3436 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3437 Details : Large single O_DIRECT read and write calls can fail to allocate
3438 a sufficiently large buffer to process the request. In case of
3439 allocation failure the allocation is retried with a smaller
3440 buffer and broken into smaller requests.
3442 Severity : enhancement
3444 Description: Add -o localflock option to simulate old noflock behaviour.
3445 Details : This will achieve local-only flock/fcntl locks coherentness.
3450 Description: versioning check is incomplete
3451 Details : Checking the version difference of client vs. server, report
3452 error if the gap is too big.
3457 Description: add support PG_writeback bit
3458 Details : add support for PG_writeback bit for Lustre, for more carefull
3459 work with page cache in 2.6 kernel. This also fix some deadlocks
3460 and remove hack for work O_SYNC with 2.6 kernel.
3462 Severity : enhancement
3464 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3465 Details : The uninit_groups feature works in conjunction with the kernel
3466 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3467 up the pass1 processing of e2fsck. This is a read-only feature
3468 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3469 mount filesystems that have had this feature enabled.
3471 Severity : enhancement
3473 Description: Improve multi-block allocation algorithm to avoid fragmentation
3474 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3475 allocation locality and avoid filesystem fragmentation.
3477 ------------------------------------------------------------------------------
3479 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3481 * Support for kernels:
3482 2.6.9-42.0.3.EL (RHEL 4)
3483 2.6.5-7.276 (SLES 9)
3484 2.4.21-47.0.1.EL (RHEL 3)
3485 2.6.12.6 vanilla (kernel.org)
3486 2.6.16.21-0.8 (SLES10)
3487 * Recommended e2fsprogs version: 1.39.cfs2-0
3489 * The backwards-compatible /proc/sys/portals symlink has been removed
3490 in this release. Before upgrading, please ensure that you change
3491 any configuration scripts or /etc/sysctl.conf files that access
3492 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3493 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3494 in advance of the upgrade on any system running Lustre 1.4.6 or
3495 newer, since /proc/sys/lnet was added in that version.
3496 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3500 Frequency : only when quota is used
3502 Description: avoid scanning export list for quota master
3503 Details : Change the algorithms to avoid scanning export list in order
3504 to improve the efficiency.
3507 Frequency : MDS failover only, very rarely
3509 Description: "went back in time" messages on mds failover
3510 Details : The greatest transno may be lost when the current operation
3511 finishes with an error (transno==0) and the client's last_rcvd
3512 record is over-written. Save the greatest transno in the
3513 mds_last_transno for this case.
3516 Frequency : always for specific kernels and striping counts
3518 Description: client may get "Matching packet too big" without ACL support
3519 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3520 when trying to access files in certain configurations. The
3521 clients should in fact be denied when mounting because they do
3522 not understand ACLs.
3525 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3527 Description: many clients connecting with IO in progress causes connect timeouts
3528 Details : Avoid synchronous journal commits to avoid delays caused by many
3529 clients connecting/disconnecting when bulk IO is in progress.
3530 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3531 OST_IO_PORTAL to avoid delays behind potentially many pending
3535 Frequency : occasionally with multiple writers to a single file
3537 Description: shared writes to file may result in wrong size reported by stat()
3538 Details : Allow growing of kms when extent lock is cancelled
3541 Frequency : always with random mmap IO to multi-striped file
3543 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3544 Details : Only if the hole is at the end of OST object so that kms is too
3545 small. Fix is to increase kms accordingly in ll_nopage.
3548 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3550 Description: writes to a missing object would leak memory on the OST
3551 Details : If there is an inconsistency between the MDS and OST filesystems,
3552 such that the MDS references an object that doesn't exist, writes
3553 to that object will leak memory due to incorrect cleanup in the
3554 error handling path, eventually running out of memory on the OST.
3559 Description: Creating too long symlink causes lustre errors
3560 Details : Check symlink and name lengths before sending requests to MDS.
3563 Frequency : only if flock is enabled (not on by default)
3565 Description: posix locks not released on fd closure on 2.6.9+
3566 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3567 kernels, this caused such locks not to be released on fd close and
3568 then assertions on fs unmount about still used locks.
3571 Frequency : MDS failover only, very rarely
3573 Description: clients may get ASSERTION(granted_lock != NULL)
3574 Details : When request was taking a long time, and a client was resending
3575 a getattr by name lock request. The were multiple lock requests
3576 with the same client lock handle and
3577 mds_getattr_name->fixup_handle_for_resent_request found one of the
3578 lock handles but later failed with ASSERTION(granted_lock != NULL).
3583 Description: handle->h_buffer_credits > 0, assertion failure
3584 Details : h_buffer_credits is zero after truncate, causing assertion
3585 failure. This patch extends the transaction or creates a new
3589 Frequency : NFS re-export or patchless client
3590 Bugzilla : 11179, 10796
3591 Description: Crash on NFS re-export node (__d_move)
3592 Details : We do not want to hash the dentry if we don't have a lock.
3593 But if this dentry is later used in d_move, we'd hit uninitialised
3594 list head d_hash, so we just do this to init d_hash field but
3595 leave dentry unhashed.
3598 Frequency : NFS re-export or patchless client
3600 Description: NFS exports has problem with symbolic link
3601 Details : lustre client didn't properly install dentry when re-exported
3602 to NFS or running patchless client.
3605 Frequency : NFS re-export or patchless client
3607 Description: Various nfs/patchless fixes.
3608 Details : fixes reuse disconected alias for lookup process - this fixes
3609 warning "find_exported_dentry: npd != pd",
3610 fix permission error with open files at nfs.
3611 fix apply umask when do revalidate.
3614 Frequency : occasional
3616 Description: Crash on NFS re-export node
3617 Details : calling clear_page() on the wrong pointer triggered oops in
3618 generic_mapping_read().
3621 Frequency : rarely, using O_DIRECT IO
3623 Description: unaligned directio crashes client with LASSERT
3624 Details : check for unaligned buffers before trying any requests.
3627 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3629 Description: stale data returned from RAID cache
3630 Details : If only a small amount of IO is done to the RAID device before
3631 reading it again it is possible to get stale data from the RAID
3632 cache instead of reading it from disk.
3635 Frequency : always for sles10 kernel
3637 Description: sles10 support
3638 Details : ll_follow_link: compile fixes and using of nd_set_link
3639 under newer kernels.
3642 Frequency : depends on arch, kernel and compiler version, always on sles10
3645 Description: recursive or deep enough symlinks cause stack overflow
3646 Details : getting rid of large stack-allocated variable in
3650 Frequency : depends on hardware
3652 Description: lustre write performance loss in the SLES10 kernel
3653 Details : the performance loss is caused by using of write barriers in the
3654 ext3 code. The SLES10 kernel turns barrier support on by
3655 default. The fix is to undo that change for ldiskfs.
3657 ------------------------------------------------------------------------------
3659 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3661 * Support for kernels:
3662 2.6.9-42.0.3EL (RHEL 4)
3663 2.6.5-7.276 (SLES 9)
3664 2.4.21-47.0.1.EL (RHEL 3)
3665 2.6.12.6 vanilla (kernel.org)
3669 Frequency : quota enabled and large files being deleted
3671 Description: releasing more than 4GB of quota at once hangs OST
3672 Details : If a user deletes more than 4GB of files on a single OST it
3673 will cause the OST to spin in an infinite loop. Release
3674 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3679 Description: statfs data retrieved from /proc may be stale or zero
3680 Details : When reading per-device statfs data from /proc, in the
3681 {kbytes,files}_{total,free,avail} files, it may appear
3682 as zero or be out of date.
3685 Frequency : systems with MD RAID1 external journal devices
3687 Description: lconf's call to blkid is confused by RAID1 journal devices
3688 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3689 of returning all block devices that match the journal UUID.
3692 Frequency : always, for aggregate stripe size over 4GB
3694 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3695 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3696 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3699 Frequency : always if "lfs find" used on a local file/directory
3701 Description: "lfs find" segfaults if used on a local file/directory
3702 Details : The case where a directory component was not specified wasn't
3703 handled correctly. Handle this properly.
3706 Frequency : always on ppc64
3708 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3709 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3710 perform endianness conversion.
3713 Frequency : rarely (truncate to non-zero file size after write under load)
3714 Bugzilla : 10730, 10687
3715 Description: Files padded with zeros to next 4K multiple
3716 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3717 it is possible that files that are truncated to a non-zero size
3718 immediately after being written are filled with zero bytes beyond
3719 the truncated size. No file data is lost.
3721 Severity : enhancement
3723 Description: Allow recovery/failover for liblustre clients.
3724 Details : liblustre clients were unaware of failover configurations until
3727 Severity : enhancement
3729 Description: user file locks should fail when not mounting with flock option
3730 Details : Set up an error-returning stub in ll_file_operations.lock field
3731 to prevent incorrect behaviour when client is mounted without
3732 flock option. Also, set up properly f_op->flock field for
3736 Frequency : always on ia64
3738 Description: "lfs df" loops on printing out MDS statfs information
3739 Details : The obd_ioctl_data was not initialized and in some systems
3740 this caused a failure during the ioctl that did not return
3741 an error. Initialize the struct and return an error on failure.
3744 Frequency : SLES 9 only
3746 Description: Error of copying files with lustre special EAs as root
3747 Details : Client side always return success for setxattr call for lustre
3748 special xattr (currently only "trusted.lov").
3751 Frequency : rarely on clusters with both ia64+i386 clients
3753 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3754 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3755 clients are doing concurrent writes to the same file it is
3756 possible that the ia64 clients may LASSERT because the OST
3757 extent locks are not PAGE_SIZE aligned. Ensure that grown
3758 locks are always aligned on the request boundary.
3761 Frequency : specific use, occasional
3763 Description: Overwriting in use executable truncates on-disk binary image
3764 Details : If one node attempts to overwrite an executable in use by
3765 another node, we now correctly return ETXTBSY instead of
3766 truncating the file.
3768 Severity : enhancement
3770 Description: Async OSC create to avoid the blocking unnecessarily.
3771 Details : If a OST has no remain object, system will block on the creating
3772 when need to create a new object on this OST. Now, ways use
3773 pre-created objects when available, instead of blocking on an
3774 empty osc while others are not empty. If we must block, we block
3775 for the shortest possible period of time.
3780 Description: chmod on Lustre root is propagated to other clients
3781 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3782 invalid by the follow_mount time.
3785 Frequency : liblustre clients only
3787 Description: Race in 'instant cancel' lock handling could lead to such locks
3788 never to be granted in case of SMP MDS
3789 Details : Do not destroy not yet granted but cbpending locks in
3793 Frequency : replay/resend of open
3795 Description: non null lock assetion failure in mds_intent_policy
3796 Details : Trying to replay/resend lockless open requests resulted in
3797 mds_open() returning 0 with no lock. Now it sets a flag if
3798 a lock is going to be returned.
3800 Severity : enhancement
3802 Description: Checksum enhancements
3803 Details : New checksum enhancements allow for resending RPCs that failed
3806 Severity : enhancement
3808 Description: Tunables on number of dirty pages in cacche
3809 Details : Allow to set limit on number of dirty pages cached.
3814 Description: client crash on unmount - lock still has references
3815 Details : In some error handling cases it was possible to leak a lock
3816 reference on a client while accessing a file. This was not
3817 harmful to the client during operation, but would cause the
3818 client to crash when the filesystem is unmounted.
3821 Frequency : specific case, rare
3823 Description: ETXTBSY on mds though file not in use by client
3824 Details : ETXTBSY is no longer incorrectly returned when attempting to
3825 chmod or chown a directory that the user previously tried to
3826 execute or a currently-executing binary.
3829 Frequency : extremely rare except on liblustre-based clients
3831 Description: Lustre space not freed when files are deleted
3832 Details : Clean up open-unlinked files after client eviction. Previously
3833 the unlink was skipped and the files remained as orphans.
3838 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3839 Details : In some cases it was possible to send a blocking callback to
3840 a client doing a glimpse, even though that client didn't get
3841 a lock granted. When the glimpse lock is cancelled on the OST
3842 the freed lock is left on the waiting list and corrupted the list.
3845 Frequency : all core dumps
3847 Description: Broke core dumps to lustre
3848 Details : Negative dentry may be unhashed if parent does not have UPDATE
3849 lock, but some callers, e.g. do_coredump, expect dentry to be
3850 hashed after successful create, hash it in ll_create_it.
3852 ------------------------------------------------------------------------------
3854 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3856 * Support for kernels:
3857 2.6.9-42.0.2.EL (RHEL 4)
3858 2.6.5-7.276 (SLES 9)
3859 2.4.21-40.EL (RHEL 3)
3860 2.6.12.6 vanilla (kernel.org)
3864 Frequency : always on RHEL 3
3866 Description: Number of open files grows over time
3867 Details : The number of open files grows over time, whether or not
3868 Lustre is started. This was due to a filp leak introduced
3869 by one of our kernel patches.
3871 ------------------------------------------------------------------------------
3873 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3875 * Support for kernels:
3876 2.6.9-42.EL (RHEL 4)
3877 2.6.5-7.276 (SLES 9)
3878 2.4.21-40.EL (RHEL 3)
3879 2.6.12.6 vanilla (kernel.org)
3884 Bugzilla : 5719, 9635, 9792, 9684
3885 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3886 Details : If a server is under heavy load and cannot reply to new
3887 connection requests before the client resends the (re)connect,
3888 the connection handling code can behave badly if two service
3889 threads are concurrently handing separate (re)connections from
3890 the same client. Add better locking to the connection handling
3891 code, and ensure that only a single connection will be processed
3892 for a given client UUID, even if the lock is dropped.
3894 Severity : enhancement
3896 Description: add TCP zero-copy support to kernel
3897 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3898 sends if the hardware supports scatter-gather and checksumming.
3899 This allows socklnd to do client-write and server-read more
3900 efficiently and reduce CPU utilization from skbuf copying.
3903 Frequency : only if NFS exporting from client
3905 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3906 Details : If setting attributes on a file created under NFS that had
3907 never been opened it would be possible to oops the client
3908 if the file had no objects.
3911 Frequency : always for liblustre
3913 Description: liblustre client does MDS+OSTs setattr RPC for each write
3914 Details : When doing a write from a liblustre client, the client
3915 incorrectly issued an RPC to the MDS and each OST the file was
3916 striped over in order to update the timestamps. When writing
3917 with small chunks and many clients this could overwhelm the MDS
3918 with RPCs. In all cases it would slow down the write because
3919 these RPCs are unnecessary.
3921 Severity : enhancement
3923 Description: allow number of MDS service threads to be changed at module load
3924 Details : It is now possible to change the number of MDS service threads
3925 running. Adding "options mds mds_num_threads={N}" to the MDS's
3926 /etc/modprobe.conf will set the number of threads for the next
3927 time Lustre is restarted (assuming the "mds" module is also
3928 reloaded at that time). The default number of threads will
3929 stay the same, 32 for most systems.
3934 Description: OST crash if filesystem is unformatted or corrupt
3935 Details : If an OST is started on a device that has never been formatted
3936 or if the filesystem is corrupt and cannot even mount then the
3937 error handling cleanup routines would dereference a NULL pointer.
3942 Description: NULL pointer deref in llap_from_page.
3943 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3944 incorrect) mapping if the page was truncated/reclaimed while it was
3945 searched for. Check for this condition and skip such pages when
3946 doing readahead. Introduce extra check to llap_from_page() to
3947 verify page->mapping->host is non-NULL (so page is not anonymous).
3950 Frequency : Sometimes when using sys_sendfile
3952 Description: "page not covered by a lock" warnings from ll_readpage
3953 Details : sendfile called ll_readpage without right page locks present.
3954 Now we introduced ll_file_sendfile that does necessary locking
3955 around call to generic_file_sendfile() much like we do in
3959 Frequency : with certain MDS communication failures at client mount time
3961 Description: NULL pointer deref after failed client mount
3962 Details : a client connection request may delayed by the network layer
3963 and not be sent until after the PTLRPC layer has timed out the
3964 request. If the client fails the mount immediately it will try
3965 to clean up before the network times out the request. Add a
3966 reference from the request import to the obd device and delay
3967 the cleanup until the network drops the request.
3970 Frequency : occasionally during client (re)connect
3972 Description: assertion failure during client (re)connect
3973 Details : processing a client connection request may be delayed by the
3974 client or server longer than the client connect timeout. This
3975 causes the client to resend the connection request. If the
3976 original connection request is replied in this interval, the
3977 client may trip an assertion failure in ptlrpc_connect_interpret()
3978 which thought it would be the only running connect process.
3981 Frequency : only with obd_echo servers and clients that are rebooted
3983 Description: kernel BUG accessing uninitialized data structure
3984 Details : When running an obd_echo server it did not start the ping_evictor
3985 thread, and when a client was evicted an uninitialized data
3986 structure was accessed. Start the ping_evictor in the RPC
3987 service startup instead of the OBD startup.
3989 Severity : enhancement
3990 Bugzilla : 10193 (patchless)
3991 Description: Remove dependency on various unexported kernel interfaces.
3992 Details : No longer need reparent_to_init, exit_mm, exit_files,
3993 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3996 Frequency : rare (only users of deprecated and unsupported LDAP config)
3998 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3999 Details : LDAP apparently contains 'lustreName' attributes instead of
4000 'name'. A simple remapping of the name is sufficient.
4003 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4005 Description: waiting_locks_callback trips kernel BUG if client is evicted
4006 Details : Running with the dump_on_timeout debug flag turned on makes
4007 it possible that the waiting_locks_callback() can try to dump
4008 the Lustre kernel debug logs from an interrupt handler. Defer
4009 this log dumping to the expired_lock_main() thread.
4011 Severity : enhancement
4013 Description: Support NFS exporting on 2.6 kernels.
4014 Details : Implement non-rawops metadata methods for NFS server to use without
4015 changing NFS server code.
4018 Frequency : very rare (synthetic metadata workload only)
4020 Description: two racing renames might cause an MDS thread to deadlock
4021 Details : Running the "racer" program may cause one MDS thread to rename
4022 a file from being the source of a rename to being the target of
4023 a rename at exactly the same time that another thread is doing
4024 so, and the second thread has already enqueued these locks after
4025 doing a lookup of the target and is trying to relock them in
4026 order. Ensure that we don't try to re-lock the same resource.
4029 Frequency : only very large systems with liblustre clients
4031 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4032 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4033 all OSTs in parallel. This allows the eviction of stale liblustre
4034 clients to proceed much faster than if they were done in series,
4035 and also offers similar improvements for other set_info RPCs.
4040 Description: excessive CPU usage during initial read phase on client
4041 Details : During the initial read phase on a client, it would agressively
4042 retry readahead on the file, consuming too much CPU and impacting
4043 performance (since 1.4.5.8). Improve the readahead algorithm
4044 to avoid this, and also improve some other common cases (read
4045 of small files in particular, where "small" is files smaller than
4046 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4051 Description: MDS crash when receiving packet with unknown intent.
4052 Details : Do not LBUG in unknown intent case, just return -EFAULT
4054 Severity : enhancement
4055 Bugzilla : 9293, 9385
4056 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4057 Details : Do not serialize getattr (non-intent version) and statfs.
4060 Frequency : occasional, when OST network is overloaded/intermittent
4062 Description: client evicted by OST after bulk IO timeout
4063 Details : If a client sends a bulk IO request (read or write) the OST
4064 may evict the client if it is unresposive to its data GET/PUT
4065 request. This is incorrect if the network is overloaded (takes
4066 too long to transfer the RPC data) or dropped the OST GET/PUT
4067 request. There is no need to evict the client at all, since
4068 the pinger and/or lock callbacks will handle this, and the
4069 client can restart the bulk request.
4072 Frequency : Always when mmapping file with no objects
4074 Description: client crashes when mmapping file with no objects
4075 Details : Check that we actually have objects in a file before doing any
4076 operations on objects in ll_vm_open, ll_vm_close and
4082 Description: Request leak when working with deleted CWD
4083 Details : Introduce advanced request refcount tracking for requests
4084 referenced from lustre intent.
4086 Severity : Enhancement
4088 Description: Cache open file handles on client.
4089 Details : MDS now will return special lock along with openhandle, if
4090 requested and client is allowed to hold openhandle, even if unused,
4091 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4092 closing files for every read/write openration.
4094 Severity : Enhancement
4096 Description: Cache open negative dentries on client when possible.
4097 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4098 negative dentries on lock revocation.
4103 Description: Remounting a client read-only wasn't possible with a zconf mount
4104 Details : It wasn't possible to remount a client read-only with llmount.
4106 Severity : enhancement
4107 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4108 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4109 MPICH is included to add Lustre-specific ADIO interfaces.
4110 This is based closely on the UFS ADIO layer and only differs
4111 in file creation, in order to allow the OST striping to be set.
4112 This is user-contributed code and not supported by CFS.
4117 Description: extended inode attributes (immutable, append-only) work improperly
4118 when 2.4 and 2.6 kernels are used on client/server or vice versa
4119 Details : Introduce kernel-independent values for these flags.
4121 Severity : enhancement
4124 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4125 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4126 granularity. Now, Fractional MB granularity is be supported,
4127 this is very useful for low memory system.
4129 Severity : enhancement
4131 Description: Getattr by fid
4132 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4133 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4134 have locks and avoids getattr rpc requests for it on every stat.
4137 Frequency : Always, for filesystems larger than 2TB
4139 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4140 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4141 filesystems larger than 2TB to be mounted with mballoc enabled.
4144 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4146 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4147 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4148 filesystems larger than 2TB is mounted then the kernel will
4149 silently corrupt the start of the filesystem. CONFIG_LBD is
4150 enabled for all CFS-supported kernels, but the possibility of
4151 this happening with a modified kernel config exists.
4153 Severity : enhancement
4155 Description: add client O_DIRECT support for 2.6 kernels
4156 Details : It is now possible to do O_DIRECT reads and writes to files
4157 in the Lustre client mountpoint on 2.6 kernel clients.
4159 Severity : enhancement
4161 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4162 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4163 parallel, not waiting for response from every OST before sending
4164 a rpc to the next OST.
4169 Description: setattr vs write race when updating file timestamps
4170 Details : Client processes that update a file timestamp into the past
4171 right after writing to the file (e.g. tar) it is possible that
4172 the updated file modification time can be reset to the current
4173 time due to a race between processing the setattr and write RPC.
4175 Severity : enhancement
4177 Description: Bring 'lfs find' closer in line with regular Linux find.
4178 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4179 -print0 options and obtains all the needed info through the lustre
4182 Severity : enhancement
4184 Description: support up to 1024 configured devices on one node
4185 Details : change obd_dev array from statically allocated to dynamically
4186 allocated structs as they are first used to reduce memory usage
4191 Description: Flush dirty partially truncated pages during truncate
4192 Details : Immediatelly flush partially truncated pages in filter_setattr,
4193 this way we completely avoid having any pages in page cache on OST
4194 and can retire ugly workarounds during writes to flush such pages.
4199 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4200 Details : protect inode from truncation within vfs_unlink() context
4201 just take a reference before calling vfs_unlink() and release it
4202 when parent's i_sem is free.
4207 Description: last_id value checked outside lock on OST caused LASSERT failure
4208 Details : If there were multiple MDS->OST object precreate requests in
4209 flight, it was possible that the OST's last object id was checked
4210 outside a lock and incorrectly tripped an assertion. Move checks
4211 inside locks, and discard old precreate requests.
4214 Frequency : always, if extents are used on OSTs
4216 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4217 Details : OSTs using the extents format would not zero the high 16 bits of
4218 the index physical block number. This is not a problem for any
4219 OST filesystems smaller than 16TB, and no kernels support ext3
4220 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4221 new/modified files) and can be fixed for existing filesystems
4222 with e2fsprogs-1.39-cfs1.
4227 Description: import connection selection may be incorrect if timer wraps
4228 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4229 import connections to be ignored if the 32-bit jiffies counter
4230 wraps. Use a 64-bit jiffies counter.
4233 Frequency : very large clusters immediately after boot
4235 Description: LNET request buffers exhausted under heavy short-term load
4236 Details : If a large number of client requests are generated on a service
4237 that has previously never seen so many requests it is possible
4238 that the request buffer growth cannot keep up with the spike in
4239 demand. Instead of dropping incoming requests, they are held in
4240 the LND until the RPC service can accept more requests.
4243 Frequency : Sometimes during replay
4245 Description: Assertion failure in ll_local_open after replay.
4246 Details : If replay happened on an open request reply before we were able
4247 to set replay handler, reply will become not swabbed tripping the
4248 assertion in ll_local_open. Now we set the handler right after
4249 recognising of open request
4252 Frequency : very rare
4254 Description: kernel reports "badness in vsnprintf"
4255 Details : Reading from the "recovery_status" /proc file in small chunks
4256 may cause a negative length in lprocfs_obd_rd_recovery_status()
4257 call to vsnprintf() (which is otherwise harmless). Exit early
4258 if there is no more space in the output buffer.
4260 Severity : enhancement
4262 Description: clear OBD RPC statistics by writing to them
4263 Details : It is now possible to clear the OBD RPC statistics by writing
4264 to the "stats" file.
4269 Description: Client mtime is not the same on different clients after utimes
4270 Details : In some cases, the client was using the utimes() syscall on
4271 a file cached on another node. The clients now validate the
4272 ctime from the MDS + OSTs to determine which one is right.
4277 Description: Inability to activate failout mode
4278 Details : lconf script incorrectly assumed that in pythong string's numeric
4279 value is used in comparisons.
4282 Frequency : always with multiple stripes per file
4284 Description: Inefficient object allocation for mutli-stripe files
4285 Details : When selecting which OSTs to stripe files over, for files with
4286 a stripe count that divides evenly into the number of OSTs,
4287 the MDS is always picking the same starting OST for each file.
4288 Return the OST selection heuristic to the original design.
4293 Description: mount failures may take full timeout to return an error
4294 Details : Under some heavy load conditions it is possible that a
4295 failed mount can wait for the full obd_timeout interval,
4296 possibly several minutes, before reporting an error.
4297 Instead return an error as soon as the status is known.
4299 Frequency : quota enabled and large files being deleted
4301 Description: releasing more than 4GB of quota at once hangs OST
4302 Details : If a user deletes more than 4GB of files on a single OST it
4303 will cause the OST to spin in an infinite loop. Release
4304 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4309 Description: statfs data retrieved from /proc may be stale or zero
4310 Details : When reading per-device statfs data from /proc, in the
4311 {kbytes,files}_{total,free,avail} files, it may appear
4312 as zero or be out of date.
4315 Frequency : systems with MD RAID1 external journal devices
4317 Description: lconf's call to blkid is confused by RAID1 journal devices
4318 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4319 of returning all block devices that match the journal UUID.
4322 Frequency : always, for aggregate stripe size over 4GB
4324 Description: assertion fails when trying to use 4GB stripe size
4325 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4326 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4329 Frequency : always on ppc64
4331 Description: the first write on an ext3 filesystem with mballoc got stuck
4332 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4333 perform endianness conversion.
4335 ------------------------------------------------------------------------------
4337 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4339 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4340 INTEROPERATE with older versions automatically. Please read the
4341 user documentation before upgrading any part of a live system.
4342 * WARNING: Lustre networking configuration changes are required with
4343 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4346 * Support for kernels:
4347 2.6.9-22.0.2.EL (RHEL 4)
4348 2.6.5-7.244 (SLES 9)
4349 2.6.12.6 vanilla (kernel.org)
4352 Severity : enhancement
4353 Bugzilla : 7981/8208
4354 Description: Introduced Lustre Networking (LNET)
4355 Details : LNET is new networking infrastructure for Lustre, it includes
4356 a reorganized network configuration mode (see the user
4357 documentation for full details) as well as support for routing
4358 between different network fabrics. Lustre Networking Devices
4359 (LNDs) for the supported network fabrics have also been
4360 created for this new infrastructure.
4362 Severity : enhancement
4363 Description: Introduced Access control lists
4364 Details : clients can set ACLs on files and directories in order to have
4365 more fine-grained permissions than the standard Unix UGO+RWX.
4366 The MDS must be started with the "-o acl" mount option.
4368 Severity : enhancement
4369 Description: Introduced filesystem quotas
4370 Details : Administrators may now establish per-user quotas on the
4373 Severity : enhancement
4375 Description: Configuration change for the XT3
4376 The PTLLND is now used to run Lustre over Portals on the XT3
4377 The configure option(s) --with-cray-portals are no longer used.
4378 Rather --with-portals=<path-to-portals-includes> is used to
4379 enable building on the XT3. In addition to enable XT3 specific
4380 features the option --enable-cray-xt3 must be used.
4385 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4386 Details : A race between allocating a new llcd and re-getting the llcd_lock
4387 allowed another thread to grab newly-allocated llcd.
4389 Severity : enhancement
4391 Description: 2.6 OST async journal commit and locking fix to improve performance
4392 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4393 2.6 kernels are now async as they already were in 2.4 kernels so
4394 that they can commit concurrently with the network bulk transfer.
4395 For block-allocated files the filter allocation semaphore is held
4396 to avoid filesystem fragmentation during allocation. BKL lock
4397 removed for 2.6 xattr operations where it is no longer needed.
4402 Description: lconf incorrectly determined whether two IP networks could talk
4403 Details : In some more complicated routing and multiple-network
4404 configurations, lconf will avoid trying to make a network
4405 connection to a disjoint part of the IP space. It was doing the
4406 math incorrectly for one set of cases.
4411 Description: Fix for potential infinite loop processing records in an llog.
4412 Details : If an llog record is corrupted/zeroed, it is possible to loop
4413 forever in llog_process(). Validate the llog record length
4414 and skip the remainder of the block on error.
4417 Frequency : occasional (liblustre only)
4419 Description: liblustre could not open files whose last component is a symlink
4420 Details : sysio_path_walk() would incorrectly pass the open intent to
4421 intermediate path components.
4424 Frequency : rare (liblustre only with non-standard tuning)
4425 Bugzilla : 7201 (7350)
4426 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4427 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4428 after it has been released, only do this for referenced locks
4430 Severity : enhancement
4432 Description: specifying an (invalid) directory default stripe_size of -1
4433 would reset the directory default striping
4434 Details : stripe_size -1 was used internally to signal directory stripe
4435 removal, now use "all default" to signal dir stripe removal
4436 as a directory striping of "all default" is not useful
4439 Frequency : common for large clusters running liblustre clients
4441 Description: doing an ls when liblustre clients are running is slow
4442 Details : sending a glimpse AST to a liblustre client waits for every AST
4443 to time out, as liblustre clients will not respond. Since they
4444 cannot cache data we refresh the OST lock LVB from disk instead.
4446 Severity : enhancement
4448 Description: doing an ls at the same time as file IO can be slow
4449 Details : enqueue and other "small" requests can be blocked behind many
4450 large IO requests. Create a new OST IO portal for non-IO
4451 requests so they can be processed faster.
4454 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4456 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4457 created file with mode 000
4458 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4459 allow an ftruncate() as a non-root user to a file with mode 000.
4460 Lustre now respects this flag to disable mode checks when
4461 truncating a file owned by the user
4464 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4467 Description: Revoking locks from clients that went dead or catatonic might take
4469 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4470 cancellation of such locks instant on servers without waiting for
4471 any reply from clients. Clients drops these locks when cancel
4472 notification from server is received without replying.
4475 Frequency : liblustre-only, when liblustre client dies or becomes busy
4477 Description: Doing ls on Linux clients can take a long time with active
4479 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4480 granting such locks to it in the first place if possible. Locks
4481 are taken by proxy on the OST during the read or write and
4482 dropped immediately afterward. Add connect flags handling, do
4483 not grant locks to liblustre clients for glimpse ASTs.
4485 Severity : enhancement
4487 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4488 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4489 sequential reads and behaves badly with applications doing random
4490 reads. Improve it by reading ahead at least read region, and
4491 avoiding excessive large RPC for small reads.
4493 Severity : enhancement
4495 Description: Creating more than 1000 files for a single job may cause a load
4496 imbalance on the OSTs if there are also a large number of OSTs.
4497 Details : qos_prep_create() uses an OST index reseed value that is an
4498 even multiple of the number of available OSTs so that if the
4499 reseed happens in the middle of the object allocation it will
4500 still utilize the OSTs as uniformly as possible.
4505 Description: OST or MDS may oops in ping_evictor_main()
4506 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4507 but doesn't restart at beginning of obd_exports_timed list
4510 Severity : enhancement
4512 Description: improve by-nid export eviction on the MDS and OST
4513 Details : allow multiple exports with the same NID to be evicted at one
4514 time without re-searching the exports list.
4517 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4519 Description: MDS may oops in groups_free()
4520 Details : in rare race conditions a newly allocated group_info struct is
4521 freed again, and this can be NULL. The 2.4 compatibility code
4522 for groups_free() checked for a NULL pointer, but 2.6 did not.
4525 Frequency : common for liblustre clients doing little filesystem IO
4526 Bugzilla : 9352, 7313
4527 Description: server may evict liblustre clients accessing contended locks
4528 Details : if a client is granted a lock or receives a completion AST
4529 with a blocking AST already set it would not reply to the AST
4530 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4533 Frequency : lfs setstripe, only systems with more than 160 OSTs
4535 Description: unable to set striping with a starting offset beyond OST 160
4536 Details : llapi_create_file() incorrectly limited the starting stripe
4537 index to the maximum single-file stripe count.
4540 Frequency : LDAP users only
4542 Description: lconf did not handle in-kernel recovery with LDAP properly
4543 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4545 Severity : enhancement
4547 Description: bind OST threads to NUMA nodes to improve performance
4548 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4549 node and do their allocations there to localize memory access
4551 Severity : enhancement
4553 Description: llmount can determine client NID directly from Myrinet (GM)
4554 Details : the client NID code from gmnalnid was moved directly into
4555 llmount, removing the need to use this or specifying the
4556 client NID explicitly when mounting GM clients with zeroconf
4559 Frequency : if client is started with down MDS
4561 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4562 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4563 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4564 ptlrpc_import_delayed_req() because we didn't check if the
4565 request was interrupted, and we also didn't break out of the
4566 event loop if there was no timeout
4571 Description: data loss during non-page-aligned writes to a single file from
4572 both multiple nodes and multiple threads on one node at same time
4573 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4574 inconsistency led to false short-reads, that were cached and later
4575 used by ->prepare_write() to fill in partially written page,
4576 leading to data loss.
4579 Frequency : always, if lconf --abort_recovery used
4581 Description: lconf --abort_recovery fails with 'Operation not supported'
4582 Details : lconf was attempting to abort recovery on the MDT device and not
4585 Severity : enhancement
4587 Description: remove cleanup logs
4588 Details : replace lconf-generated cleanup logs with lustre internal
4589 cleanup routines. Eliminates the need for client-cleanup and
4592 Severity : enhancement
4594 Description: add support for EAs (user and system) on lustre filesystems
4595 Details : it is now possible to store extended attributes in the Lustre
4596 client filesystem, and with the user_xattr mount option it
4597 is possible to allow users to store EAs on their files also
4599 Severity : enhancement
4601 Description: Add possibility (config option) to show minimal available OST free
4603 Details : When compiled with --enable-mindf configure option, statfs(2)
4604 (and so, df) will return least minimal free space available from
4605 all OSTs as amount of free space on FS, instead of summary of
4606 free spaces of all OSTs.
4608 Severity : enhancement
4610 Description: do not expand extent locks acquired on OST-side
4611 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4612 by server: they are not cached anyway.
4615 Frequency : when mmap is used/binaries executed from Lustre
4617 Description: Unmmap pages before throwing them away from read cache.
4618 Details : llap_shrink cache now attempts to unmap pages before discarding
4619 them (if unmapping failed - do not discard). SLES9 kernel has
4620 extra checks that trigger if this unmapping is not done first.
4625 Description: lconf didn't resolve symlinks before checking to see whether a
4626 given mountpoint was already in use
4629 Frequency : when migrating failover services
4630 Bugzilla : 6395, 9514
4631 Description: When migrating a subset of services from a node (e.g. failback
4632 from a failover service node) the remaining services would
4633 time out and evict clients.
4634 Details : lconf --force (implied by --failover) sets the global obd_timeout
4635 to 5 seconds in order to quickly disconnect, but this caused
4636 other RPCs to time out too quickly. Do not change the global
4637 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4639 Severity : enhancement
4640 Frequency : if MDS is started with down OST
4641 Bugzilla : 9439,5706
4642 Description: Allow startup/shutdown of an MDS without depending on the
4643 availability of the OSTs.
4644 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4645 Add appropriate locking and lov-osc refcounts for safe
4646 cleaning. Add osc abort_inflight calls in case the
4647 synchronize never started.
4650 Frequency : occasional (Cray XT3 only)
4652 Description: root not authorized to access files in CRAY_PORTALS environment
4653 Details : The client process capabilities were not honoured on the MDS in
4654 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4655 been accessed by an authorized user then root was able to access
4656 the file on the local client also. The root user capabilities
4657 are now allowed on the MDS, as this environment has secure UID.
4660 Frequency : occasional
4662 Description: ldiskfs "too long searching" message happens too often
4663 Details : A debugging message (otherwise harmless) prints too often on
4664 the OST console. This has been reduced to only happen when
4665 there are fragmentation problems on the filesystem.
4670 Description: Division by zero in statfs when all OSCs are inactive
4671 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4672 lov_statfs divides by value returned by lov_get_stripecnt().
4676 Bugzilla : 9489, 3273
4677 Description: First write from each client to each OST was only 4kB in size,
4678 to initialize client writeback cache, which caused sub-optimal
4679 RPCs and poor layout on disk for the first writen file.
4680 Details : Clients now request an initial cache grant at (re)connect time
4681 and so that they can start streaming writes to the cache right
4682 away and always do full-sized RPCs if there is enough data.
4683 If the OST is rebooted the client also re-establishes its grant
4684 so that client cached writes will be honoured under the grant.
4689 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4690 Details : Now I/O RPCs go to different portal number and (presumably) fast
4691 lock requests (and glimses) and other RPCs get their own service
4692 threads pool that should be able to service those RPCs
4695 Severity : enhancement
4697 Description: Ability to exchange lustre version between client and servers and
4698 issue warnings at client side if client is too old. Also for
4699 liblustre clients there is ability to refuse connection of too old
4701 Details : New 'version' field is added to connect data structure that is
4702 filled with version info. That info is later checked by server and
4706 Frequency : rare, liblustre only.
4707 Bugzilla : 9296, 9581
4708 Description: Two simultaneous writes from liblustre at offset within same page
4709 might proceed at the same time overwriting eachother with stale
4711 Details : I/O lock withing llu_file_prwv was released too early, before data
4712 actually was hitting the wire. Extended lock-holding time until
4713 server acknowledges receiving data.
4716 Frequency : extremely rare. Never observed in practice.
4718 Description: avoid generating lustre_handle cookie of 0.
4719 Details : class_handle_hash() generates handle cookies by incrementing
4720 global counter, and can hit 0 occasionaly (this is unlikely, but
4721 not impossible, because initial value of cookie counter is
4722 selected randonly). Value of 0 is used as a sentinel meaning
4723 "unassigned handle" --- avoid it. Also coalesce two critical
4724 sections in this function into one.
4726 Severity : enhancement
4728 Description: allow liblustre clients to delegate truncate locking to OST
4729 Details : To avoid overhead of locking, liblustre client instructs OST to
4730 take extent lock in ost_punch() on client's behalf. New connection
4731 flag is added to handle backward compatibility.
4733 Severity : enhancement
4734 Bugzilla : 4928, 7341, 9758
4735 Description: allow number of OST service threads to be specified
4736 Details : a module parameter allows the number of OST service threads
4737 to be specified via "options ost ost_num_threads={N}" in the
4738 OSS's /etc/modules.conf or /etc/modprobe.conf.
4742 Bugzilla : 6146, 9635, 9895
4743 Description: servers crash with bad pointer in target_handle_connect()
4744 Details : In rare cases when a client is reconnecting it was possible that
4745 the connection request was the last reference for that export.
4746 We would temporarily drop the export reference and get a new
4747 one, but this may have been the last reference and the export
4748 was just destroyed. Get new reference before dropping old one.
4750 Severity : enhancement
4751 Frequency : if client is started with failover MDS
4753 Description: Allow multiple MDS hostnames in the mount command
4754 Details : Try to read the configuration from all specified MDS
4755 hostnames during a client mount in case the "primary"
4758 Severity : enhancement
4760 Description: Stop sending data to evicted clients as soon as possible.
4761 Details : Check if the client we are about to send or are sending data to
4762 was evicted already. (Check is done every second of waiting,
4763 for which l_wait_event interface was extended to allow checking
4764 of exit condition at specified intervals).
4767 Frequency : rare, normally only when NFS exporting is done from client
4769 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4771 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4774 Frequency : always, if srand() is called before liblustre initialization
4776 Description: Liblustre uses system PRNG disturbing its usage by user application
4777 Details : Introduce internal to lustre fast and high-quality PRNG for
4778 lustre usage and make liblustre and some other places in generic
4779 lustre code to use it.
4781 Severity : enhancement
4782 Bugzilla : 9477, 9557, 9870
4783 Description: Verify that the MDS configuration logs are updated when xml is
4784 Details : Check if the .xml configuration logs are newer than the config
4785 logs stored on the MDS and report an error if this is the case.
4786 Request --write-conf, or allow starting with --old_conf.
4788 Severity : enhancement
4790 Description: Handle symlinks in the path when checking if Lustre is mounted.
4791 Details : Resolve intermediate symlinks when checking if a client has
4792 mounted a filesystem to avoid duplicate client mounts.
4797 Description: lconf can hit an error exception but still return success.
4798 Details : The lconf command catches the Command error exception at the top
4799 level script context and will exit with the associated exit
4800 status, but doesn't ensure that this exit status is non-zero.
4805 Description: failure of ptlrpc thread startup can cause oops
4806 Details : Starting a ptlrpc service thread can fail if there are a large
4807 number of threads or the server memory is very fragmented.
4808 Handle this without oopsing.
4811 Frequency : always, only if liblustre and non-default acceptor port was used
4813 Description: liblustre cannot connect to servers with non-default acceptor port
4814 Details : tcpnal_set_default_params() was not called and was therefore
4815 ignoring the environment varaible TCPNAL_PORT, as well as other
4816 TCPNAL_ environment variables
4821 Description: two objects could be created on the same OST for a single file
4822 Details : If an OST is down, in some cases it was possible to create two
4823 objects on a single OST for a single file. No problems other
4824 than potential performance impact and spurious error messages.
4828 Bugzilla : 5681, 9562
4829 Description: Client may oops in ll_unhash_aliases
4830 Details : Client dcache may become inconsistent in race condition.
4831 In some cases "getcwd" can fail if the current directory is
4837 Description: Inode refcounting problems in NFS export code
4838 Details : link_raw functions used to call d_instantiate without obtaining
4839 extra inode reference first.
4843 Bugzilla : 9942, 9903
4844 Description: Referencing freed requests leading to crash, memleaks with NFS.
4845 Details : We used to require that call to ll_revalidate_it was always
4846 followed by ll_lookup_it. Also with revalidate_special() it is
4847 possible to call ll_revalidate_it() twice for the same dentry
4848 even if first occurence returned success. This fix changes semantic
4849 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4850 reference on a request referred from the intent.
4851 ll_intent_release() then releases such a request.
4854 Frequency : rare, normally benchmark loads only
4856 Description: unlinked inodes were kept in memory on the client
4857 Details : If a client is repeatedly creating and unlinking files it
4858 can accumulate a lot of stale inodes in the inode slab cache.
4859 If there is no other client load running this can cause the
4860 client node to run out of memory. Instead flush old inodes
4861 from client cache that have the same inode number as a new inode.
4864 Frequency : SLES9 2.6.5 kernel and long filenames only
4865 Bugzilla : 9969, 10379
4866 Description: utime reports stale NFS file handle
4867 Details : SLES9 uses out-of-dentry names in some cases, which confused
4868 the lustre dentry revalidation. Change it to always use the
4872 Frequency : rare, unless heavy write-truncate concurrency is continuous
4873 Bugzilla : 4180, 6984, 7171, 9963, 9331
4874 Description: OST becomes very slow and/or deadlocked during object unlink
4875 Details : filter_destroy() was holding onto the parent directory lock
4876 while truncating+unlinking objects. For very large objects this
4877 may block other threads for a long time and slow overall OST
4878 responsiveness. It may also be possible to get a lock ordering
4879 deadlock in this case, or run out of journal credits because of
4880 the combined truncate+unlink. Solution is to do object truncate
4881 first in one transaction without parent lock, and then do the
4882 final unlink in a new transaction with the parent lock. This
4883 reduces the lock hold time dramatically.
4886 Frequency : rare, 2.4 kernels only
4888 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4889 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4890 before unmounting their filesystem in order to avoid deadlock.
4891 The kernel_locked() function in 2.4 kernels only checks whether
4892 the kernel lock is held, not whether it is this process that is
4893 holding it as 2.6 kernels do.
4898 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4899 Details : The client ptlrpc code may be trying to reconnect to a down
4900 server before a previous connection attempt has timed out.
4901 Increase the reconnect interval to be longer than the connection
4902 timeout interval to avoid sending duplicate connections to
4906 Frequency : echo_client brw_test command
4908 Description: fix echo_client to work with OST preallocated code
4909 Details : OST preallocation code (5137) didn't take echo_client IO path
4910 into account: echo_client calls filter methods outside of any
4911 OST thread and, hence, there is no per-thread preallocated
4912 pages and buffers to use. Solution: hijack pga pages for IO. As
4913 a byproduct, this avoids unnecessary data copying.
4917 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4918 Description: Client can oops in mdc_commit_close() after open replay
4919 Details : It was possible for the MDS to return an open request with no
4920 transaction number in mds_finish_transno() if the client was
4921 evicted, but without actually returning an error. Clients
4922 would later try to replay that open and may trip an assertion
4923 Simplify the client close codepath, and always return an error
4924 from the MDS in case the open is not successful.
4927 Frequency : rare, 2.6 OSTs only
4929 Description: OST may deadlock under high load on fragmented files
4930 Details : If there was a heavy load and highly-fragmented OST filesystems
4931 it was possible to have all the OST threads deadlock waiting on
4932 allocation of biovecs, because the biovecs were not released
4933 until the entire RPC IO was completed. Instead, release biovecs
4934 as soon as they are complete to ensure forward IO progress.
4936 Severity : enhancement
4938 Description: Support for specifying external journal device at mount
4939 Details : If an OST or MDS device is formatted with an external journal
4940 device, this device major/minor is stored in the ext3 superblock
4941 and may not be valid for failover. Allow detecting and
4942 specifying the external journal at mount time.
4947 Description: Mounting an MDS with pending unlinked files may cause oops
4948 Details : target_finish_recovery() calls mds_postrecov() which returned
4949 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4950 considers this an error and immediately begins cleaning up the
4951 lov, just after starting the mds_lov process
4953 Severity : enhancement
4955 Description: Implement 'lfs df' to report actual free space on per-OST basis
4956 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4957 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4958 usage of objects; '-h' to report in human readable format.
4960 ------------------------------------------------------------------------------
4962 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4969 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4970 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4971 at startup during orphan unlinking. Instead, skip buddy update
4972 before setup, as it will be regenerated after recovery is complete.
4977 Description: If an OST is inactive, its locks might reference stale inodes.
4978 Details : lov_change_cbdata() must iterate over all namespaces, even if
4979 they are inactive to clear inode references from the lock.
4981 Severity : enhancement
4982 Frequency : occasional, if non-standard max_dirty_mb used
4984 Description: Client will block write RPCs if not enough grant
4985 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4986 then the client will block writes while waiting for another RPC
4987 to complete instead of consuming its dirty limit. With change
4988 we get improved performance when max_dirty_mb is small.
4990 Severity : enhancement
4991 Bugzilla : 3389, 6253
4992 Description: Add support for supplementary groups on the MDS.
4993 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4994 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4995 lookups for user supplementary groups into a cache.
4999 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5000 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5001 for liblustre O_EXCL is enforced by MDS.
5005 Description: readdir never returns NULL in liblustre.
5006 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5007 dentry in struct dirent.
5011 Frequency : liblustre only, depends on application IO pattern
5012 Description: liblustre clients evicted if not contacting servers
5013 Details : Don't put liblustre clients into the ping_evictor list, so
5014 they will not be evicted by the pinger ever.
5016 Severity : enhancement
5018 Description: Add ability to evict clients by NID from MDS.
5019 Details : By echoing "nid:$NID" string into
5020 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5021 $NID would be instantly evicted from this MDS and from all active
5022 OSTs connected to it.
5026 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5027 Details : lookup_it_finish() used to query file size from OSTs that was not
5032 Description: service threads change working directory to that of init
5033 Details : Starting lustre service threads may pin the working directory
5034 of the parent thread, making that filesystem busy. Threads
5035 now change to the working directory of init to avoid this.
5039 Frequency : during shutdown only
5040 Description: shutdown with a failed MDS or OST can cause unmount to hang
5041 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5046 Frequency : 2.6 only
5047 Description: chmod/chown may include an extra supplementary group
5048 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5049 supplementary group and if none is specified this is used.
5052 Bugzilla : 5479 (6816)
5054 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5055 Details : If lookup is in progress on a file that is unlinked we might try
5056 to revalidate the inode and fail in revalidate after lookup is
5057 complete and ll_file_open() enqueues the open again but
5058 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5061 Frequency : always, if lconf --abort_recovery used
5063 Description: lconf --abort_recovery fails with 'Operation not supported'
5064 Details : lconf was attempting to abort recovery on the MDT device and not
5067 ------------------------------------------------------------------------------
5069 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5074 Frequency : rare (only unsupported configurations with a node running as an
5076 Bugzilla : 6514, 5137
5077 Description: Mounting a Lustre file system on a node running as an OST could
5079 Details : OSTs now preallocates memory needed to write out data at
5080 startup, instead of when needed, to avoid having to
5081 allocate memory in possibly low memory situations.
5082 Specifically, if the file system is mounted on on OST,
5083 memory pressure could force it to try to write out data,
5084 which it needed to allocate memory to do. Due to the low
5085 memory, it would be unable to do so and the node would
5086 become unresponsive.
5088 Severity : enhancement
5090 Description: Addition of lconf --service command line option
5091 Details : lconf now accepts a '--service <arg>' option, which is
5092 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5094 Severity : enhancement
5096 Description: Failover mode is now the default for OSTs.
5097 Details : By default, OSTs will now run in failover mode. To return to
5098 the old behaviour, add '--failout' to the lmc line for OSTs.
5100 Severity : enhancement
5102 Description: Health checks are now provided for MDS and OSTs
5103 Details : Additional detailed health check information on MSD and OSTs
5104 is now provided through the procfs health_check value.
5107 Frequency : occasional, depends on IO load
5109 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5110 after numerous create/delete cycles
5111 Details : The ext3 inode allocation policy would not allocate new inodes
5112 very well on the OSTs because there are no new directories
5113 being created. Instead we look for groups with free space if
5114 the parent directories are nearly full.
5119 Description: Network or server problems during mount may cause partially
5120 mounted clients instead of returning an error.
5121 Details : The config llog parsing code may overwrite the error return
5122 code during mount error handling, returning success instead
5128 Description: MDS can fail to allocate large reply buffers
5129 Details : After long uptimes the MDS can fail to allocate large reply
5130 buffers (e.g. zconf client mount config records) due to memory
5131 fragmentation or consumption by the buffer cache. Preallocate
5132 some large reply buffers so that these replies can be sent even
5133 under memory pressure.
5137 Frequency : rare (liblustre)
5138 Description: fsx running with liblustre complained that using truncate() to
5139 extend the file doesn't work. This patch corrects that issue.
5140 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5141 ATTR_SIZE and lsm use in llu_setattr_raw.
5145 Frequency : rare, only 2.6 kernels
5146 Description: Unusual file access patterns on the MDS may result in inode
5147 data being lost in very rare circumstances.
5148 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5149 code in the 2.6 kernel caused the inode and/or EA data not to
5150 be read from disk, causing single-file corruption.
5154 Frequency : rare, only 2.6 filesystems using extents
5155 Description: Heavy concurrent write and delete load may cause data corruption.
5156 Details : It was possible under high-load situations to have an extent
5157 metadata block in the block device cache from a just-unlinked
5158 file overwrite a newly-allocated data block. We now unmap any
5159 metadata buffers that alias just-allocated data blocks.
5163 Frequency : filesystems with default stripe_count larger than 77
5164 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5165 Details : lconf specifies an inode size of 4096 bytes when the default
5166 stripe_count is larger than 77. This conflicts with the default
5167 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5168 this case to avoid pinning too much memory for large EAs.
5170 ------------------------------------------------------------------------------
5172 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5177 Frequency : rare (extremely heavy IO load with hundreds of clients)
5179 Description: Client is evicted, gets IO error writing to file
5180 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5181 caused clients to be evicted for AST timeouts. The fixes in
5182 bug 5192 mean we no longer need to have such short AST timeouts
5183 so ldlm_timeout has been increased.
5186 Frequency : occasional during --force or --failover shutdown under load
5187 Bugzilla : 5949, 4834
5188 Description: Server oops/LBUG if stopped with --force or --failover under load
5189 Details : a collection of import/export refcount and cleanup ordering
5190 issues fixed for safer force cleanup
5193 Frequency : only filesystems larger than 120 OSTs
5194 Bugzilla : 5990, 6223
5195 Description: lfs getstripe would oops on a very large filesystem
5196 Details : lov_getconfig used kfree on vmalloc'd memory
5199 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5200 Bugzilla : 6242, 6243
5201 Description: reading from files that had been truncated to a non-zero size
5202 but never opened returned no data
5203 Details : ll_file_read() reads zeros from no-object files to EOF
5208 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5209 Details : There's little harm in aborting MDS/OSS recovery and letting it
5210 try again, so I removed the LASSERT and return an error instead.
5212 Severity : enhancement
5214 Description: New debugging infrastructure for tracking down data corruption
5215 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5216 (b) cover more of the client-side code path, and (c) try to narrow
5217 down where problems occurred
5221 Bugzilla : 3819, 4364, 4397, 6313
5222 Description: Racing close and eviction MDS could cause assertion in mds_close
5223 Details : It was possible to get multiple mfd references during close and
5224 client eviction, leading to one thread referencing a freed mfd.
5226 Severity: : enhancement
5227 Bugzilla : 3262, 6359
5228 Description: Attempts to reconnect to servers are now more aggressive.
5229 Details : This builds on the enhanced upcall-less recovery that was added
5230 in 1.4.2. When trying to reconnect to servers, clients will
5231 now try each server in the failover group every 10 seconds. By
5232 default, clients would previously try one server every 25 seconds.
5237 Description: After recovery, certain operations trigger a failed
5238 assertion on a client.
5239 Details : Failing over an mds, using lconf -d --failover, while a
5240 client was doing a readdir() call would cause the client to
5241 LBUG after recovery completed and the readdir() was resent.
5243 Severity : enhancement
5245 Description: Default groups are now added by lconf
5246 Details : You can now run lconf --group <servicename> without having to
5247 manually add groups with lmc.
5250 Frequency : occasional
5252 Description: Nodes with an elan id of 0 trigger a failed assertion
5255 Frequency : always when accessing e.g. tty/console device nodes
5257 Description: tty and some other devices nodes cannot be used on lustre
5258 Details : file's private_data field is used by device data and lustre
5259 values in there got lost. New field was added to struct file to
5260 store fs-specific private data.
5263 Frequency : when exporting Lustre via NFS
5265 Description: NFSD failed occasionally when looking up a path component
5266 Details : NFSD is looking up ".." which was broken in ext3 directories
5267 that had grown large enough to become hashed.
5270 Frequency : Clusters with multiple interfaces not on the same subnet
5272 Description: Nodes will repeatedly try to reconnect to an interface which it
5273 cannot reach and report an error to the log.
5274 Details : Extra peer list entries will be created by lconf with some peers
5275 unreachable. lconf now validates the peer before adding it.
5278 Frequency : Only if a default stripe is set on the filesystem root.
5280 Description: Setting a default stripe on the filesystem root prevented the
5281 filesystem from being remounted.
5282 Details : The client was sending extra request flags in the root getattr
5283 request and did not allocate a reply buffer for the dir EA.
5286 Frequency : occasional, higher if lots of files are accessed by one client
5287 Bugzilla : 6159, 6097
5288 Description: Client trips assertion regarding lsm mismatch/magic
5289 Details : While revalidating inodes the VFS looks up inodes with ifind()
5290 and in rare cases can find an inode that is being freed.
5291 The ll_test_inode() code will free the lsm during ifind()
5292 when it finds an existing inode and then the VFS later attaches
5293 this free lsm to a new inode.
5297 Bugzilla : 6422, 7030
5298 Description: MDS deadlock between mkdir and client eviction
5299 Details : Creating a new file via mkdir or mknod (starting a transaction
5300 and getting the ns lock) can deadlock with client eviction
5301 (gets ns lock and trying to finish a synchronous transaction).
5304 Frequency : occasional
5305 Description: While starting a server, the fsfilt_ext3 module could not be
5307 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5308 kernels. Previously, lconf would still use the ext3 name
5309 when trying to load modules. Now, it will correctly use
5310 ext3 on 2.4 and ldiskfs on 2.6.
5312 Severity : enhancement
5313 Description: The default stripe count has been changed to 1
5314 Details : The interpretation of the default stripe count (0, to lfs
5315 or lmc) has been changed to mean striping across a single
5316 OST, rather than all available. For general usage we have
5317 found a stripe count of 1 or 2 works best.
5319 Severity : enhancement
5320 Description: Add support for compiling against Cray portals.
5321 Details : Conditional compiling for some areas that are different
5325 Frequency : occasional
5326 Bugzilla : 6409, 6834
5327 Description: Creating files with an explicit stripe count may lead to
5328 a failed assertion on the MDS
5329 Details : If some OSTs are full or unavailable, creating files may
5330 trigger a failed assertion on the MDS. Now, Lustre will
5331 try to use other servers or return an error to the
5335 Frequency : occasional
5337 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5338 caused lock timeout during readahead (since 1.4.2).
5339 Details : Processes doing readahead might match a lock that hasn't been
5340 granted yet if there are overlapping and conflicting lock
5341 requests. The readahead process waits on ungranted lock
5342 (original lock is CBPENDING), while OST waits for that process
5343 to cancel CBPENDING read lock and eventually evicts client.
5345 Severity : enhancement
5347 Description: Initial enabling of flock support for clients
5348 Details : Implements fcntl advisory locking and file status functions.
5349 This feature is provided as an optional mount flag (default
5350 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5351 locking are implemented yet, and those that are are not guaranteed
5352 to be completely correct in production environments.
5353 mount -t lustre -o [flock|noflock] ...
5356 Frequency : occasional
5358 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5359 assertion in the ext3 JBD (journaling) layer.
5360 Details : The b_committed_data struct is protected by the big kernel lock
5361 in 2.4 kernels, serializing journal_commit_transaction() and
5362 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5363 access to this struct. In 2.6 kernels there is finer grained
5364 locking to improve SMP performance of the JBD layer.
5368 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5373 Description: merge and cleanup kernel patches.
5376 -----------------------------------------------------------------------------
5378 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5380 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5381 versions of Lustre. Please update all servers and clients to
5382 version 1.4.2 or later at the same time. You must also run
5383 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5384 to update the configuration logs.
5386 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5387 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5388 - swab llog records of type '0' so we get proper header size/idx (5861)
5389 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5390 - fix rename of one directory over another leaking an inode (5953)
5391 - avoid SetPageDirty on 2.6 (5981)
5392 - don't re-add just-being-destroyed locks to the waiting list (5653)
5393 - when creating new directories, inherit the parent's custom
5394 striping settings if present parent (3048)
5395 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5396 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5397 - don't hold client locks on temporary worklist from l_lru (5666)
5398 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5399 - automatically evict dead clients (5921)
5400 - Update file size properly in create+truncate+fstat case (6196)
5401 - Do not unhash mountpoint dentries, do not allow removal of
5403 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5404 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5405 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5406 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5407 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5408 - Correctly handle reads of files with no objects (6243)
5409 - lctl recover will also mark a device active if deactivate used (5933)
5411 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5412 - allow --write-conf on an MDS with different nettype than client (5619)
5413 - don't write config llogs to MDS for mounts not from that MDS (5617)
5414 - lconf should create multiple TCP connections from a client (5201)
5415 - init scripts are now turned off by default; run chkconfig --on
5416 lustre and chkconfig --on lustrefs to use them
5417 - upcalls are no longer needed for clients to recover to failover
5419 - add --abort-recovery option to lconf to abort recovery on device
5421 - add support for an arbitrary number of OSTs (3026)
5422 - Quota support protocol changes.
5423 - forward compatibility changes to wire structs (6007)
5424 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5425 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5426 - improved "lustre status" script
5427 - initialize blocksize for non-regular files (6062)
5428 - added --disable-server and --disable-client configure options (5782)
5429 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5430 - Vanilla 2.4.29 support
5431 - increase maximum number of obd devices to 520 (6242)
5432 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5433 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5434 SLES 9 2.6.5 kernels
5435 - limit stripes per file to 160 (the maximum EA size) (6093)
5437 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5440 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5441 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5442 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5443 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5444 - small ext3 extents cleanups and fixes (5733)
5445 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5446 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5447 - avoid needless client->OST connect, fix handle mismatch (5317)
5448 - fix DLM error path that led to out-of-sync client, long delays (5779)
5449 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5450 - fix several locking issues related to i_size (5492,5624,5654,5672)
5451 - don't move pending lock onto export if it is already evicted (5683)
5452 - fix kernel oops when creating .foo in unlinked directory (5548)
5453 - fix deadlock in obdfilter statistics vs. object create (5811)
5454 - use time_{before,after} to avoid timer jiffies wrap (5882)
5455 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5456 - Do not leak request if server was not able to process it (5154)
5457 - If mds_open unable to find parent dir, make that negative lookup(5154)
5458 - don't create new directories with extent-mapping (5909, 5936)
5460 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5461 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5462 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5463 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5464 - support for RHEL4 kernel on i686 (5773)
5465 - provide error messages when incompatible logs are encountered (5898)
5467 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5468 * version 1.4.0.10 (1.4.1 release candidate 1)
5470 - don't keep a lock reference when lock is not granted (4238)
5471 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5472 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5473 - drop import inflight refcount on signal_completed_replay error (5255)
5474 - unlock page after async write error during send (3677)
5475 - handle missing objects in filter_preprw_read properly (5265)
5476 - no transno return for symlink open, don't save no-trasno open (3440)
5477 - don't try to complete elan receive that already failed (4012)
5478 - free RPC server reply state on error (5406)
5479 - clean up thread from ptlrpc_start_thread() on error (5160)
5480 - readahead could read extra page into cache that wasn't ejected (5388)
5481 - prevent races in class_attach/setup/cleanup/detach (5260)
5482 - don't dereference de->d_inode after l_dput of de (5458)
5483 - use "int" for stripe value returned from lock_to_stripe (5544)
5484 - mballoc allocation and error-checking fixes in 2.6 (5504)
5485 - block device patches to fix I/O request sizes in 2.6 (5482)
5486 - look up hostnames for IB nals (5602)
5487 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5488 - don't start multiple acceptors for the same port (5277)
5489 - fix incorrect LASSERT in mds_getattr_name (5635)
5490 - export a proc file for general "ping" checking (5628)
5491 - fix "lfs check" to not block when the MDS is down (5628)
5493 - service request history (4965)
5494 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5495 - create an "evict_client" /proc entry on OSTs, like the MDS has
5496 - fix mount usage message, return errors per mount(8) (5168)
5497 - change grep [] to grep "[]" in tests so they work in more UMLs
5498 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5499 - remove ancient LOV_MAGIC_V0 EA support (5047)
5500 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5501 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5502 - allow readahead allocations to fail when low on memory (5383)
5503 - mmap locking landed again, after considerable improvement (2828)
5504 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5506 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5509 - send OST transaction number in read/write reply to free req (4966)
5510 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5511 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5512 - the watchdog thread now runs as interruptible (5246)
5513 - flock/lockf fixes (but it's still disabled, pending 5135)
5514 - don't use EXT3 constants in llite code (5094)
5515 - memory shortage at startup could cause assertion (5176)
5517 - reorganization of lov code
5518 - single portals codebase
5520 - add extents/mballoc support (5025)
5521 - direct I/O reads in the obdfilter (4048)
5522 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5524 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5527 - send OST transaction number in read/write reply to free req (4966)
5528 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5529 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5530 - the watchdog thread now runs as interruptible (5246)
5531 - handle missing objects in filter_preprw_read properly (5265)
5532 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5533 - drop import inflight refcount on signal_completed_replay error (5255)
5534 - unlock page after async write error during send (3677)
5535 - return original error code on reconstructed replies (3761)
5536 - no transno return for symlink open, don't save no-trasno open (3440)
5538 - add pid to ldlm debugging output (4922)
5539 - bump the watchdog timeouts -- we can't handle 30sec yet
5540 - extra debugging for orphan dentry/inode bug (5259)
5542 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5545 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5546 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5547 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5548 - fix NULL dereference in /proc/sys/portals/routes (4827)
5549 - allow failed mdc_close() operations to be interrupted (4561)
5550 - stop precreate on OST before MDS would time out on it (4778)
5551 - don't send partial-page writes before EOF from client (4410)
5552 - discard client grant for sub-page writes on large-page clients (4520)
5553 - don't free dentries not owned by NFS code, check generation (4806)
5554 - fix lsm leak if mds_create_objects() fails (4801)
5555 - limit debug_daemon file size, always print CERROR messages (4789)
5556 - use transno after validating reply (3892)
5557 - process timed out requests if import state changes (3754)
5558 - update mtime on OST during writes, return in glimpse (4829)
5559 - add mkfsoptions to LDAP (4679)
5560 - use ->max_readahead method instead of zapping global ra (5039)
5561 - don't interrupt __l_wait_event() during strace
5563 - add software watchdogs to catch hung threads quickly (4941)
5564 - make lustrefs init script start after nfs is mounted
5565 - fix CWARN/ERROR duplication (4930)
5566 - return async write errors to application if possible (2248)
5567 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5568 - print NAL number in %x format (4645)
5569 - update barely-supported suse-2.4.21-171 series (4842)
5570 - support for sles 9 %post scripts
5571 - support for building 2.6 kernel-source packages
5572 - support for sles km_* packages
5574 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5577 - ignore -ENOENT errors in osc_destroy (3639)
5578 - notify osc create thread that OSC is being cleaned up (4600)
5579 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5580 - reconstruct ost_handle() like mds_handle() (4657)
5581 - create a new thread to do import eviction to avoid deadlock (3969)
5582 - let lconf resolve symlinked-to devices (4629)
5583 - don't unlink "objects" from directory with default EA (4554)
5584 - hold socknal file ref over connect in case target is down (4394)
5585 - allow more than 32000 subdirectories in a single directory (3244)
5586 - fix blocks count for O_DIRECT writes (3751)
5587 - OST returns ENOSPC from object create when no space left (4539)
5588 - don't send truncate RPC if file size isn't changing (4410)
5589 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5590 - bind to privileged port in socknal and tcpnal (3689)
5592 - rate limit CERROR/CWARN console message to avoid overload (4519)
5593 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5594 - basic mmap support (3918)
5595 - kernel patch series update from b1_4 (4711)
5597 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5600 - avoid crash during MDS cleanup with OST shut down (2775)
5601 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5602 - don't use bad inodes on the MDS (3744)
5603 - dynamic object preallocation to improve recovery speed (4236)
5604 - don't hold spinlock over lock dumping or change debug flags (4401)
5605 - don't zero obd_dev when it is force cleaned (3651)
5606 - print grants to console if they go negative (4431)
5607 - "lctl deactivate" will stop automatic recovery attempts (3406)
5608 - look for existing locks in ldlm_handle_enqueue() (3764)
5609 - don't resolve lock handle twice in recovery avoiding race (4401)
5610 - revalidate should check working dir is a directory (4134)
5612 - don't always mark "slow" obdfilter messages as errors (4418)
5614 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5617 - don't close LustreDB during write_conf until it is done (3860)
5618 - fix typo in lconf for_each_profile (3821)
5619 - allow dumping logs from multiple threads at one time (3820)
5620 - don't allow multiple threads in OSC recovery (3812)
5621 - fix debug_size parameters (3864)
5622 - fix mds_postrecov to initialize import for llog ctxt (3121)
5623 - replace config semaphore with spinlock (3306)
5624 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5625 - don't allow enqueue to complete on a destroyed export (3822)
5626 - down write_lock before checking llog header bitmap (3825)
5627 - recover from lock replay timeout (3764)
5628 - up llog sem before sending rpc (3652)
5629 - reduce ns lock hold times when setting kms (3267)
5630 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5631 - fix NULL deref and obd_dev leak on setup error (3312)
5632 - replace some LBUG about llog ops with error handling (3841)
5633 - don't match INVALID dentries from d_lookup and spin (3784)
5634 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5635 - fix invalid assertion in ptlrpc_set_wait (3880)
5637 - add libwrap support for the TCP acceptor (3996)
5638 - add /proc/sys/portals/routes for non-root route listing (3994)
5639 - allow setting MDS UUID in .xml (2580)
5640 - print the stack of a process that LBUGs (4228)
5642 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5645 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5646 - make sure to unset replay flag from failed open requests (3440)
5647 - if default stripe count is 0, use OST count for inode size (3636)
5648 - update parent mtime/ctime on client for create/unlink (2611)
5649 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5650 - free recovery state on server during a forced cleanup (3571)
5651 - unregister_reply for resent reqs (3063)
5652 - loop back devices mounting and status check on 2.6 (3563)
5653 - fix resource-creation race that can provoke i_size == 0 (3513)
5654 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5655 - more debugging for page-accounting assertion (3746)
5656 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5657 - avoid deadlock after precreation failure (3758)
5658 - fix race and lock order deadlock in orphan handling (3450, 3750)
5659 - add validity checks when grabbing inodes from l_ast_data (3599)
5661 - add /proc/.../recovery_status to obdfilter (3428)
5662 - lightweight CDEBUG infrastructure, debug daemon (3668)
5663 - change default OSC RPC parameters to be better on small clusters
5664 - turn off OST read cache for files smaller than 32MB
5665 - install man pages and include them in rpms (3100)
5666 - add new init script for (un)mounting lustre filesystems (2593)
5667 - run chkconfig in %post for init scripts (3701)
5668 - drop scimac NAL (unmaintained)
5670 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5673 - clean kiobufs before and after use (3485)
5674 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5675 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5676 - change init script to fail more clearly if not run as root (1528)
5677 - allow clients to reconnect during replay (1742)
5678 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5679 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5680 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5681 - fix cond_resched (3554)
5682 - only evict unfinished clients after recovery (3515)
5683 - allow bulk resend, prevent data loss (3570)
5684 - dynamic ptlrpc request buffer allocation (2102)
5685 - don't allow unlinking open directory if it isn't empty (2904)
5686 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5687 - remove extraneous obd dereference causing LASSERT failure (3334)
5688 - don't use get_cycles() when creating temp. files on the mds (3156)
5689 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5690 - handle EEXIST for set-stripe, set proper directory name (3336)
5692 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5693 - fix ksocknal_fmb_callback() error messages (2918)
5695 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5698 - don't copy lvb into (possibly NULL) reply on error (2983)
5699 - don't deref dentry after dput, don't free lvb on error (2922)
5700 - use the kms to determine writeback rpc length (2947)
5701 - increment oti_logcookies when osc is inactive (2948)
5702 - update client's i_blocks count via lvb messages (2543)
5703 - handle intent open/close of special files properly (1557)
5704 - mount MDS with errors=remount-ro, like obdfilter (2009)
5705 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5706 - don't use cancelling-locks' kms values (2947)
5707 - use highest lock extent for kms, not last one (2925)
5708 - don't dereference ERR_PTR() dentry in error handling path (3107)
5709 - fix thread race in portals_debug_dumplog() (3122)
5710 - create lprocfs device entries at setup instead of at attach (1519)
5711 - common AST error handler, don't evict client on completion race (3145)
5712 - zero nameidata in detach_mnt in 2.6 (3118)
5713 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5714 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5715 - initialize RPC timeout timer earlier for 2.6 (3219)
5716 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5717 - print nal/nid for unknown nid (3258)
5718 - additional checks for oscc recovery before doing precreate (3284)
5719 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5720 - don't crash in mdc_close for bad permissions on open (3285)
5721 - zero i_rdev for non-device files (3147)
5722 - clear page->private before handing to FS, better assertion (3119)
5723 - tune the read pipeline (3236)
5724 - fix incorrect decref of invalidated dentry (2350)
5725 - provide read-ahead stats and refine rpc in flight stats (3328)
5726 - don't hold journal transaction open across create RPC (3313)
5727 - update atime on MDS at close time (3265)
5728 - close LDAP connection when recovering to avoid server load (3315)
5729 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5730 - don't leak open file on MDS after open resend (3325)
5731 - serialize filter_precreate and filter_destroy_precreated (3329)
5732 - loop device shouldn't call sync_dev() for nul device (3092)
5733 - clear page cache after eviction (2766)
5734 - resynchronize MDS->OST in background (2824)
5735 - refuse to mount the same filesystem twice on same mountpoint (3394)
5736 - allow llmount to create routes for mounting behind routers (3320)
5737 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5738 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5739 - fix rare problem with rename on htree directories (3417)
5741 - allow default OST striping configuration per directory (1414)
5742 - fix compilation for qswnal for 2.6 kernels (3125)
5743 - increase maximum number of MDS request buffers for large systems
5744 - change liblustreapi to be useful for external progs like lfsck (3098)
5745 - increase local configuration timeout for slow disks (3353)
5746 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5748 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5751 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5752 - don't overwrite extent policy data in reply if lock was blocked (2901)
5753 - drop filter export grants atomically with removal from device (2663)
5754 - del obd_self_export from work_list in class_disconnect_exports (2908)
5755 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5756 - swab reply message in mdc_close, other PPC fixes (2464)
5757 - fix destroying of named logs (2325)
5758 - overwrite old logs when running lconf --write_conf (2264)
5759 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5760 - fix race in target_handle_connect (2898)
5761 - mds_reint_create() should take same inode create lock (2926)
5762 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5763 - don't close files for self_export to avoid uninitialized obd (2936)
5764 - allow MDS with the same name as client node (2939)
5765 - hold dentry reference for closed log files for unlink (2325)
5766 - reserve space for all logs during transactions (2059)
5767 - don't evict page beyond end of stripe extent (2925)
5768 - don't oops on a deleted current working directory (2399)
5769 - handle hard links to targets without a parent properly (2517)
5770 - don't dereference NULL lock when racing during eviction (2867)
5771 - don't grow lock extents when lots of conflicting locks (2919)
5773 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5776 - account for cache space usage on clients to avoid data loss (974)
5777 - lfsck support in lustre kernel code (2349)
5778 - reduce journal credits needed for BRW writes (2370)
5779 - orphan handling to avoid losing space on client/server crashes
5780 - ptlrpcd can be blocked, stopping ALL progress (2477)
5781 - use lock value blocks to assist in proper KMS, faster stat (1021)
5782 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5783 - recovery for initial connections (2355)
5784 - fixes for mds_cleanup_orphans (1934)
5785 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5786 - block all file creations until orphan recovery completes (1901)
5787 - client remove rq_connection from request struct (2423)
5788 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5789 - recovery timer race (2670)
5790 - mdc_close recovey bug (2532)
5791 - ptlrpc cleanup bug (2710)
5792 - mds timeout on local locks (2588)
5793 - namespace lock held during RPCs (2431)
5794 - handle interrupted sync write properly (2503)
5795 - don't try to handle a message that hasn't been replied to (2699)
5796 - client assert failure during cleanup after abort recovery (2701)
5797 - leak mdc device after failed mount (2712)
5798 - ptlrpc_check_set allows timedout requests to complete (2714)
5799 - wait for inflight reqs when ptlrpcd finishes (2710)
5800 - make sure unregistered services are removed from the srv_list
5801 - reset bulk XID's when resending them (caught by 1138 test)
5802 - unregister_bulk after timeout
5803 - fix lconf error (2694)
5804 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5805 - readahead locks pages, leaves pending causing memory pressure (2673)
5806 - increase OST request buffers to 4096 on large machines (2729)
5807 - fix up permission of existing directories in simple_mkdir (2661)
5808 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5809 - don't assign transno to errored transactions (2742)
5810 - don't delete objects on OST if given a bogus objid from MDS (2751)
5811 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5812 - if rq_no_resend, then timeout request after recovery (2432)
5813 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5814 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5815 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5816 - MDS recovery shouldn't delete objects that it has given out (2730)
5817 - if enqueue arrives after completion, don't clobber LVB (2819)
5818 - don't unlock pages twice when trigger_group_io returns error (2814)
5819 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5820 - don't write pages to disk if there was an error (1450)
5821 - don't ping imports that have recovery disabled (2676)
5822 - take buffered bytes into account when balancing socknal conn (2817)
5823 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5824 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5825 - remove little-endian swabbing of llog records (1987)
5826 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5827 - retry reposting request buffers if they fail (1191)
5828 - grow extent at grant time to avoid granting a revoked lock (2809)
5829 - lock revoke doesn't evict page if covered by a second lock (2765)
5830 - disable VM readahead to avoid reading outside lock extents (2805)
5832 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5833 - updated kernel patches for hp-2.4.20 kernel (2681)
5835 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5838 - fix truncated write corruption (2366)
5839 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5841 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5842 - ensure len(uuid) < 37 in lmc (1171)
5843 - fix ia64 OOPS in llog_test (2255)
5844 - zero end of page at obdfilter for partial page writes (2648)
5845 - don't leave stale dentries around after renames (bug 2428)
5846 - fix timeouts when evicting a client with a single lock held (2642)
5847 - set deadline for the initial HELLO message to drain (2634)
5848 - print out dotted-quad IP addresses in the socknal (2302)
5850 - additional debugging for MDS client eviction problem (2443)
5851 - fix mkfsoptions support for osts (2603, 2604)
5853 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5856 - add series for the vanilla 2.6.0 kernel
5857 - add series for the vanilla 2.4.24 kernel
5858 - add series for a cray x86/64 UL kernel drop
5859 - fix xattr patches for the vanilla 2.4.19 series
5861 - generate true UUIDs in lmc (1171)
5862 - have portals stack dumping break in UML (2466)
5863 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5864 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5865 - wake callback waiting threads on client eviction (2460)
5866 - Add --ptldebug and --subsystem to lmc (1719)
5867 - update assertion to allow safe interrupt allocation
5868 - set rq_no_resend for cancel requests (2432)
5869 - recalculate ptlrpcd timeout after resend (2494)
5870 - call vfs_rmdir when removing pending directories (2368)
5871 - fix renaming a file to itself (2429)
5872 - lmc creates a default one-stripe lov (2454)
5873 - expand procfs space to handle large clusters (2326)
5874 - increase UML stack to avoid overflow
5875 - update lconf's list of debug and subsystem masks
5876 - fix lfs find --obd (2510)
5877 - /proc tunable for disabling filter read caching (2591)
5878 - stop rpm packages from altering slapd.conf (2301)
5879 - disable nagle in the socknal under 0conf (2578)
5880 - choose mds inode size based on stripe count (2572)
5881 - fix kernel-source rpm problems (2516)
5883 - add --disable-doc to avoid pdf generation (2421)
5884 - update documentation, tests, type-os, comments
5885 - avoid format warnings on ia64
5886 - remove the TOE NAL
5887 - tiny code cleanups by removing unused fields
5889 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5892 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5893 - remove the most blatant lies from BUILDING file (2371)
5894 - change default debug level to reasonable production setting
5895 - reduce client side cache size to reduce cache flush time
5896 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5897 - make TCP zerocopy and pinger support enabled by default (2476)
5898 - sync writes completed after process exits caused crashes (2319)
5899 - maintain correct mount count on the MDS (2356)
5900 - backout 1557, because 2316 wasn't really fixed
5901 - better file I/O statistics gathering in /proc
5902 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5903 - another kernel patch to fix zero-copy TCP function export
5904 - don't take duplicate lock when processing re-sent getattr (2420)
5905 - lctl uses obd_self_export instead of creating new conn (2353)
5906 - MDS/OST recovery case which requires object creation asserted (2425)
5907 - move lfs from /usr/sbin to /usr/bin in packages
5908 - fix race between mds_client_add and mds_client_free (2417)
5909 - use kmalloc instead of slabs in portals (2430)
5910 - don't create duplicate records when a failover MDS is present (2442)
5911 - remove unnecessary mount age check (2332)
5912 - don't remove directory inodes from locks prematurely (2451)
5913 - don't break if MDS service name is the same as hostname (2103)
5914 - fix races in client write RPC generation when cache full (2482)
5916 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5919 - remove now-unused request->rq_obd (278)
5920 - if an allocation fails, print out how much memory we've used (1933)
5921 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5922 - add the "configurable stack size" patch to most series files (1256)
5923 - ability to write large log records, for 100+ OST configs (2306)
5924 - fix NULL deref when filter_prep fails (2314)
5925 - fix operator precedence error in filter_sync
5926 - dynamic allocation of socknal TX descriptors (2315)
5927 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5928 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5929 - ensure that CWARN messages in Portals always get to the syslog
5930 - __init/__exit are not for prototype decls (ldlm_init/exit)
5931 - x86-64 compile warning fixes
5932 - fix gateway LMC keyword conflict (2318)
5933 - fix MDS lock inversions in getattr/reint paths (1844)
5934 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5935 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5936 - DLM race condition prevented some lock evictions (2328)
5937 - ENOMEM detection and retry on socknal sends (2230)
5938 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5939 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5940 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5941 - better error messages when a client is rejected during recovery (1505)
5942 - avoid cancelling locks which were never granted, after failure (2330)
5943 - fix i_sem/journal inversion in mds_client_add (2333)
5944 - fix truncate/getattr lock cycle deadlock (2334)
5945 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5946 - fix two rare exit paths which could leak an l_lock() ref (2321)
5947 - fencepost error in MDS/OST orphan recovery (2226)
5948 - make log record alignment 8 bytes (1988)
5949 - lstripe now fails when requested offset > ost_count (2237)
5950 - ensure that all kernel series have a complete list.h (1607)
5951 - fix crashes in special-file operations (2316)
5952 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5954 - allow configurable automake binary, for testing new versions
5955 - small update to the lfs documentation
5957 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5959 * fix negative export reference count in fsfilt_sync (2312)
5961 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5962 * release candidate 0.9.1
5964 - orphans are moved into the PENDING directory for possible recovery
5965 - replayed opens now open by fid for orphan/rename safety (1042)
5966 - last close of an orphan inode generates a transno (683)
5967 - chdir() and mount() now pin the directory entry (1020)
5968 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5969 - discard very old requests without processing them (1502)
5970 - remove some common, well-understood CERRORs (1505)
5971 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5972 - clear "grant" flags in OST replies until OST grant code lands (1644)
5973 - fix read performance by not clobbering i_blksize on client (1598)
5974 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5975 - make LRU size a /proc tunable, clears locks when reduced (707)
5976 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5977 - server locks take references on exports now (1558)
5978 - build fixes for 2.4.20-rh trees (1663)
5979 - return an error from lov_create if all OSCs are inactive (1751)
5980 - fix import levels when a reconnect happens without a timeout (1597)
5981 - exit early from mds_open if we get a lookup error (1749)
5982 - partial page read at EOF wouldn't wait for disk before sending (1642)
5983 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5984 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5985 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5986 - write/truncate lock inversion (1639)
5987 - Don't auto-load obdclass, portals modules during cleanup (1495)
5988 - fix timestamps from jumping to "now" (1763)
5989 - extra journal assertions (1648)
5990 - add an extra multiunlink test (1771)
5991 - fix read_record/write_record API (1776)
5992 - fix leak of offset_extent, possible incorrect i_size later (1772)
5993 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5994 - Debugging for the kqswnal_get_idle_tx problems (1820)
5995 - Allow recovery to be attempted multiple times (1536)
5996 - Write out MDS last_rcvd file after it is first created (1600)
5997 - Fix tx_descriptor leak in failed transmit situations (1827)
5998 - ext3 journaling fixes for assertion failure after IO error (1871)
5999 - class_export_put() on freed export after completion AST error (1896)
6000 - Fix revalidate looping in VFS (1322)
6001 - Don't access a freed export during MDS_REINT timeout (1521)
6002 - Add open-unlink recovery support on the MDS (1673,1764)
6003 - Return an error if no MDS data was read from last_rcvd (1946)
6004 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6005 - Don't setup a disk device that doesn't match exported UUID (317)
6006 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6007 - avoid committing NULL handle in force close
6008 - local.sh is now a one-stripe LOV configuration
6009 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6010 - fix ext3 htree duplicate directory entry corruption (1516)
6011 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6012 - update to new LOV EA format (2097)
6013 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6015 2003-06-15 Phil Schwan <phil@clusterfs.com>
6018 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6019 - per-import/export recovery handling (958, 931, 959)
6020 - multiple last-rcvd slots, for serving multiple FSes (949)
6021 - connections are again shared between multiple imp/exports (963, 964)
6022 - "umount -f" would hang if any requests needed to be sent (393, 978)
6023 - avoid pinning large req buffer by copying for queued messages (989)
6024 - add "uuid" to "lctl device" command to help upcalls (991)
6025 - "open" RPCs with transnos would confuse recovery counters (1037)
6026 - do proper endian conversion of all wire messages (288, 340, 891)
6027 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6028 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6029 - fix ptlrpc_connection leak in target_handle_connect (1174)
6030 - fix import refcounting bug in OST and MDS cleanup (1134)
6031 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6032 - fix very unlikely obd_types race condition (501)
6033 - remove osc_open hack for echo_client (1187)
6034 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6035 - a failure in read_inode2 leads to deadlock (1139)
6036 - cancel ack-locks as soon as transaction is committed (1072)
6037 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6038 - make sure to commitrw after any preprw to avoid deadlock (1162)
6039 - failing to execute a file in a lustre FS would lock inode (1203)
6040 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6041 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6042 - obd_uuid2tgt would walk off the end of the list (1255)
6043 - on IA64 the osc would give portals incorrect bulk size (1258)
6044 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6045 - fix lock inversion caused by new llite matching code (1282)
6046 - limit the number of dirty pages on a client to 10MB (1286)
6047 - timed out locks were not being corrected cancelled (1289)
6048 - fix O_DIRECT above 4GB on IA-32 (1292)
6049 * major user-visible changes
6050 - fail out/fail over policy now controlled by the upcall (993)
6052 - add OBD_PING to check server availability and failure (954)
6053 - lustre messages are now sent in sending host order (288, 340, 891)
6054 - add eadatalen to MDS getattr reply (340)
6055 - OST read replies may contain second buffer, with per-page status (593)
6057 2003-03-11 Phil Schwan <phil@clusterfs.com>
6060 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6061 - failed open()s could cause deadlock; fixed (867, 869)
6062 - stop cancelling OST locks when files are closed (481)
6063 - overlapping XID spaces caused network corruption (851, 853)
6064 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6065 - setattr_raw added, to do single-RPC, server-side setattrs
6066 - lmc/lconf syntax change for OST UUIDs
6067 - fix crashy race condition between ptlrpc_free_req and osc_close
6068 - don't use request in mdc_enqueue if we hit a timeout (889)
6069 - don't set the inode i_size for regular files from the MDS (896)
6070 - handle out of order completion AST (842)
6071 - don't LBUG if a lock request times out after receiving AST (913)
6072 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6073 - if a bad lock AST arrives, send an error instead of dropping entirely
6074 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6075 - fix leak in bulk IO when only partially completed (899, 900, 926)
6076 - fix O_DIRECT for ia64 (55)
6077 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6078 - C-z after timeout could hang a process forever; fixed (977)
6080 - client-side I/O cache (678, 924, 929, 941, 970)
6082 - READPAGE and SETATTRs which don't take server-side locks get
6085 2003-02-11 Phil Schwan <phil@clusterfs.com>
6088 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6089 - Fix fsfilt_extN_readpage() to read a full page of directory
6090 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6091 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6092 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6093 - Most intent operations were removed, in favour of a new RPC mode
6094 that does a single RPC to the server and bypasses most of the VFS
6095 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6096 - Aggressively cancel local locks on DLM servers
6097 - mds_reint_unlink sends EA to the client if it's the last nlink.
6098 client uses that EA to unlink OST objects.
6099 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6100 - recursive symlinks were fixed (439)
6101 - fixed NULL deref in DEBUG_REQ
6102 - filter_update_lastobjid no longer calls sync, which annoyed extN
6103 - fixed multi-client small-writes to a single file problem (445)
6104 - fixed mtime updates during file writes (607)
6105 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6106 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6107 - fixed Linux OST size reporting problem (444, 656)
6108 - OST now updates object mtime with writes or setattr (607, 619)
6109 - client verifies file size before zeroing page past EOF (445)
6110 - OST now writes last allocated objid to disk with allocation (108)
6111 - LOV on echo now works (409)
6113 - mds_reint_unlink sends a new buffer, with the EA included. this
6114 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6115 set if a regular file was being unlinked, and it was the last link
6116 - use PtlGet from the target for bulk writes (315)
6117 - OST now updates object mtime with writes or setattr (607, 619)
6118 - LDLM now has a grant-time callback to revalidate locked items, if
6120 - Many MDS operations were reorganized to combat race conditions
6122 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6123 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6125 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6128 - Fully reactivate OST imports after reconnection (512, others)
6129 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6130 - More graceful error handling for truncating on dead OST (515)
6131 - Don't error out unless we're actually accessing dead stripes (474)
6132 - Fix garbage sizes when stripes are missing (410)
6133 - LRU counters were broken, causing constant lock purge (433, 432)
6134 - garbage on read from stripes with failed OSTs (441)
6135 - mark OSCs as active before reconnecting during recovery (438)
6136 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6137 - lfind did not preserve OST order in output (443)
6138 - symlinks cause hung clients, incorrect data (439)
6139 - stop dereferencing request after dropping refcount (457)
6140 - don't LASSERT(spin_is_locked) on non-SMP (455)
6141 - fixes for many rename() bugs
6142 - fstat didn't correctly synchronize attributes (399)
6143 - server must handle lock cancellation during blocking AST prep (487)
6144 - bulk descriptors were free()d too soon (511)
6145 - fix paths in lconf, which would load incorrect modules (451, 507)
6146 - fix confusing lconf 'host not found' error message (386)
6147 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6148 - fix race condition in mdc_blocking_ast() for inode access (526)
6149 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6150 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6151 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6152 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6153 - fix lstripe in several ways (kernel side) (527)
6154 - fix request leak in ldlm_cli_enqueue (262)
6155 - incorrect OSC was marked inactive after OST failure
6156 - call mds_fs_cleanup before unmounting filesystem (524)
6157 - fix races between taking ns_lock and ldlm_lock_change_resource
6158 - fix races updating LOV export open file list
6159 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6160 - fix recovery NULL deref in ldlm_cli_cancel_unused
6161 - fix some DLM races by using new hash table for lock handles (419)
6162 - permit the client to specify desired inodes, at replay
6163 - duplicate requests when we queue them for replay reintegration
6164 - fix last_rcvd offset calculation
6165 - sync after each recovered transaction, so we always make progress
6166 - never, not always, ERESTART requests without transnos
6167 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6169 - skip replay if the MDS says that the client is already connected
6170 - don't check for a recovery-enabled export to match lctl's UUID
6171 - don't INC_USE_COUNT for phantom exports
6172 - don't crash when cleaning up phantom exports (567)
6173 - don't double-finish or set replay data for errored mdc_open requests
6174 - abort requests when they time out, so we don't get old replies
6175 - send/receive replies for AST messages again
6176 - if the client says that it doesn't have the lock, cancel it on the
6178 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6179 mark it as destroyed, it will all work out when decref is called
6180 - fix module use counts (22, 581)
6182 - ASTs now expect a reply (server cancels lock on error reply)
6184 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6187 - fix many simultaneous client startup (392)
6188 - fix dentry->d_it clobbering
6189 - credentials weren't being shipped for readdir/getattr operations
6190 - remove invalid assertions triggered during some concurrent MD
6192 - proper Lustre versions added (336, 389)
6193 - fix memory leak for create error case (398)
6194 - fix LOV locking bug that would get cli/srv out of sync
6195 - fix echo client over LOV (409)
6196 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6197 - fix double-O_EXCL intent crash (424)
6198 - avoid sending multiple lock CANCELs (352)
6200 - MDS can do multi-client recovery (modulo bugs in new code)
6202 - many updates, edits, cleanups
6204 2002-11-18 Phil Schwan <phil@clusterfs.com>
6207 - fix null d_it dereference (346)
6208 - fix full OST/dbench hang (333)
6209 - fix permission problem with file removal (286)
6210 - fix removal of OSCs from LOV when they fail
6211 - fix NULL deref during bulk timeout (214)
6212 - fix problems related to multiple filesystems on one MDS (241)
6213 - fixed serious subtle metadata locking bugs
6214 - free locks on clients when inodes are removed due to memory
6216 - fix inode pointer in lock data (285)
6217 - partial support for multiple MDS on a single host (241)
6218 - data locks weren't cancelled at clear_inode time (290, 311)
6219 - intent locks could lead to unbounded lock growth (205)
6220 - added a maximum lock count, an LRU list, and a flusher
6221 - fix multiple rename (365)
6222 - properly abstracted the echo client
6223 - OSC locked 1 byte too many; fixed
6224 - rewrote brw callback code:
6225 - fixed recovery bugs related to LOVs (306)
6226 - fixed too-many-pages-in-one-write crash (191)
6227 - fixed (again) crash in sync_io_timeout (214)
6228 - probably fixed callback-related race (385)
6230 - Add capability to MDS protocol
6231 - LDLM cancellations and callbacks on different portals
6233 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6236 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6237 - defer open object destruction to close time (601981, 138)
6238 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6239 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6240 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6241 - don't call obd_setattr in ll_file_release for destroyed objects
6242 * protocol change to lustre_msg: move |version| and add |flags|
6243 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6244 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6245 * added replay of create, unlink, link and rename operations during
6246 MDS failover; recovery should be much more robust now
6247 * remove failed OSCs from LOVs (only lov_create uses this so far)
6248 * the lustre-HOWTO was brought (more) up to date (582544)
6250 2002-10-23 Phil Schwan <phil@clusterfs.com>
6253 - in-use dentries weren't being reused properly (617851)
6254 - prevent multiple LDLM setup (599178)
6255 - fix LOV size calculations for truncate (617853)
6256 - fix client handling of MDS intent errors (POSIX)
6257 - fix permission bug in lovstripe.c test (624321)
6258 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6259 - truncate past end of file could corrupt data
6260 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6261 - a race in recovery could return ETIMEDOUT to apps (623947)
6262 - building outside the source directory was fixed
6263 * the lustre-HOWTO was brought (more) up to date (582544)
6264 * major progress was made on recovery functionality
6266 2002-10-10 Phil Schwan <phil@clusterfs.com>
6269 - recovery deadlock fix
6270 - rm -rf causes LBUG fix (617817)
6271 - file open by multiple tasks fix (618962)
6272 - directory permissions bugs (602707 and 620007)
6273 - journal_stop fixed with locking (611313)
6274 - O_APPEND failures resolved (618273, perhaps 614459)
6275 - lconf PATH fix (619770)
6276 - IA64 build fix (621450)
6277 - RPC buffer sizes scale with amount of memory
6279 2002-10-01 Phil Schwan <phil@clusterfs.com>
6282 - locks would be cancelled without throwing away data pages,
6283 resulting in inconsistent data (605627)
6284 - inode attributes were not always being refreshed (605627, 612449)
6285 - lconf now continues to cleanup after lctl reports an error
6286 - MDS now enforces user permissions (602707)
6287 - lprocfs cleanup fixed, but not yet enabled (614157)
6288 - fixed infinite server hang, should a client not respond to an AST
6289 - avoid going into recovery if user calls readlink() with a buffer
6290 that's too small (613941)
6291 - AST RPCs no longer require replies (614867) -- this may be changed
6292 - don't crash server if client sends an IOV that's too big (611336)
6293 - fixed lock conversion deadlock (611892)
6294 - fixed the following of symlinks (614622)
6295 * recovery: the server can remove locks from a client that dies, other
6296 clients can make progress
6297 * more extN patch fixes
6298 * compile-time configurable ptlrpc buffer allocations
6300 - collaborative read cache document
6301 - Lustre Lite Performance CDR document-in-progress
6303 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6306 - fix typo in patch-2.4.18
6308 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6311 - clear ptlrpc request each time in handle_incoming_request()
6312 - unlink of files now destroys the object on the OST
6314 2002-09-19 Peter Braam <braam@clusterfs.com>
6316 * add hard link support
6317 * change obdfile creation method
6318 * kernel patch changed
6320 2002-09-19 Peter Braam <braam@clusterfs.com>
6323 - stack overflow bug in extN fixed
6325 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6327 * documentation updates
6328 - add man pages for config tools
6329 - update tests/README to describe testing with new config tools
6330 - finish metadata API descriptions
6331 * bug fixes and cleanups
6332 - statfs workaround for 16TB limit
6333 - LOV stripe allocation improved, can stripe on subset of OSTs
6334 - LOV file size/IO offset was wrong for files > 4GB in size
6335 - object EA data was being dropped, caused files to be unreadable
6336 - memory overflow with non-LOV OST caused memory corruption
6337 - fixed regression tests to work with new config tools, obdfilter
6338 - fixed bug when directory size became larger than 1 block
6339 - fixed bug (for single client case) when PWD was deleted
6340 - invalidate local directory pages when doing intent-based ops
6341 - avoid LDLM oops when lock callback contained bad data
6343 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6345 * documentation updates
6346 * bug fixes and cleanups
6347 - configuration tools
6350 - 64-bit compile warnings
6351 - 64-bit internal statfs data
6353 * test_brw on persistent OST devices
6355 * lprocfs (disabled)
6357 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6359 * documentation updates
6360 * bug fixes and cleanups
6361 * configuration tools
6363 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6366 * many small fixes to 0_5_4
6367 * io/network handling
6368 * thinkos in MDS operations
6370 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6373 * crucial basic fixes to 0.5.3
6374 * IOR, Iozone work over Elan
6377 2002-08-07 Phil Schwan <phil@clusterfs.com>
6378 * version 0_5_3, our first alpha
6379 * we use the new Portals iovs
6380 * documentation updates
6381 * bug fixes and cleanups
6382 * small changes in the DLM wire protocol
6384 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6385 * version 0_5_1 with some initial stability,
6386 * locking on MD and file I/O.
6387 * documentation updates
6388 * several bug fixes since 0.5.0
6389 * small changes in wire protocol
6391 2002-07-18 Phil Schwan <phil@clusterfs.com>
6393 * delivered as Lustre Light Alpha
6394 * fixed a crash after handling invalid MDS requests
6395 * fixed directory pages for architectures with non-4k pages sizes
6397 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6398 * release version v0_4_4
6399 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6401 * Fixes a number of interruption problems with OST operations.
6402 * Update documentation for portals header changes
6403 * Move all wire protocol structs/defines to lustre_idl.h
6404 * Fixes symlink length bug.
6405 * Add tcpdump to repository.
6407 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6408 * release version v0_4_3
6409 * Fixes statfs for inodes on extN.
6410 * Fixes bug in runtests which would delete /etc/hosts.
6411 * Use 64-bit object IDs wherever possible (not into VFS though)
6412 Remove ost_get_info, which is unused by lustre, and out of date.
6414 2002-07-03 Peter Braam <braam@clusterfs.com>
6415 * release version v0_4_2 Fixes a lookup error (type not passed)
6416 * move forward to head of Portals
6417 * move forward to latest Lustre kernel
6419 2002-06-25 Peter Braam <braam@clusterfs.com>
6420 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob