3 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server known to build on patched kernels:
6 2.6.32-431.29.2.el6 (RHEL6.5)
7 3.0.101-0.35 (SLES11 SP3)
8 * Client known to build on unpatched kernels:
9 2.6.32-431.29.2.el6 (RHEL6.5)
10 3.0.101-0.35 (SLES11 SP3)
13 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
14 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
15 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
16 stack overflow. For more information, please refer to bugzilla 17630.
17 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
18 the re-exporting nodes to be mounted with "32bitapi" mount option
20 --------------------------------------------------------------------------------
22 07-30-2014 Intel Corporation
24 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
25 for currently supported client and server kernel versions.
26 * Server known to build on patched kernels:
27 2.6.32-431.20.3.el6 (RHEL6.5)
28 3.0.101-0.31 (SLES11 SP3)
29 * Client known to build on unpatched kernels:
30 2.6.32-431.20.3.el6 (RHEL6.5)
31 3.0.101-0.31 (SLES11 SP3)
34 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
35 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
36 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
37 stack overflow. For more information, please refer to bugzilla 17630.
38 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
39 the re-exporting nodes to be mounted with "32bitapi" mount option
41 --------------------------------------------------------------------------------
43 10-31-2013 Intel Corporation
45 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
46 * Server builds for kernels:
47 2.6.32-358.18.1.el6 (RHEL6.4)
48 * Client builds for unpatched kernels:
49 2.6.32-358.18.1.el6 (RHEL6.4)
50 3.0.92-0.8 (SLES11 SP3)
52 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
53 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
54 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
55 stack overflow. For more information, please refer to bugzilla 17630.
56 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
57 the re-exporting nodes to be mounted with "32bitapi" mount option
59 --------------------------------------------------------------------------------
61 05-30-2013 Intel Corporation
63 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
64 * Server builds for kernels:
65 2.6.32-358.6.2.el6 (RHEL6.4)
66 * Client builds for unpatched kernels:
67 2.6.32-358.6.2.el6 (RHEL6.4)
68 3.0.74-0.6.6 (SLES11 SP2)
70 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
71 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
72 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
73 stack overflow. For more information, please refer to bugzilla 17630.
74 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
75 the re-exporting nodes to be mounted with "32bitapi" mount option
77 --------------------------------------------------------------------------------
79 9-30-2012 Whamcloud, Inc.
81 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
82 * Server support for kernels:
83 2.6.32-279.5.1.el6 (RHEL6)
84 * Client support for unpatched kernels:
85 2.6.32-279.5.1.el6 (RHEL6)
87 2.6.18-238.19.1.el5 (RHEL5)
88 2.6.16.60-0.69.1 (SLES10)
89 * Recommended e2fsprogs version: 1.41.90.wc4
90 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
91 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
92 stack overflow. For more information, please refer to bugzilla 17630.
93 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
94 the re-exporting nodes to be mounted with "32bitapi" mount option
96 Severity : enhancement
98 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
100 Severity : enhancement
102 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
104 Severity : enhancement
106 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
108 Severity : enhancement
110 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
112 Severity : enhancement
114 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
116 --------------------------------------------------------------------------------
118 04-30-2012 Whamcloud, Inc.
120 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
121 * Server support for kernels:
122 2.6.32-220.4.2.el6 (RHEL6)
123 * Client support for unpatched kernels:
124 2.6.32-220.4.2.el6 (RHEL6)
126 2.6.18-238.19.1.el5 (RHEL5)
127 2.6.16.60-0.69.1 (SLES10)
128 * Recommended e2fsprogs version: 1.41.90.wc4
129 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
130 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
131 stack overflow. For more information, please refer to bugzilla 17630.
132 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
133 the re-exporting nodes to be mounted with "32bitapi" mount option
135 Severity : enhancement
137 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
139 Severity : enhancement
141 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
143 Severity : enhancement
145 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
147 Severity : enhancement
149 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
151 --------------------------------------------------------------------------------
153 09-30-2011 Whamcloud, Inc.
155 * Server support for kernels:
156 2.6.18-238.19.1.el5 (RHEL5)
157 2.6.32-131.6.1.el6 (RHEL6)
158 * Client support for unpatched kernels:
159 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
160 2.6.18-238.19.1.el5 (RHEL5)
161 2.6.32-131.6.1.el6 (RHEL6)
162 2.6.29.4-167.fc11 (FC11)
163 2.6.16.60-0.69.1 (SLES10)
165 * Recommended e2fsprogs version: 1.41.90.wc3
166 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
167 Lustre filesystem with 4K stack may cause a stack overflow. For more
168 information, please refer to bugzilla 17630.
169 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
170 mounted with "32bitapi" mount option
172 Severity : enhancement
173 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
175 Severity : enhancement
176 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
178 Severity : enhancement
179 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
182 Frequency : only with SLES10
184 Description: Use OFED "KMP" provided by Novell
185 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
186 Lustre is now built against this package. That means you need to
187 install the ofed-kmp package from Novell for the patchless client.
188 Note that the ofed-kmp that Novell ships may not exactly match the
189 kernel version but should still be compatible.
191 Severity : enhancement
193 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
194 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
196 --------------------------------------------------------------------------------
198 2010-07-15 Oracle, Inc.
200 * Server support for kernels:
201 2.6.18-164.11.1.el5 (RHEL 5)
202 2.6.18-164.11.1.0.1.el5 (OEL 5)
203 * Client support for unpatched kernels:
204 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
205 2.6.18-164.11.1.el5 (RHEL 5),
206 2.6.18-164.11.1.0.1.el5 (OEL 5)
207 2.6.16.60-0.42.8 (SLES 10),
209 2.6.29.4-167.fc11 (FC11)
210 * Recommended e2fsprogs version: 1.41.10-sun2
212 Severity : enhancement
214 Description: Distribute lfs_migrate script with lustre
215 Details : The lfs_migrate mentioned in the manual is now included in the
216 lustre tree in lustre/scripts and will distributed in lustre RPMs.
218 Severity : enhancement
220 Description: Failover nid list added to import proc file.
222 Severity : enhancement
224 Description: "som_preview" mount option required to use SOM
225 Details : The clients will need to use som_preview mount option
229 Description: Lustre does not do 1MB IOs to HW RAID
230 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
232 Severity : enhancement
234 Description: lfs quota output is not very convenient for awk/sed-parsing
235 Details : Some positions in lfs quota output table could be empty or
236 non-empty which made it hard to parse it with scripts, now
237 a dash is put instead of space where there is not supposed
242 Description: Update to OFED 1.5.1
243 Details : OFED 1.5.1 will be used in environments that don't use
244 vendor supplied OFED.
248 Description: Remove lnet. prefix from lctl params display
249 Details : Also renamed lnet.memused to lnet_memused to
254 Description: Use conf_param -d to remove permanant settings.
258 Description: Add list_param -R
259 Details : lctl list_param now supports recursive mode.
262 Bugzilla : 21829, 22312
263 Description: Fix llobdstat.
264 Details : 1. Now updating the write_bytes counter
265 2. Port script fixes from 1.8.
269 Description: decrease the usage of memory on clients.
270 Details : 1. On clients, recycle dentries and inodes unused.
271 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
276 Description: llobdstat fix and enhancement
277 Details : add a counter to set a limit to how many samples will be returned
278 fix a wildcard in the path to limit to obdfilter stats only
282 Description: lfs check only allowed for root
283 Details : Code cleanups and added a regresion test of non-root use of
288 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
289 Details : Fix hr->hr_index overload.
293 Description: Fix FIEMAP functionality.
297 Description: FC11 patchless client support added
301 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
302 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
306 Description: New acc-sm script and various test-framework enhancments.
310 Description: Previously landed patch was reverted.
314 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
318 Description: Add ldiskfs maxdirsize mount option.
322 Description: Fix mkfs.lustre support for 16TB LUNs.
326 Description: Print mmp_check_interval
327 Details : print mmp_check_interval and make it possible to abort
328 mount operation in case it takes too long.
332 Description: Add cascading_rw to lustre mpi tests.
336 Description: Store SOM data in EA.
337 Details : Also several SOM related cleanups.
341 Description: Add refcount to md_open_data
345 Description: Use refcount on lov_request_set to prevent memory corruption.
349 Description: Statahead: use parent dir's i_mutex for synchronization.
353 Description: loadgen improvements.
357 Description: Enable peer health detection in ksocklnd and ko2iblnd.
361 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
365 Description: Fix race on obd_devcie:md_stats
369 Description: Fix SLES 11 client builds.
370 Details : Resolved uninitalized pointer gcc warnings.
374 Description: Allow non-root access for "lfs check".
375 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
379 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
381 Severity : enhancement
383 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
384 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
388 Description: Sanity test_27p fails.
389 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
390 Sanity tests 27 were cleaned up and improved.
394 Description: Unify req format on client/servers
395 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
396 buffers and swabbers approach..
401 Description: drop join file stuff.
403 Severity : enhancement
405 Description: Add 2.6.27 and SLES11 (patchless client) support
409 Description: Fix in ptlrpc_expire_one_request() to print the signed time
412 Severity : enhancement
414 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
417 Severity : enhancement
419 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
421 Severity : enhancement
423 Description: correctly handle big reply message.
424 Details : send LNet event if reply is bigger then buffer and adjust this buffer
429 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
431 Severity : enhancement
433 Description: Add support for OEL5.
437 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
439 Severity : enhancement
441 Description: "weak-modules" support
442 Details : Implement "weak-modules" support which enables kernel modules
443 to be used with any kernel that implements the same kABI. In
444 order to achieve this modules are now installed in
445 /lib/modules/$(uname -r)/updates/kernel on all distributions.
449 Description: client could not reconnect to OST because of an active request.
450 Details : The client now retries to reconnect to the same server, if a
451 connect request failed with EBUSY or -EAGAIN. Test case 35b
452 conf-sanity.sh has been created.
456 Description: Don't allow make backward step on assignin osc next id.
457 Details : race between allocation next id and ll_sync thread can be cause
458 of set wrong osc next id and can be kill valid ost objects.
462 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
466 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
470 Description: Error handling in osc_statfs_interpret() has been improved.
471 Details : Check in osc_statfs_interpret() for EBADR.
475 Description: Do not send statfs() requests to OSTs disabled by administrator.
476 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
478 Severity : enhancement
480 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
482 Severity : enhancement
484 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
486 Severity : enhancement
488 Description: Enhancement for lfs(1) command to use numeric uid/gid.
492 Description: Improved cleanup procedure for conf-sanity test_46a.
496 Description: Do not start bulk transfer, if client tried to reconnect.
497 Details : Do not start bulk transfer, if client tried to reconnect,
498 but failed due to active rpcs. Check for exp_abort_active_req
499 has been implemented.
503 Description: Prevent inconsistences between linux and lustre mount structures.
504 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
505 Make the sleep interruptible.
507 Severity : enhancement
509 Description: provide server to client comms path
510 Details : server to client communications path via new LDLM_SET_INFO rpc
514 Description: truncate starts GFP_FS allocation under transaction causing deadlock
515 Details : ldiskfs_truncate calls grab_cache_page which may start page
516 allocation under an open transaction. This may lead to
517 calling prune_icache with consequent lustre reentrance.
521 Description: send correctly lsm on open replay
522 Details : MDS is trust to LSM size on replay open, but client can set wrong size
527 Description: enable client interface failover
528 Details : When a child reconnects from another NID, properly update export
529 nid hash position and ldlm reverse import.
531 Severity : enhancement
533 Description: Communicate OST degraded/readonly state via statfs to MDS
534 Details : Flags in the statfs returned from OSTs indicate whether the
535 OST is in a degraded RAID state, or if the filesystem has
536 turned read-only after a filesystem error is detected.
541 Description: don't panic if EPROTO was hit when reading symlink
542 Details : correctly handling request reference in error cases.
544 Severity : enhancement
546 Description: Add LustreNetLink, a kernel-userspace communcation path.
548 Severity : enhancement
550 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
555 Description: access to llog context before init.
556 Details : move handling CATALOGS file at osc layer and forbid access to llog
560 Frequency : rare, only when client is writing to OST on local node
562 Description: Avoid deadlock for local client writes
563 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
564 memory freeing context. This allows OST threads to set the
565 PF_MEMALLOC flag on task structures in order to allocate memory
566 from reserved pools and complete IO.
567 Use GFP_HIGHUSER for OST allocations for non-local client writes,
568 so that the OST threads generate memory pressure and allow
569 inactive pages to be reclaimed.
571 Severity : enhancement
573 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
578 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
579 Details : move ldlm namespace creation in setup phase to avoid grab
580 _lprocfs_lock with cli_sem held.
584 Description: Temporarily disable grant shrink.
585 Details : Disable the feature for debugging.
588 Frequency : only when formatting test filesystems
590 Description: Unable to run several mkfs.lustre on loop devices at the same time
591 Details : mkfs.lustre returns error 256 on the concurrent loop devices
592 formatting. The solution is to proper handle the error.
594 Severity : enhancement
596 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
598 Severity : enhancement
600 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
605 Description: resolve race between obd_disconnect and class_disconnect_exports
606 Details : if obd_disconnect will be called to already disconnected export he
607 forget release one reference and osc module can't unloaded.
609 Severity : enhancement
611 Description: Allow tuning service thread via /proc
612 Details : For each service a new
613 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
614 created that can be used to set min/max thread counts, and get the
615 current number of running threads.
620 Descriptoin: don't return error if have partially created objects for file.
621 Details : lov_update_create_set uses set->set_success as index for created
622 objects, so if some requests will be failed, they will have hole
623 at end of array and qos_shrink_lsm for allocate correct lsm.
625 Severity : enhancement
627 Description: Update OFED support to 1.4.1
632 Description: Reduce small size read RPC
633 Details : Set read-ahead limit for every file and only do read-ahead when
634 available read-ahead pages are bigger than 1M to avoid small size
637 Severity : enhancement
639 Description: move AT tunable parameters for more consistent usage
640 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
642 Severity : enhancement
644 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
645 Details : allow skip disconnected ost for send statfs request and hide error
651 Description: fix racy locking of mballoc block bitmaps causing BUG
652 Details : The locking of the mballoc buddy bitmap and the in-memory
653 block bitmap was using two different spin locks in some
654 cases. This made it possible to incorrectly access the
655 mballoc bitmap while another process was modifying it,
656 causing a sanity assertion to fail. While no on-disk corruption
657 was reported, there was some risk of this happening.
660 Frequency : rare, on llog test 6
662 Descriptoin: don't allow connect to already connected import
663 Details : allowing connect to already connected import is hide connecting problem.
668 Descriptoin: allow kill process which wait statahead result
669 Details : for some reasons 'ls' can stick in waiting result from statahead,
670 in this case need way for kill this process.
672 Severity : enhancement
674 Description: Add state history info file, enhance import info file
675 Details : Track import connection state changes in a new osc/mdc proc file;
676 add overview-type data to the osc/mdc import proc file.
678 Severity : enhancement
680 Description: MDS create should not wait for statfs RPC while holding DLM lock.
682 Severity : enhancement
684 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
689 Descriptoin: OST grant too much space to client even there are not enough space.
690 Details : Client will shrink its grant cache to OST if there are no write
691 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
692 this grant cache if there are already not enough avaible space
693 (left_space < total_clients * 32M).
696 Frequency : start MDS on uncleanly shutdowned MDS device
698 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
699 Details : stay in waiting mds<>ost recovery finished produce random bugs
700 due race between two ll_sync thread for one lov target. send
701 ACTIVATE event only if connect realy finished and import have
705 Frequency : rare, connect and disconnect target at same time
707 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
708 Details : don't call obd_disconnect under lov_lock. this long time
709 operation and can block ptlrpcd which answer to connect request.
714 Descriptoin: don't lose wakeup for imp_recovery_waitq
715 Details : recover_import_no_retry or invalidate_import and import_close can
716 both sleep on imp_recovery_waitq, but we was send only one wakeup
720 Frequency : always with long access acl
722 Descriptoin: mds can't pack reply with long acl.
723 Details : mds don't control size of acl but they limited by reint/getattr
726 Severity : enhancement
728 Description: Update to SLES10 kernel-2.6.16.60-0.33.
730 Severity : enhancement
732 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
735 Frequency : start MDS on uncleanly shutdowned MDS device
737 Descriptoin: aborting recovery hang on MDS
738 Details : don't throttle destroy RPCs for the MDT.
741 Frequency : on remount
743 Description: external journal device not working after the remount
744 Details : clear dev_rdonly flag for external journal devices in
750 Description: shutdown vs evict race
751 Details : client_disconnect_export vs connect request race.
752 if client will evicted at this time - we start invalidate
753 thread without referece to import and import can be freed
757 Frequency : rare, need acl's on inode.
759 Description: client can't handle ost additional correctly
760 Details : if ost was added after client connected to mds client can have
761 hit lnet_try_match_md ... to big messages to wide striped files.
762 in this case need teach client to handle config events about add
763 lov target and update client max ea size at that event.
765 Severity : enhancement
767 Description: Changelogs
768 Details : Changelogs are a lightweight mechanism to track filesystem
769 metadata and namespace changes. The changelog is recorded
770 permanently on the MDTs, and is periodically "consumed" / purged
771 when records are no longer needed.
773 Severity : enhancement
775 Description: compact fld format with extents
776 Details : Store range of seq rather than every seq in FLD. Seq
777 controller update FLD rather than clients. In Case of CMD, mdt0
778 has FLD, all other metadata server act as non persistent proxy
779 for FLD queries and cache fld entries in fld cache.
784 Description: don't skip ost target if they assigned to file
785 Details : Drop slow OSCs if we can, but not for requested start idx.
786 This means "if OSC is slow and it is not the requested
787 start OST, then it can be skipped, otherwise skip it only
788 if it is inactive/recovering/out-of-space.
792 Description: more cleanup in mds_lov
793 Details : not send LOV EA under replay, we can't know about they size at this
794 time. Don't allow client connect to mds before any ost connected,
795 for avoid problems with LOV EA size and returning EIO to client.
797 Severity : enhancement
799 Description: Interoperability at server side (Disk interoperability)
801 Severity : enhancement
803 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
805 Severity : enhancement
807 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
809 Severity : enhancement
811 Description: New client IO stack (CLIO).
813 Severity : enhancement
815 Description: Commit on sharing. Eliminate inter-client dependencies between
816 uncommitted transactions by doing transaction commits.
817 Thereby clients may recovery independently.
820 Frequency : Create a symlink file with a very long name
822 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
823 Details : If there is no extra space in the request for early cancels,
824 ldlm_req_handles_avail() returns 0 instead of a negative value.
826 Severity : enhancement
828 Description: Add /proc entry for import status
829 Details : The mdc, osc, and mgc import directories now have
830 an import directory that contains useful import data for debugging
833 Severity : enhancement
835 Description: Re-disable certain /proc logging
836 Details : Enable and disable client's offset_stats, extents_stats and
837 extents_stats_per_process stats logging on the fly.
840 Frequency : Only on FC kernels 2.6.22+
842 Description: oops in statahead
843 Details : Do not drop reference count for the dentry from VFS when lookup,
844 VFS will do that by itself.
846 Severity : enhancement
848 Description: Generic /proc file permissions
849 Details : Set /Proc file permissions in a more generic way to enable non-
850 root users operate on some /proc files.
854 Description: Hitting mdc_commit_close() ASSERTION
855 Details : Properly handle request reference release in
856 ll_release_openhandle().
860 Description: quota recovery deadlock during mds failover
861 Details : This patch includes att18982, att18236, att18237 in bz14840.
863 1. fix osts hang when mds does failover with quotaon
864 2. prevent watchdog storm when osts threads wait for the
869 Frequency : only patchless client
870 Description: add workaround for race between add/remove dentry from hash
872 Severity : enhancement
874 Description: Allow OST glimpses to return PW locks
878 Description: LBUG when llog conf file is full
879 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
883 Description: Prevent import from entering FULL state when server in recovery
887 Description: service mount cannot take device name with ":"
888 Details : Only when device name contains ":/" will mount treat it as
894 Description: replace ptlrpcd with the statahead thread to interpret the async
895 statahead RPC callback
899 Frequency : on recovery
900 Description: I/O failures after umount during fail back
901 Details : if client reconnected to restarted server we need join to recovery
902 instead of find server handler is changed and process self
903 eviction with cancel all locks.
905 Severity : enhancement
907 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
909 Severity : enhancement
911 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
913 Severity : enhancement
915 Description: Upcall on Lustre log has been dumped
916 Details : Allow for a user mode script to be called once a Lustre log has
917 been dumped. It passes the filename of the dumped log to the
918 script, the location of the script can be specified via
919 /proc/sys/lnet/debug_log_upcall.
924 Description: avoid idr_remove called for id which is not allocated.
925 Details : Move assigment s_dev for clustered nfs to end of initialization,
926 to avoid problem with error handling.
931 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
932 Details : When connection is reused this not moved from CONN_UNUSED_HASH
933 into CONN_USED_HASH and this prodice warning when put connection
934 again in unused hash.
937 Severity : enhancement
939 Description: File striping can now be set to use an arbitrary pool of OSTs.
941 Severity : enhancement
943 Description: Export bytes_read/bytes_write count on OSC/OST.
947 Description: Early reply size mismatch, MGC loses connection
948 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
949 the connect flags are properly negotiated.
954 Description: quotas are not honored with O_DIRECT
955 Details : all writes with the flag O_DIRECT will use grants which leads to
956 this problem. Now using OBD_BRW_SYNC to guard this.
960 Description: add quota statistics
961 Details : 1. sort out quota proc entries and proc code.
962 2. add quota statistics
964 Severity : enhancement
966 Description: enable quota support for HEAD.
970 Description: Properly propagate oinfo flags from lov to osc for statfs
971 Details : restore missing copy oi_flags to lov requests.
973 Severity : enhancement
975 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
976 : l_getgroups(8), lst(8), routerstat(8)
977 Details : included man pages for llobdstat(8), llstat(8),
978 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
980 Severity : enhancement
982 Description: configure's --enable-quota should check the
983 : kernel .config for CONFIG_QUOTA
984 Details : configure is terminated if --enable-quota is passed but
985 : no quota support is in kernel
989 Description: Remove portals compatibility
990 Details : Remove portals compatibility, not interoperable with releases
995 Description: Resolve device initialization race
996 Details : Prevent proc handler from accessing devices added to the
997 obd_devs array but yet be intialized.
999 Severity : enhancement
1001 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1003 Severity : enhancement
1005 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1010 Description: Using wrong pointer in osc_brw_prep_request
1011 Details : Access to array[-1] can produce panic if kernel compiled with
1012 CONFIG_PAGE_ALLOC enabled
1016 Description: Client runs out of low memory
1017 Details : Consider only lowmem when counting initial number of llap pages
1021 Description: *optional* service tags registration
1022 Details : if the "service tags" package is installed on a Lustre node
1023 When the filesystem is mounted, a local-node service tag will
1024 be created. See http://inventory.sun.com/ for more information
1025 about the Service Tags asset management system.
1029 Description: Kernel BUG tries to release flock
1030 Details : Lustre does not destroy flock lock before last reference goes
1031 away. So always drop flock locks when client is evicted and
1032 perform unlock regardless of successfulness of speaking to MDS.
1036 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1040 Description: sanity test 65a fails if stripecount of -1 is set
1041 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1046 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1047 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1050 Severity : enhancement
1052 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1056 Description: do not process already freed flock
1057 Details : flock can possibly be freed by another thread before it reaches
1058 to ldlm_flock_completion_ast.
1062 Description: LBUG during stress test
1063 Details : Need properly lock accesses the flock deadlock detection list.
1067 Description: oops in page fault handler
1068 Details : kernel page fault handler can return two special 'pages' in error
1069 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1073 Description: timeout with invalidate import.
1074 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1075 should be handled by ptlrpcd. This produce long age waiting and
1076 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1078 Severity : enhancement
1080 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1084 Description: enable MGS and MDT services start separately
1085 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1086 a co-located MGS without starting the MGS, which is a complement
1087 to 'nosvc' mount option.
1091 Frequency : after recovery
1092 Description: precreate to many object's after del orphan.
1093 Details : del orphan st in oscc last_id == next_id and this triger growing
1094 count of precreated objects. Set flag LOW to skip increase count
1095 of precreated objects.
1099 Frequency : rare, on clear nid stats
1100 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1101 Details : when clean nid stats sometimes try destroy live entry,
1102 and this produce panic in free.
1106 Description: Stack overflow during MDS log replay
1107 ease stack pressure by using a thread dealing llog_process.
1111 Description: wait until IO finished before start new when do lock cancel.
1112 Details : VM protocol want old IO finished before start new, in this case
1113 need wait until PG_writeback is cleared until check dirty flag and
1114 call writepages in lock cancel callback.
1116 Severity : enhancement
1118 Description: using special macro for print time and cleanup in includes.
1122 Description: mds_mfd_close() ASSERTION(rc == 0)
1123 Details : In mds_mfd_close(), we need protect inode's writecount change
1124 within its orphan write semaphore to prevent possible races.
1128 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1132 Frequency : rare, on shutdown ost
1133 Description: don't hit live lock with umount ost.
1134 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1135 use shrink_dcache_sb instead.
1139 Description: don't panic with use echo client
1140 Details : echo client pass NULL as client nid pointer and this produce null
1141 pointer dereference.
1145 Description: fix build on ppc32
1146 Details : compile code with -m64 flag produce wrong object file for ppc32.
1150 Description: add message levels for liblustreapi
1154 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1155 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1156 will result in an -EAGAIN and in the exit call path we call
1157 alloc_rr() which will with increasing aggressiveness attempt to
1158 aquire precreated objects on the minimum number of required OSCs.
1162 Description: Use old size assignment to avoid deadlock
1163 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1164 the scheduling while holding a spinlock. We do not need locking
1165 for size in ll_update_inode() because size is only updated from
1166 the MDS for directories or files without objects, so there is no
1167 other place to do the update, and concurrent access to such inodes
1168 are protected by the inode lock.
1172 Description: resolve "_IOWR redefined" build error on SLES10
1176 Description: dump the memory debugging after all modules are unloaded to
1177 suppress false negative in conf_sanity test 39
1179 Severity : enhancement
1181 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1186 Description: fix SLES kernel versioning
1187 Details : the kernel version for our SLES 10 kernel did not include a "-"
1188 before the "smp" at the end. while this was not a problem in
1189 general, it did mean that software trying to use the kernel
1190 version to try to detect a vendor specific kernel would fail.
1191 this was most evident by the OFED build scripts.
1195 Description: Don't update lov_desc members until making sure they are valid
1196 Details : When updating lov_desc members via proc fs, need fix their
1197 validities before doing the real update.
1201 Description: don't put request into delay list while invalidate in flight.
1202 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1203 import in flight. this produce timeout for invalidate and sometimes
1204 can cause stale data.
1208 Frequency : on ppc only
1209 Description: not convert ost objects for directory because it's not exist.
1210 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1214 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1216 Severity : enhancement
1218 Description: Update to support 2.6.22.14 vanilla kernel.
1222 Frequency : rare, on recovery
1223 Description: read procfs can produce deadlock in some situation
1224 Details : Holding lprocfs lock which send rpc can produce block for destroy
1225 obd objects and this also block reconnect with -EALREADY. This
1226 isn't fix all lprocfs bugs - but make it rare.
1228 Severity : enhancement
1230 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1233 Frequency : frequent on X2 node
1235 Description: mdc_set_open_replay_data LBUG
1236 Details : Set replay data for requests that are eligible for replay.
1240 Description: lustre_mgs: operation 101 on unconnected MGS
1241 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1242 MGC, and late on MGC cannot successfully connect to MGS and a lot
1243 of the error messages complaining that MGS is not connected.
1247 Frequency : on network error
1248 Description: panic with double free request if network error
1249 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1250 but it's true only for synchronus enqueue, for async enqueue
1251 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1254 Severity : enhancement
1256 Description: client-side metadata stat-ahead during readdir(directory readahead)
1257 Details : perform client-side metadata stat-ahead when the client detects
1258 readdir and sequential stat of dir entries therein
1261 Frequency : on start mds
1263 Description: Implement get_info(last_id) in obdfilter.
1266 Frequency : occasional
1268 Description: Correctly check stale fid, not start epoch if ost not support SOM
1269 Details : open with flag O_CREATE need set old fid in op_fid3 because
1270 op_fid2 was overwritten with new generated fid, but mds can answer
1271 with one of these two fids and both is not stale. Setattr
1272 incorrectly started an epoch and assume will be called
1273 done_writing, but without SOM done_writing ever being called.
1276 Frequency : rare, depends on device drivers and load
1278 Description: MDS or OSS nodes crash due to stack overflow
1279 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1280 In some cases, in conjunction with device drivers that use a lot
1281 of stack the MDS (or possibly OSS) service threads could overflow
1282 the stack. One change which was identified to consume additional
1283 stack has been reworked to avoid the extra stack usage.
1286 Frequency : occasional
1288 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1289 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1290 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1291 oscc_has_objects and see OSC has no objests and call
1292 oscc_internal_create to resend create request.
1294 Severity : enhancement
1296 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1298 Severity : enhancement
1300 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1303 Frequency : very rare
1305 Description: Fix replay if there is an un-replied request and open
1306 Details : In some cases, older replay request will revert the
1307 mcd->mcd_last_xid on MDS which is used to record the client's
1308 latest sent request.
1310 Severity : enhancement
1312 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1314 Severity : enhancement
1316 Description: Add rhel5 support to HEAD.
1318 Serverity : enhancement
1320 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1325 Description: Don't allow skipping OSTs if index has been specified.
1326 Details : Don't allow skipping OSTs if index has been specified, make
1327 locking in internal create lots better.
1331 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1332 Details : access to bitfield in structure is always rounded to long
1333 and this produce problem with not atomic change any bit.
1337 Description: Lustre make rpms failed.
1338 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1339 builds Lustre rpms from tarball.
1342 Frequency : rare on shutdown ost
1344 Description: If llog cancel was not send before clean_exports phase, this can
1345 produce deadlock in llog code.
1346 Details : If llog thread has last reference to obd and call class_import_put
1347 this produce deadlock because llog_cleanup_commit_master wait when
1348 last llog_commit_thread exited, but this never success because was
1349 called from llog_commit_thread.
1353 Description: allow userland application know is lost one of stripes.
1354 Details : fill lvb_blocks with error code on ost and return it to
1355 application if error flag found.
1359 Description: NULL lov_tgts causing MDS oops
1360 Details : more safe checks for NULL lov_tgts for avoid oops.
1362 Severity : enhancement
1364 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1368 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1369 Details : If an OST is down the MDS will hang indefinitely in
1370 obd_statfs_rqset() waiting for the statfs data. While for
1371 MDS QOS usage of statfs, it should not stuck in waiting.
1373 Severity : enhancement
1375 Description: remote_acl support
1376 Details : Support ACL-based permission check for remote user.
1377 Support setfacl/getfacl for remote user with the utils
1378 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1379 the system "{s,g}etfacl" utils.
1381 Severity : enhancement
1383 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1385 Severity : enhancement
1387 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1391 Description: make mgs_setparam() handle fsname containing dash
1392 Details : fsname containing a dash does not work with lctl conf_param
1394 Severity : enhancement
1396 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1398 Severity : enhancement
1400 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1404 Frequency : rare, at shutdown
1405 Description: access already free / zero obd_namespace.
1406 Details : if client_disconnect_export was called without force flag set,
1407 and exist connect request in flight, this can produce access to
1408 NULL pointer (or already free pointer) when connect_interpret
1409 store ocd flags in obd_namespace.
1413 Frequency : only at startup
1414 Description: not alloc memory with spinlock held.
1415 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1416 if any spinlock held.
1418 Severity : enhancement
1420 Description: make lustre randomly fail allocating memory
1421 Details : Make lustre randomly failed allocating memory for testing purpose.
1423 Severity : enhancement
1425 Description: lost problems with lov objid file
1426 Details : Fixes some scability and access to not inited memory problems
1427 in work with lov objdid file.
1432 Description: lfs find does not continue on file error
1433 Details : Continue other files processing when a file/dir is absent.
1437 Description: Inconsistent usage of lustre_pack_reply()
1438 Details : Standardize the usage of lustre_pack_reply() such that it
1439 always generate a CERROR on failure.
1442 Frequency : occasional
1444 Description: MDS hang or stay in waiting lock
1445 Details : If client receive lock with CBPENDING flag ldlm need send lock
1446 cancel as separate rpc, to avoid situation when cancel request
1447 can't processed due all i/o threads stay in wait lock.
1451 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1455 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1459 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1463 Description: Update to RHEL4 latest kernel.
1465 Severity : enhancement
1467 Description: Build SLES10 patchless client fails
1468 Details : The configure was broken by run ./configure with
1469 --with-linux-obj=.... argument for patchless client. When the
1470 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1471 can't search header adequately. Use absolute path such as
1472 -I($LINUX)/include instead.
1476 Description: interrupt oig_wait produce painc on resend.
1477 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1478 private set, and this produce situation when rq_ptlrpcd_data not
1479 copyed to new allocated request and triggered LBUG on assert
1480 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1481 ptlrpcd set if request is changed and can be safety changed to use
1486 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1487 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1488 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1489 these macros and replace them with their difinitions instead.
1493 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1494 Details : Update which_patch & target file for SLES10 latest kernel.
1497 Description: Remove client patches from SLES 10 kernel.
1498 Details : This causes SLES 10 clients to behave as patchless clients
1499 even on a Lustre-patched (server) kernel.
1501 Severity : enhancement
1503 Description: self-adjustable client's lru lists
1504 Details : use adaptive algorithm for managing client cached locks lru
1505 lists according to current server load, other client's work
1506 pattern, memory activities, etc. Both, server and client
1507 side namespaces provide number of proc tunables for controlling
1510 Severity : enhancement
1512 Description: light-weight GSS support
1513 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1514 and reduce performance overhead.
1516 Severity : enhancement
1518 Description: Linux keyring support
1519 Details : Support using service of Linux keyring for Lustre GSS internal
1520 context refresh/cache mechanism.
1524 Description: Fix errors in lfs documentation
1525 Details : Fixes man pages
1529 Description: don't use GFP_* in generic Lustre code.
1530 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1534 Description: obdclass is limited by single OBD_ALLOC(idarray)
1535 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1539 Description: warnings with build patchless client with vanila 2.6.19 and up
1540 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1541 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1545 Frequency : Only for SLES
1546 Description: sanity_quota fail test_1
1547 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1548 file, which makes TSTID[2] inunique.
1552 Frequency : Only for SLES9
1553 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1554 Details : Update target/ChangeLog/which_patch .
1558 Frequency : Only for test_10 in sanity-quota.sh
1559 Description: a bug in quota test code
1560 Details : checking a test flag in a wrong place
1564 Frequency : Only for test_13 in sanity-quota.sh
1565 Description: a bug in quota test script
1566 Details : shouldn't check overfull allocated quota
1570 Description: jbd statistics
1571 Details : Port older jbd statistics patch for sles10
1573 Severity : enhancement
1575 Description: Add printing inode info into message about error in writepage.
1580 Description: reply_lock_interpret crash due to race with it and lock cancel.
1581 Details : Do not replay locks that are being cancelled. Do not reference
1582 locks by their address during replay, just by their handle.
1586 Frequency : When flocks are used.
1587 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1588 Details : Flock locks might destroy just granted lock if it could be merged
1589 with another existing flock, this is done in completion handler,
1590 so teach ldlm_cli_enquque_fini that this is a valid case for
1596 Description: Oops in read and write path when failing to allocate lock.
1597 Details : Check if lock allocation failed and return error back.
1601 Description: lstripe command fails for valid OST index
1602 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1603 instead of lov->desc.ld_active_tgt_count.
1607 Description: sanity.sh failed test 103
1608 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1613 Description: df doesn't work properly if diskfs blocksize != 4K
1614 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1618 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1619 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1621 Severity : enhancement
1623 Description: Adaptive timeouts
1624 Details : RPC timeouts adapt to changing server load and network
1625 conditions to reduce resend attempts and improve recovery time.
1629 Description: llapi_file_create() does not allow some changes
1630 Details : add llapi_file_open() that allows specifying the mode and
1631 open flags, and also returns an open file handle.
1635 Description: merge and cleanup kernel patches.
1636 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1640 Description: Add journal checksum support.(Kernel part)
1641 Details : The journal checksum feature adds two new flags i.e
1642 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1643 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1644 indicates that the commit block contains the checksum for
1645 the blocks described by the descriptor blocks. Now commit
1646 record can be sent to disk without waiting for descriptor
1647 blocks to be written to disk. This behavior is controlled
1648 using JBD2_FEATURE_ASYNC_COMMIT flag.
1652 Description: OSS needs mutliple precreate threads
1653 Details : Add ability to start more than one create thread per OSS.
1657 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1658 Details : execute lfs setstripe on client
1662 Description: mds_obd_create error creating tmp object
1663 Details : When the user sets quota on root, llog will be affected and can't
1664 create files and write files.
1667 Frequency : Always on ia64 patchless client, and possibly others.
1669 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1670 Details : This allows the patchless client to be loaded on architectures
1671 without this export.
1675 Description: RedHat Update kernel for RHEL5
1676 Details : Add the kernel config file for RHEL5.
1680 Description: RedHat Update kernel for RHEL5
1681 Details : Modify the kernel config file more closer RHEL5.
1685 Description: Build failure against Centos5 (RHEL5)
1686 Details : Define PAGE_SIZE when it isn't present.
1690 Description: test_7 on sanity-quota.sh can't be used on separate machines
1691 Details : add facet to handle it
1695 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1696 Details : with NFS, the anon dentry's parent was set to itself in
1697 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1698 corresponding dentry other than use rec->ur_name.
1700 Severity : enhancement
1702 Description: lfs setstripe enhancement
1703 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1707 Description: enable data checksumming by default
1708 Details : enable checksum by default, allow --disable-checksum
1709 configure option and "-o nochecksum" mount option. Checksums
1710 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1714 Description: lustre support for RHEL5
1715 Details : Add support for RHEL5.
1719 Description: Client eviction due to failover config
1720 Details : after a connection loss, the lustre client should attempt to
1721 reconnect to the last active server first before trying the
1722 other potential connections.
1726 Description: when mds and osts use different quota unit(32bit and 64bit),
1727 quota will be released repeatly.
1728 Details : void sending multiple quota reqs to mds, which will keep the
1729 status between the reqs.
1733 Description: osts not allocated evenly to files
1734 Details : change the condition to increase offset_idx
1737 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1739 Description: Data corruption for OSTs that are formatted larger than 2TB
1741 Details : When generating the bio request for lustre file writes the
1742 sector number would overflow a temporary variable before being
1743 used for the IO. The data reads correctly from Lustre (which
1744 will overflow in a similar manner) but other file data or
1745 filesystem metadata may be corrupted in some cases.
1749 Description: Tune the kernel for good SCSI performance.
1750 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1751 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1756 Description: rewrite ext2-derived code in obdclass/uuid.c
1757 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1758 from scratch preserving functionality.
1762 Description: "lfs find -obd UUID" prints directories
1763 Details : "lfs find -obd UUID" will return all directory names instead
1764 of just file names. It is incorrect because the directories
1765 do not reside on the OSTs.
1769 Description: use i_size_read and i_size_write in 2.6 port
1770 Details : replace inode->i_size access with i_size_read/write()
1773 Frequency : when removing large files
1775 Description: scheduling issue during removal of large Lustre files
1776 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1777 It causes scheduling issues when removing large files (17TB in the
1781 Frequency : only with liblustre clients on XT3
1783 Description: evictions taking too long
1784 Details : allow llrd to evict clients directly on OSTs
1787 Frequency : only on ppc
1789 Description: /proc/fs/lustre/devices broken on ppc
1790 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1791 We should make sure the type of 'index' is loff_t and then cast
1792 explicitly as needed below. Do not assign an explicitly cast
1796 Frequency : only for rhel5
1798 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1799 Details : Modify the target file & which_kernel.
1801 Severity : enhancement
1803 Description: omit set fsid for export NFS
1804 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1805 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1809 Frequency : Always, for kernels after 2.6.16
1810 Description: Fix warning idr_remove called for id=.. which is not allocated.
1811 Details : Last kernels save old s_dev before kill super and not allow
1812 to restore from callback - restore it before call kill_anon_super.
1816 Description: buffer overruns could theoretically occur
1817 Details : llapi_semantic_traverse() modifies the "path" argument by
1818 appending values to the end of the origin string, and a
1819 overrun may occur. Adding buffer overrun check in liblustreapi.
1823 Description: Fix error on 'ls .' at the top of the Lustre mount.
1824 Details : Don't revalidate dentry if it is a root dentry.
1828 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1829 Details : Modify vm-tunables-rhel4.patch.
1833 Description: Kernel config for 2.6.18-vanilla.
1834 Details : Modify targets/2.6-vanilla.target.in.
1835 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1836 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1837 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1838 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1842 Description: improve handling recoverable errors
1843 Details : if request processig with error which can be recoverable on server
1844 request should be resend, otherwise page released from cache and
1849 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1850 Details : Modify target file & which_patch.
1851 A flaw was found in the IA32 system call emulation provided
1852 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1853 value could be stored in the %RAX register, which could trigger an
1854 out-of-bounds system call table access. An untrusted local user
1855 could exploit this flaw to run code in the kernel
1856 (ie a root privilege escalation). (CVE-2007-4573).
1860 Description: O_DIRECT bypasses client statistics.
1861 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1862 read_ahead_stats were updated. copied the stats section from
1863 osc_send_oap_rpc() into async_internal().
1867 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1871 Description: change order of libsysio includes
1872 Details : '#include sysio.h' should always come before '#include xtio.h'
1876 Description: handle "serious error: objid * already exists" more gracefully
1877 Details : If LAST_ID value on disk is smaller than the objects existing in
1878 the O/0/d* directories, it indicates disk corruption and causes an
1879 LBUG(). If the object is 0-length, then we should use the existing
1880 object. This will help to avoid a full fsck in most cases.
1882 Severity : enhancement
1884 Description: adapt the lustre_config script to support the upgrade case
1885 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1886 server targets to 1.6 in parallel.
1890 Description: To avoid grant space > avaible space when the disk is almost
1891 full. Without this patch you might see the error "grant XXXX >
1892 available" or some LBUG about grant, when the disk is almost
1894 Details : In filter_check_grant, for non_grant cache write, we should
1895 check the left space by if (*left > ungranted + bytes), instead
1896 of (*left > ungranted), because only we are sure the left space
1897 is enough for another "bytes", then the ungrant space should be
1898 increase. In client, we should update cl_avail_grant only there
1899 is OBD_MD_FLGRANT in the reply.
1903 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1904 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1905 described in CVE-2007-4573. This problem could allow untrusted
1906 local users to gain root access.
1909 Frequency : when using O_DIRECT and quotas
1911 Description: Incorrect file ownership on O_DIRECT output files
1912 Details : block usage reported by 'lfs quota' does not take into account
1913 files that have been written with O_DIRECT.
1918 Description: touch file failed when fs is not full
1919 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1920 otherwise we can get ENOSP while fs is not full.
1924 Description: parallel lock callbacks
1925 Details : Instead of sending blocking and completion callbacks as separated
1926 requests, adding them to a set and sending in parallel.
1929 Frequency : only for Cray XT3
1930 Bugzilla : 12829/13455
1931 Description: Changing primary group doesn't change the group lustre assigns to
1933 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1934 overridden with the primary group provided by the group upcall,
1935 whereas the supplied fsgid can be trusted if it is in the list of
1936 supplementary groups returned by the group upcall.
1938 Severity : enhancement
1940 Description: Allow masking D_WARNING, D_ERROR messages from console
1941 Details : Console messages can now be disabled via lnet.printk.
1945 Description: User code with malformed file open parameter crashes client node
1946 Details : Before packing join_file req, all the related reference should be
1947 checked carefully in case some malformed flags cause fake
1948 join_file req on client.
1952 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1953 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1954 mode is equal to the granted mode. However, it should also check
1955 that LDLM_FL_CP_REQD is not set.
1959 Description: Heavy nfs access might result in deadlocks
1960 Details : After ELC code landed, it is now improper to enqueue any mds
1961 locks under och_sem, because enqueue might want to decide to
1962 cancel open locks for same inode we are holding och_sem for.
1966 Description: Client eviction while running blogbench
1967 Details : A lot of unlink operations with concurrent I/O can lead to a
1968 deadlock causing evictions. To address the problem, the number of
1969 oustanding OST_DESTROY requests is now throttled to
1970 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1971 callbacks are processed in priority.
1975 Description: enable ACLs on MDS by default
1976 Details : ACLs must be enabled on MDS by default.
1979 Frequency : PPC/PPC64 only
1981 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1982 Details : fix duplicated definitions between asm-ppc64/types.h and
1983 lustre_types.h on PPC.
1986 Frequency : PPC/PPC64 only
1988 Description: asm-ppc/segment.h does not exist
1989 Details : fix compile issue on PPC.
1993 Description: better handle error messages in extents code
1996 Frequency : RHEL4 only
1998 Description: mkfs is very slow on IA64/RHEL4
1999 Details : A performance regression has been discovered in the MPT Fusion
2000 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2001 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2002 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2004 Severity : enhancement
2006 Description: SNMP support enhancement
2007 Details : Adding total number of sampled request for an MDS node in snmp
2010 Severity : enhancement
2012 Description: Optimize ldlm waiting list processing for PR extent locks
2013 Details : When processing waiting list for read extent lock and meeting read
2014 lock that is same or wider to it that is not contended, skip
2015 processing rest of the list and immediatelly return current
2016 status of conflictness, since we are guaranteed there are no
2017 conflicting locks in the rest of the list.
2021 Description: Time out and refuse to reconnect
2022 Details : When the failover node is the primary node, it is possible
2023 to have two identical connections in imp_conn_list. We must
2024 compare not conn's pointers but NIDs, otherwise we can defeat
2025 connection throttling.
2029 Description: port llog fixes from b1_6 into HEAD
2030 Details : Port llog reference couting and some llog cleanups from b1_6
2031 (bug 10800) into HEAD, for protect from panic and access to already
2032 free llog structures.
2036 Description: Detect stride IO mode in read-ahead
2037 Details : When a client does stride read, read-ahead should detect that and
2038 read-ahead pages according to the detected stride pattern.
2042 Description: data checksumming impacts single node performance
2043 Details : add support for several checksum algorithm. Currently, only CRC32
2044 and Adler-32 are supported. The checksum type can be changed on
2045 the fly via /proc/fs/lustre/osc/*/checksum_type.
2049 Description: use adler32 for page checksums
2050 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2055 Description: build for x2 fails
2056 Details : fix compile issue on Cray systems.
2060 Description: Properly match for duplicate locks
2061 Details : Due to different lock order from skiplists code, we need to
2062 traverse entire list for now
2065 Frequency : only on PPC/SLES10
2067 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2068 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2069 be defined. Add a hack in mkfs_lustre.c to work around this compile
2074 Description: LASSERT on MDS when client holding flock lock dies
2075 Details : ldlm pool logic depends on number of granted locks equal to
2076 number of released locks which is not true for flock locks, so
2077 just exclude such locks from consideration.
2081 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2082 Details : Use fsfilt_sync() for both the whole filesystem sync and
2083 individual file sync to eliminate dangerous inode locking
2084 with I_LOCK that can lead to a deadlock.
2088 Description: performance in 1.6.3
2089 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2090 sure that 1MB requests are not fragmented by the block layer.
2092 Severity : enhancement
2094 Description: organize the server-side client stats on per-nid basis
2095 Details : Change the structure of stats under obdfilter and mds to
2103 The "uuid"s file would list the uuids of _active_ exports.
2104 And the clear entry is to clear all stats and stale nids.
2106 Severity : enhancement
2108 Description: eliminate client locks in face of contention
2109 Details : file contention detection and lockless i/o implementation
2110 for contended files.
2114 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2117 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2119 Description: MDS may lose file striping (and hence file data) in some cases
2120 Details : If there are additional extended attributes stored on the MDS,
2121 in particular ACLs, SELinux, or user attributes (if user_xattr
2122 is specified for the client mount options) then there is a risk
2123 of attribute loss. Additionally, the Lustre file striping
2124 needs to be larger than default (e.g. striped over all OSTs),
2125 and an additional attribute must be stored initially in the
2126 inode and then increase in size enough to be moved to the
2127 external attribute block (e.g. ACL growing in size) for file
2132 Description: skiplist implementation simplification
2133 Details : skiplists are used to group compatible locks on granted list
2134 that was implemented as tracking first and last lock of each
2135 lock group the patch changes that to using doubly linked lists
2139 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2140 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2141 because LASSERT on both the data supplied by a client, and the
2142 data on disk is dangerous and incorrect.
2144 Severity : enhancement
2146 Description: Slow truncate/writes to huge files at high offsets.
2147 Details : Directly associate cached pages to lock that protect those pages,
2148 this allows us to quickly find what pages to write and remove
2149 once lock callback is received.
2153 Description: more ldlm soft lockups
2154 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2155 starve other threads from the resource lock for a long time in
2156 case of long waiting queue, so change the debug level from
2157 D_OTHER to the less frequently used D_INFO.
2159 Severity : enhancement
2161 Description: add -gid, -group, -uid, -user options to lfs find
2165 Description: Hung threads in invalidate_inode_pages2_range
2166 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2167 once one is completed. As a result, some RPCs are stuck in the
2168 queue and are never sent.
2172 Description: filter threads hungs on waiting journal commit
2173 Details : Cleanup filter group llog code, then only filter group llog will
2174 be only created in the MDS/OST syncing process.
2178 Description: Procfs and llog threads access destoryed import sometimes.
2179 Details : Sync the import destoryed process with procfs and llog threads by
2180 the import refcount and semaphore.
2182 Severity : enhancement
2184 Description: openlock cache of b1_6 port to HEAD
2189 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2190 Details : If insertion of an extent fails, then discard the inode
2191 preallocation and free data blocks else it can lead to duplicate
2196 Description: don't always update ctime in ext3_xattr_set_handle()
2197 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2198 In some cases the ctime should not be updated, for example for
2199 2.0->1.8 compatibility it is necessary to delete an xattr and it
2200 should not update the ctime.
2204 Bugzilla : 15713/16362
2205 Description: Assertion in iopen_connect_dentry in 1.6.3
2206 Details : looking up an inode via iopen with the wrong generation number can
2207 populate the dcache with a disconneced dentry while the inode
2208 number is in the process of being reallocated. This causes an
2209 assertion failure in iopen since the inode's dentry list contains
2210 both a connected and disconnected dentry.
2214 Description: assertion failure in ldlm_handle2lock()
2215 Details : fix a race between class_handle_unhash() and class_handle2object()
2216 introduced in lustre 1.6.5 by bug 13622.
2221 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2222 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2227 Description: oops in sd_iostats_seq_show()
2228 Details : unloading/reloading the scsi low level driver triggers a kernel
2229 bug when trying to access the sd iostat file.
2234 Description: Kernel panics during QLogic driver reload
2235 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2236 patch, causing memory corruption.
2241 Description: journal_dev option does not work in b1_6
2242 Details : pass mount option during pre-mount.
2244 Severity : enhancement
2246 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2247 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2248 extent information of a file. It can be used to map logical blocks
2249 in a file to physical blocks in the block device.
2253 Description: LDLM soft lockups - improvement
2254 Details : It is be possible to send the lock handle along with each read
2255 or write request because the client is already doing a lock match
2256 itself so there isn't any reason the OST should have to re-do that
2260 Frequency : only on Cray X2
2262 Description: X2 build failures
2263 Details : fix build failures on Cray X2.
2267 Description: xid & resent requests
2268 Details : Initialize RPC XID from clock at startup (randomly if clock is
2271 Severity : enhancement
2273 Description: Add lustre_start utility to start or stop multiple Lustre servers
2278 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2279 Details : In case of memory pressure, list_del() can be called twice on
2280 req->rq_history_list, causing a kernel oops.
2284 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2285 Details : fix stack overflow in the distributed lock manager by defering
2286 export eviction after a failed AST to the elt thread instead of
2287 handling it in the dlm interpret routine.
2291 Description: Convert some comments to new format.
2292 Details : Update documenting comments to match doxygen conventions.
2296 Description: Grammar fixes.
2297 Details : A couple of trivial spelling fixes.
2301 Description: OSD_COUNTERS-mandatory
2302 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2303 to simplify the code.
2307 Description: simplify cmm_device freeing logic.
2308 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2312 Description: Add lockdep support to dt_object_operations locking interface.
2313 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2314 parameter indicating lock ordering. Update mdd code to use new
2319 Description: Introduce failloc constants for lockless IO tests.
2320 Details : Add two new failloc constants to test lockless IO. Only one of
2321 them in implemented---another is checked in yet to be landed
2326 Description: Add lockdep support for inode mutex.
2327 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2328 in the situations where Lustre has to lock more than one inode
2333 Description: Add optional invariants checking support.
2334 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2335 configure switch. This macro is to be used for consistency and
2336 sanity checks that are too expensive to be left in `production'
2341 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2342 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2347 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2348 Details : Use spin_lock_nested() in (the only) situation where more than
2349 one ldlm_lock is locked simultaneously. Also, fix possible
2350 dead-lock in ldlm_lock_change_resource() by enforcing particular
2355 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2356 Details : Instead of specifying each ldlm_lock call-back through separate
2357 parameter, wrap them into struct ldlm_callback_suite.
2361 Description: Kill join_lru obd method and its callers.
2362 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2363 in memory---a requirement implemented through ->o_join_lru() obd
2364 method in HEAD. Remove this method and its users.
2368 Description: Add asynchronous ldlm ENQUEUE completion handler.
2369 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2370 case that stock ldlm_completion_ast() cannot handle as it waits
2371 until lock is granted. Introduce new ldlm_completion_ast_async()
2372 for this. Also comment ldlm_completion_ast().
2376 Description: ldlm_error <-> errno conversion.
2377 Details : Add functions to map (rather arbitrary) between LDLM error codes
2378 and standard errno values. CLIO needs this to prevent LDLM specific
2379 constants from escaping ldlm and osc.
2383 Description: Kill unused ldlm_handle2lock_ns() function.
2384 Details : Kill unused ldlm_handle2lock_ns() function.
2388 Description: Add lu_ref support to ldlm_lock
2389 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2390 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2391 and ldlm_resource. LDLM interface has to be changed, because code that
2392 releases a reference on a lock, has to "know" what reference this is.
2393 In the most frequent case
2395 lock = ldlm_handle2lock(handle);
2397 LDLM_LOCK_PUT(lock);
2399 no changes are required. When any other reference (received _not_ from
2400 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2401 instead of LDLM_LOCK_PUT().
2403 Arguably, changes are pervasive, and interface requires some discipline
2404 for proper use. On the other hand, it was very instrumental in finding
2405 a few leaked lock references.
2409 Description: Add ldlm_lock_addref_try().
2410 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2411 attempts to addref a lock that might be being canceled
2416 Description: Add ldlm_weigh_callback().
2417 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2418 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2419 hard-coded `number of pages' logic.
2423 Description: Add lockdep annotations to llog code.
2424 Details : Use appropriately tagged _nested() locking calls in the places
2425 where llog takes more than one ->lgh_lock lock.
2429 Description: Add loi_kms_set().
2430 Details : Wrap kms updates into a helper function.
2434 Description: Constify instances of struct lsm_operations.
2435 Details : Constify instances of struct lsm_operations.
2439 Description: lu_conf support.
2440 Details : On a server, a file system object is uniquely identified
2441 by a fid, which is sufficient to locate and load all object
2442 state (inode). On a client, on the other hand, more data are
2443 necessary instantiate an object. Change lu_object_find() and
2444 friends to take additional `lu_conf' argument describing object.
2445 Typically this includes layout information.
2449 Description: lu_context fixes.
2450 Details : Introduce new lu_context functions that are needed on the client
2451 side, where some system threads (ptlrpcd) are shared by multiple
2452 modules, and so cannot be stopped during module shutdown.
2456 Description: Add start and stop methods to lu_device_type_operations.
2457 Details : Introduce two new methods in lu_device_type_operations, that are
2458 invoked when first instance of a given type is created and last one
2459 is destroyed respectively. This is need by CLIO.
2463 Description: Add lu_ref support to struct lu_device.
2464 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2465 track leaked references.
2469 Description: Introduce lu_kmem_descr.
2470 Details : lu_kmem_descr and its companion interface allow to create
2471 and destroy a number of kmem caches at once.
2475 Description: Fix lu_object finalization race.
2476 Details : Fix a race between lu_object_find() finding an object and its
2477 concurrent finalization. This race is (most likely) not possible
2478 on the server, but might happen on the client.
2482 Description: Introduce lu_ref interface.
2483 Details : lu_ref is a debugging module allowing to track references to
2484 a given object. It is quite cpu expensive, and has to be
2485 explicitly enabled with --enable-lu_ref. See usage description
2490 Description: Factor lu_site procfs stats into a separate function.
2491 Details : Separate lu_site stats printing code into a separate function
2492 that can be reused on a client.
2496 Description: Constify instances of struct {lu,dt,md}_device_operations.
2497 Details : Constify instances of struct {lu,dt,md}_device_operations.
2501 Description: Introduce struct md_site and move meta-data specific parts of
2502 struct lu_site here.
2503 Details : Move md-specific fields out of struct lu_site into special struct
2504 md_site, so that lu_site can be used on a client.
2508 Description: Kill mdd_lov_destroy().
2509 Details : Remove unused mdd code.
2513 Description: Add st_block checking to multistat.c.
2514 Details : Add st_block checking to multistat.c.
2518 Description: Add lu_ref support to struct obd_device.
2519 Details : Add lu_ref tracking to obd_device.
2523 Description: Kill obd_set_fail_loc().
2524 Details : Remove unused code.
2528 Description: Add special type for ptlrpc_request interpret functions.
2529 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2530 there. Actual usage will be in CLIO.
2534 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2535 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2536 doesn't work with lockdep.
2540 Description: Add rwv.c test program.
2541 Details : New testing program exercising readv(2) and writev(2) (Qian).
2545 Description: Add sendfile.c test program.
2546 Details : New testing program exercising sendfile(2) (Jay).
2550 Description: Ratelimit a message that can be very frequent.
2551 Details : Ratelimit a memory allocation failure message that can
2556 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2557 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2561 Description: Liblustre build fixes.
2562 Details : Liblustre build fixes.
2566 Description: libcfs: add cfs_{need,cond}_resched() interface.
2567 Details : libcfs: add cfs_{need,cond}_resched() definition and
2568 implementations for Linux, NT, and liblustre.
2570 Severity : enhancement
2572 Description: More exported tunables for mballoc
2573 Details : Add support for tunable preallocation window and new tunables for
2574 large/small requests
2578 Description: Detect corruption of block bitmap and checking for preallocations
2579 Details : Checks validity of on-disk block bitmap. Also it does better
2580 checking of number of applied preallocations. When corruption is
2581 found, it turns filesystem readonly to prevent further corruptions.
2585 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2586 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2587 window to 0. If it does hit cache miss, and read-pattern is still
2588 stride-io mode, does not reset the stride window, but also does not
2589 increase the stride window length in this case.
2593 Frequency : only for big-endian servers
2594 Description: Check if system is big-endian while mounting fs with extents feature
2595 Details : Mounting a filesystem with extents feature will fail on big-endian
2596 systems since ext3-based ldiskfs is not supported on big-endian
2597 systems. This can be over-riden with "bigendian_extents" mount option.
2599 Severity : enhancement
2601 Description: The root squash functionality
2602 Details : A security feature, which is to prevent users from being able
2603 to mount lustre on their desktop, run as root, and delete
2604 all of the files in the filesystem. The goal is accomplished by
2605 remapping user id (UID) and group id (GID) of the root user to
2606 a UID and GID specified by the system administartor via Lustre
2607 configuration management server (MGS). The functionality also
2608 allows to specify sets of clients for which the remapping does
2613 Description: Excessive recovery window
2614 Details : With AT enabled, the recovery window can be excessively long (6000+
2615 seconds). To address this problem, we no longer use
2616 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2617 timeout no longer depends on the service time, it is set to
2618 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2619 time via pb_service_time.
2623 Description: Watchdog triggered on MDS failover
2624 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2625 the OSTs know that the MDS "UUID" can be reused for the same export
2626 from a different NID, so we do not need to wait for the export to be
2630 Frequency : rare, only if using MMP with Linux RAID
2632 Description: MMP doesn't work with Linux RAID
2633 Details : While using HA for Lustre servers with Linux RAID, it is possible
2634 that MMP will not detect multiple mounts. To make this work we
2635 need to unplug the device queue in RAID when the MMP block is being
2636 written. Also while reading the MMP block, we should read it from
2637 disk and not the cached one.
2639 Severity : enhancement
2641 Description: open file using fid
2642 Details : A file can be opened using just its fid, like
2643 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2646 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2649 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2650 Details : add proper locking for creation of jbd_4k slab cache
2654 Description: MMP check in ext3_remount() fails without displaying any error
2655 Details : When multiple mount protection fails during remount, proper error
2658 Severity : enhancement
2660 Description: Allow stripe size to be up to 4G-64k
2661 Details : Fix math logic to allow large stripe sizes.
2665 Description: add check for >8TB ldiskfs filesystems
2666 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2667 Don't allow >8TB ldiskfs filesystems to be mounted without
2668 force_over_8tb mount option
2673 Descriptoin: OSS complains "too many blocks" and crashes
2674 Details : In filter_commitrw_write():
2675 1. lquota_chkquota();
2676 2. obdo_from_inode();
2677 3. lquota_pending_commit();
2678 "2." has changed uid and gid in this case, then caused
2679 lquota_chkquota() and lquota_pending_commit() to use
2680 different uid and gid. That is the root of the bug.
2684 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2685 Details : Detect and print out the 'stripe count = -1' case properly.
2689 Description: Out or order replies might be lost on replay
2690 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2691 request with tid smaller then one currently being added, add it
2692 to the start, not end of the list.
2694 --------------------------------------------------------------------------------
2696 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2698 * Support for kernels:
2699 2.6.5-7.283 (SLES 9),
2700 2.6.9-55.EL (RHEL 4),
2701 2.6.16.46-0.14 (SLES 10),
2702 2.6.18.8 vanilla (kernel.org)
2703 * Client support for unpatched kernels:
2704 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2705 2.6.16 - 2.6.22 vanilla (kernel.org)
2706 * Due to recently discovered recovery problems, we do not recommend
2707 using patchless RHEL 4 clients with this or any earlier release.
2708 * Recommended e2fsprogs version: 1.39.cfs8
2709 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2710 * Starting with this release, the ldiskfs backing filesystem required
2711 by Lustre is now in its own package, lustre-ldiskfs. This package
2712 should be installed. It is versioned separately from Lustre and
2713 may be released separately in future.
2717 Description: block reactivating mgc import until all deactivates complete
2718 Details : Fix race when failing back MDT/MGS to itself (testing)
2721 Frequency : at statup only
2723 Description: mds_lov_synchronize race leads to various problems
2724 Details : simultaneous MDT->OST connections at startup can cause the
2725 sync to abort, leaving the OSC in a bad state.
2727 Severity : enhancement
2729 Description: add optional extra BUILD_VERSION info
2730 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2731 to override the lustre version.
2734 Frequency : 2.6.18 servers only
2736 Description: ll_kern_mount() doesn't release the module reference
2737 Details : The ldiskfs module reference count never drops down to 0
2738 because ll_kern_mount() doesn't release the module reference.
2743 Description: server LBUG when using old ost_num_threads parameter
2744 Details : Accept the old ost_num_threads parameter but warn that it
2745 is deprecated, and fix an off-by-one error that caused an LBUG.
2750 Description: Transient SCSI error results in persistent IO issue
2751 Details : iobuf->dr_error is not reinitialized to 0 between two
2755 Frequency : sometimes when underlying device returns I/O errors
2757 Description: OSTs not going read-only during write failures
2758 Details : OSTs are not remounted read-only when the journal commit threads
2759 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2760 instead of the ext3 wrappers.
2764 Description: poor connect scaling with increasing client count
2765 Details : Don't run filter_grant_sanity_check for more than 100 exports
2766 to improve scaling for large numbers of clients.
2769 Frequency : SLES10 only
2771 Description: sanity-quota.sh quotacheck failed: rc = -22
2772 Details : Quotas cannot be enabled on SLES10.
2775 Frequency : liblustre clients only
2777 Description: getdirentries does not give error when run on compute nodes
2778 Details : getdirentries does not fail when the size specified as an argument
2779 is too small to contain at least one entry
2781 Severity : enhancement
2783 Description: Add LNET router traceability for debug purposes
2784 Details : If a checksum failure occurs with a router as part of the
2785 IO path, the NID of the last router that forwarded the bulk data
2786 is printed so it can be identified.
2791 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2792 Details : Due to a race condition, liblustre clients were occasionally
2793 evicted incorrectly.
2795 Severity : enhancement
2797 Description: lfs setstripe use optional parameters instead of postional
2800 Severity : enhancement
2802 Description: Nanosecond timestamp support for ldiskfs
2803 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2804 resolution timestamps. There is not yet support for this at
2805 the Lustre filesystem level.
2808 Frequency : during server recovery
2810 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2811 Details : request with rq_no_resend flag not awake l_wait_event if they get
2815 Frequency : nfs export on patchless client
2817 Description: connectathon hang when test nfs export over patchless client
2818 Details : Disconnected dentry cannot be found with lookup, so we do not need
2819 to unhash it or make it invalid
2822 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2823 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2824 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2829 Description: open req refcounting wrong on reconnect
2830 Details : If reconnect happened between getting open reply from server and
2831 call to mdc_set_replay_data in ll_file_open, we will schedule
2832 replay for unreferenced request that we are about to free.
2833 Subsequent close will crash in variety of ways.
2834 Check that request is still eligible for replay in
2835 mdc_set_replay_data().
2840 Description: disable writes to filesystem when reading health_check file
2841 Details : the default for reading the health_check proc file has changed
2842 to NOT do a journal transaction and write to disk, because this
2843 can cause reads of the /proc file to hang and block HA state
2844 checking on a healthy but otherwise heavily loaded system. It
2845 is possible to return to the previous behaviour during configure
2846 with --enable-health-write.
2848 Severity : enhancement
2850 Description: 64-bit inode version
2851 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2852 to the inode. This will be required for version-based recovery.
2857 Description: MDS fails to start if a duplicate client export is detected
2858 Details : in some rare cases it was possible for a client to connect to
2859 an MDS multiple times. Upon recovery the MDS would detect this
2860 and fail during startup. Handle this more gracefully.
2862 Severity : enhancement
2864 Description: Add -o localflock option to simulate old noflock
2866 Details : This will achieve local-only flock/fcntl locks
2872 Description: log_commit_thread vs filter_destroy race leads to crash
2873 Details : Take import reference before releasing llog record semaphore
2878 Description: Wrong request locking in request set processing
2879 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2880 imp_delayed_list, in this place should be used imp_lock.
2883 Frequency : when reconnection
2885 Description: Grant Leak when osc reconnect to OST
2886 Details : When osc reconnect ost, OST(filter) should check whether it
2887 should grant more space to client by comparing fed_grant and
2888 cl_avail_grant, and return the granted space to client instead
2889 of "new granted" space, because client will call osc_init_grant
2890 to update the client grant space info.
2893 Frequency : when client reconnect to OST
2895 Description: Grant Leak when osc do resend and replay bulk write
2896 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2897 bulk write request, because the grant info while be sync between
2898 OSC and OST when reconnect, and we should ignore the grant info
2899 these of resend/replay write req.
2904 Description: Grant space more than avaiable left space sometimes.
2905 Details : When then OST is about to be full, if two bulk writing from
2906 different clients came to OST. Accord the avaliable space of the
2907 OST, the first req should be permitted, and the second one
2908 should be denied by ENOSPC. But if the seconde arrived before
2909 the first one is commited. The OST might wrongly permit second
2910 writing, which will cause grant space > avaiable space.
2913 Frequency : when client is evicted
2915 Description: Grant might be wrongly erased when osc is evicted by OST
2916 Details : when the import is evicted by server, it will fork another
2917 thread ptlrpc_invalidate_import_thread to invalidate the
2918 import, where the grant will be set to 0. While the original
2919 thread will update the grant it got when connecting. So if
2920 the former happened latter, the grant will be wrongly errased
2921 because of this race.
2926 Description: Checking Stale with correct fid
2927 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2928 to check whether it is stale, because sometimes, we want the
2929 enqueue happened anyway, and op_data.fid2 will not be initialized.
2931 Severity : enhancement
2933 Description: update patchless client
2934 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2937 Frequency : only with 2.4 kernel
2939 Description: random memory corruption
2940 Details : size of struct ll_inode_info is to big for union inode.u and this
2941 can be cause of random memory corruption.
2946 Description: Memory leak in recovery
2947 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2948 It should also check obd_fail before fsfilt_start, otherwise if
2949 fsfilt_start return -EROFS,(failover mds during mds recovery).
2950 then the req will return with repmsg->transno = 0 and rc = EROFS.
2951 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2952 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2953 be freed no matter whether fsfilt_commit success or not.
2956 Frequency : only with huge count clients
2958 Description: Prevents from taking the superblock lock in llap_from_page for
2960 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2961 for llap_from_page call in ll_removepage prevents from taking the
2962 superblock lock for a soon died page.
2967 Description: Not check open intent error before release open handle
2968 Details : in some rare cases, the open intent error is not checked before
2969 release open handle, which may cause
2970 ASSERTION(open_req->rq_transno != 0), because it tries to release
2971 the failed open handle.
2976 Description: Set cat log bitmap only after create log success.
2977 Details : in some rare cases, the cat log bitmap is set too early. and it
2978 should be set only after create log success.
2982 Description: Accessing a block bevice can re-enable I/O when Lustre is
2983 tearing down a device.
2984 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2989 Description: service threads may hog cpus when there are a lot of requests
2990 Details : Insert cond_resched to give other threads a chance to use some CPU
2995 Description: the cat log was not initialized in recovery
2996 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2997 unlink log on mds will not be initialized until mds post
2998 recovery. And also in mds post recovery, the unlink log will
2999 initialization will be done asynchronausly, so there will be race
3000 between add unlink log and unlink log initialization.
3004 Description: brw_stats were being printed incorrectly
3005 Details : brw_stats were being printed as log2 but all of them were not
3006 recorded as log2. Also remove some code duplication arising from
3007 filter_tally_{read,write}.
3011 Frequency : rare, only in recovery.
3012 Description: ASSERTION(req->rq_type != LI_POISON) failed
3013 Details : imp_lock should be held while iterating over imp_sending_list for
3014 prevent destroy request after get timeout in ptlrpc_queue_wait.
3018 Description: replay-single.sh test 52 fails
3019 Details : A lock's skiplist need to be cleanup when it being unlinked
3020 from its resource list.
3024 Description: Short directio read returns full requested size rather than
3026 Details : Direct I/O operations should return actual amount of bytes
3027 transferred rather than requested size.
3029 Severity : enhancement
3031 Description: metadata RPC reduction (e.g. for rm performance)
3032 Details : decrease the amount of synchronous RPC between clients and servers
3033 by canceling conflicing lock before the operation on the client
3034 and packing thier handles into the main operation RPC to server.
3036 Severity : enhancement
3038 Description: Async OSC create to avoid the blocking unnecessarily.
3039 Details : If a OST has no remain object, system will block on the creating
3040 when need to create a new object on this OST. Now, ways use
3041 pre-created objects when available, instead of blocking on an
3042 empty osc while others are not empty. If we must block, we block
3043 for the shortest possible period of time.
3047 Description: improve handling recoverable errors
3048 Details : if request processig with error which can be recoverable on server
3049 request should be resend, otherwise page released from cache and
3052 Severity : enhancement
3054 Description: refine locking for avoid write wrong info into lov_objid file
3055 Details : fix possible races with add new target and write/update data in
3058 --------------------------------------------------------------------------------
3060 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3065 Frequency : on some architectures
3067 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3068 Details : Uninitialized flags sometimes cause configuration commands to
3072 Frequency : patchless clients only
3074 Description: missing __iget() symbol export
3075 Details : The __iget() symbol export is missing. To avoid the need for
3076 this on patchless clients the deathrow inode reaper is turned
3077 off, and we depend on the VM to clean up old inodes. This
3078 dependency was during via the fix for bug 12181.
3080 --------------------------------------------------------------------------------
3082 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3084 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3085 INTEROPERATE with older versions automatically. In many cases a
3086 special upgrade step is needed. Please read the
3087 user documentation before upgrading any part of a live system.
3088 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3089 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3090 * WARNING: Lustre configuration and startup changes are required with
3091 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3095 Severity : enhancement
3097 Description: MountConf
3098 Details : Lustre configuration is now managed via mkfs and mount
3099 commands instead of lmc and lconf. New obd types (MGS, MGC)
3100 are added for dynamic configuration management. See
3101 https://mail.clusterfs.com/wikis/lustre/MountConf for
3104 Severity : enhancement
3106 Description: dynamic OST addition
3107 Details : OSTs can now be added to a live filesystem
3109 Severity : enhancement
3111 Description: startup order invariance
3112 Details : MDTs and OSTs can be started in any order. Clients only
3113 require the MDT to complete startup.
3115 Severity : enhancement
3117 Description: parallel, asynchronous orphan cleanup
3118 Details : orphan cleanup is now performed in separate threads for each
3119 OST, allowing parallel non-blocking operation.
3121 Severity : enhancement
3123 Description: optimized stripe assignment
3124 Details : stripe assignments are now made based on ost space available,
3125 ost previous usage, and OSS previous usage, in order to try
3126 to optimize storage space and networking resources.
3128 Severity : enhancement
3130 Description: Permanently set tunables
3131 Details : All writable /proc/fs/lustre tunables can now be permanently
3132 set on a per-server basis, at mkfs time or on a live system.
3134 Severity : enhancement
3136 Description: Lustre message v2
3137 Details : Add lustre message format v2.
3139 Severity : enhancement
3141 Description: client OST exclusion list
3142 Details : Clients can be started with a list of OSTs that should be
3143 declared "inactive" for known non-responsive OSTs.
3147 Description: SPEC SFS validation failure on NFS v2 over lustre.
3148 Details : Changes the blocksize for regular files to be 2x RPC size,
3149 and not depend on stripe size.
3151 Severity : enhancement
3153 Description: Multiple MD RPCs in flight.
3154 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3155 To avoid overly-overloading MDS, introduce a limit on number of
3156 MDS RPCs in flight for a single client and add /proc controls
3157 to adjust this limit.
3159 Severity : enhancement
3161 Description: client read/write statistics
3162 Details : Add client read/write call usage stats for performance
3163 analysis of user processes.
3164 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3165 file access. extents_stats shows chunk size distribution.
3166 extents_stats_per_process show chunk size distribution per
3169 Severity : enhancement
3171 Description: mds statistics
3172 Details : Add detailed mds operations statistics in
3173 /proc/fs/lustre/mds/*/stats.
3177 Description: Failure of copying files with lustre special EAs.
3178 Details : Client side always return success for setxattr call for lustre
3179 special xattr (currently only "trusted.lov").
3184 Description: Refcount LNET uuids
3185 Details : The global LNET uuid list grew linearly with every startup;
3186 refcount repeated list entries instead of always adding to
3189 Severity : enhancement
3191 Description: Dynamic service threads
3192 Details : Within a small range, start extra service threads
3193 automatically when the request queue builds up.
3196 Frequency : mixed-endian client/server environments
3198 Description: mixed-endian crashes
3199 Details : The new msg_v2 system had some failures in mixed-endian
3202 Severity : enhancement
3204 Description: Easy OST removal
3205 Details : OSTs can be permanently deactivated with e.g. 'lctl
3206 conf_param lustre-OST0001.osc.active=0'
3208 Severity : enhancement
3210 Description: MGS proc entries
3211 Details : Added basic proc entries for the MGS showing what filesystems
3214 Severity : enhancement
3216 Description: provide MGS failover
3217 Details : Added config lock reacquisition after MGS server failover.
3219 Severity : enhancement
3221 Description: add Linux 2.4 support
3222 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3226 Description: a large application tries to do I/O to the same resource and dies
3227 in the middle of it.
3228 Details : Check the req->rq_arrival time after the call to
3229 ost_brw_lock_get(), but before we do anything about
3230 processing it & sending the BULK transfer request. This
3231 should help move old stale pending locks off the queue as
3232 quickly as obd_timeout.
3235 Frequency : when an incorrect nid is specified during startup
3237 Description: ptlrpc connect to non-existant node causes kernel crash
3238 Details : LNET can't be re-entered from an event callback, which
3239 happened when we expire a message after the export has been
3240 cleaned up. Instead, hand the zombie cleanup off to another
3243 Severity : enhancement
3245 Description: plain/inodebits lock performance improvement
3246 Details : Grouping plain/inodebits in granted list by their request modes
3247 and bits policy, thus improving the performance of search through
3251 Frequency : only if OST filesystem is corrupted
3253 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3254 Details : for a short time RPCs with bulk IO are in the replay list,
3255 but replay of bulk IOs is unimplemented. If the OST filesystem
3256 is corrupted due to disk cache incoherency and then replay is
3257 started it is possible to trip an assertion. Avoid putting
3258 committed RPCs into the replay list at all to avoid this issue.
3261 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3263 Description: System hang on startup
3264 Details : This bug allowed the liblustre (e.g. catamount) client to
3265 return to the app before handling all startup RPCs. This
3266 could leave the node unresponsive to lustre network traffic
3267 and manifested as a server ptllnd timeout.
3269 Severity : enhancement
3271 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3272 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3273 boolean to control whether to print peer debug info when a
3274 client's RPC times out.
3277 Frequency : only for kernels with patches from Lustre below 1.4.3
3279 Description: Remove old rdonly API
3280 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3283 Frequency : only for devices with external journals
3285 Description: Set external device read-only also
3286 Details : During a commanded failover stop, we set the disk device
3287 read-only while the server shuts down. We now also set any
3288 external journal device read-only at the same time.
3291 Frequency : when upgrading from 1.4 while trying to change parameters
3293 Description: The wrong (new) MDC name was used when setting parameters for
3294 upgraded MDT's. Also allows changing of OSC (and MDC)
3295 parameters if --writeconf is specified at tunefs upgrade time.
3298 Frequency : when setting specific ost indicies
3300 Description: QOS code breaks on skipped indicies
3301 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3302 created with --index need not be sequential.
3304 Severity : enhancement
3306 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3307 Details : The uninit_groups feature works in conjunction with the kernel
3308 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3309 up the pass1 processing of e2fsck. This is a read-only feature
3310 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3311 mount filesystems that have had this feature enabled.
3313 Severity : enhancement
3315 Description: Improve multi-block allocation algorithm to avoid fragmentation
3316 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3317 allocation locality and avoid filesystem fragmentation.
3319 ------------------------------------------------------------------------------
3321 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3323 * Support for kernels:
3324 2.4.21-47.0.1.EL (RHEL 3)
3325 2.6.5-7.283 (SLES 9)
3326 2.6.9-42.0.10.EL (RHEL 4)
3327 2.6.12.6 vanilla (kernel.org)
3328 2.6.16.27-0.9 (SLES 10)
3329 * Recommended e2fsprogs version: 1.39.cfs5
3331 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3335 Frequency : occasional, depends on client load and configuration
3336 Bugzilla : 12181, 12203
3337 Description: data loss for recently-modified files
3339 Details : In some cases it is possible that recently written or created
3340 files may not be written to disk in a timely manner (this should
3341 normally be within 30s unless client IO load is very high).
3342 The problem appears as zero-length files or files that are a
3343 multiple of 1MB in size after a client crash or client eviction
3344 that are missing data at the end of the file.
3346 This problem is more likely to be hit on clients where files are
3347 repeatedly created and unlinked in the same directory, clients
3348 have a large amount of RAM, have many CPUs, the filesystem has
3349 many OSTs, the clients are rebooted frequently, and/or the files
3350 are not accessed by other nodes after being written.
3352 The presence of the problem can be detected by looking at
3353 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3354 smaller than the second (nr_unused) then dirty files will not
3355 be flushed automatically to disk. "sync; sleep 10" should be
3356 run several times on the node before unmounting it to update
3357 Lustre (this is also safe to run on nodes without this problem).
3359 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3360 that can cause this same problem, so customers using that kernel
3361 also need to update the kernel in addition to Lustre. In order
3362 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3364 It is normal that files written just before a client crash (less
3365 than 30s) may not yet have been flushed to disk, even for local
3369 Frequency : frequent on thin XT3 nodes
3371 Description: UUID collision on thin XT3 Linux nodes
3372 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3373 randomly, since we relied on an insufficiently-seeded PRNG.
3378 Description: OSS hangs after "All ost request buffers busy"
3379 Details : A deadlock between quota and journal operations caused OSS
3380 hangs after printing "All ost request buffers busy."
3383 Frequency : always on liblustre builds
3385 Description: Cleanup compiler warnings on liblustre
3388 Frequency : always on liblustre builds on XT3
3390 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3394 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3395 > 32000 subdirectories
3396 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3397 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3398 subdirectory count crosses 32000. This will aid e2fsck to
3399 correctly handle more than 32000 subdirectories.
3402 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3404 Description: System hang on startup
3405 Details : This bug allowed the liblustre (e.g. catamount) client to
3406 return to the app before handling all startup RPCs. This
3407 could leave the node unresponsive to lustre network traffic
3408 and manifested as a server ptllnd timeout.
3410 Severity : enhancement
3412 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3413 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3414 boolean to control whether to print peer debug info when a
3415 client's RPC times out.
3420 Description: make O_SYNC working on 2.6 kernels
3421 Details : 2.6 kernels use different method for mark pages for write,
3422 so need add a code to lustre for O_SYNC work.
3427 Description: Failure to close file and release space on NFS
3428 Details : Put inode details into lock acquired in ll_intent_file_open.
3429 Use mdc_intent_lock in ll_intent_open to properly
3430 detect all kind of errors unhandled by mdc_enqueue.
3435 Description: proc file read during shutdown sometimes raced obd removal,
3437 Details : Add lock to prevent obd access after proc file removal.
3440 Frequency : Only for files larger than 4GB on 32-bit clients.
3442 Description: improperly doing page alignment of locks
3443 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3444 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3449 Description: RPCs being resent when they shouldn't be
3450 Details : Some RPCs that should not be resent are being resent. This
3451 can cause inconsistencies in the RPC state machine. Do not
3452 resend such requests.
3455 Frequency : rare, only with NFS export
3457 Description: Crash on NFS re-export node
3458 Details : under very unusual load conditions an assertion is hit in
3459 ll_intent_file_open()
3462 Frequency : only if OST filesystem is corrupted
3464 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3465 Details : for a short time RPCs with bulk IO are in the replay list,
3466 but replay of bulk IOs is unimplemented. If the OST filesystem
3467 is corrupted due to disk cache incoherency and then replay is
3468 started it is possible to trip an assertion. Avoid putting
3469 committed RPCs into the replay list at all to avoid this issue.
3474 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3475 Details : Large single O_DIRECT read and write calls can fail to allocate
3476 a sufficiently large buffer to process the request. In case of
3477 allocation failure the allocation is retried with a smaller
3478 buffer and broken into smaller requests.
3480 Severity : enhancement
3482 Description: Add -o localflock option to simulate old noflock behaviour.
3483 Details : This will achieve local-only flock/fcntl locks coherentness.
3488 Description: versioning check is incomplete
3489 Details : Checking the version difference of client vs. server, report
3490 error if the gap is too big.
3495 Description: add support PG_writeback bit
3496 Details : add support for PG_writeback bit for Lustre, for more carefull
3497 work with page cache in 2.6 kernel. This also fix some deadlocks
3498 and remove hack for work O_SYNC with 2.6 kernel.
3500 Severity : enhancement
3502 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3503 Details : The uninit_groups feature works in conjunction with the kernel
3504 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3505 up the pass1 processing of e2fsck. This is a read-only feature
3506 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3507 mount filesystems that have had this feature enabled.
3509 Severity : enhancement
3511 Description: Improve multi-block allocation algorithm to avoid fragmentation
3512 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3513 allocation locality and avoid filesystem fragmentation.
3515 ------------------------------------------------------------------------------
3517 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3519 * Support for kernels:
3520 2.6.9-42.0.3.EL (RHEL 4)
3521 2.6.5-7.276 (SLES 9)
3522 2.4.21-47.0.1.EL (RHEL 3)
3523 2.6.12.6 vanilla (kernel.org)
3524 2.6.16.21-0.8 (SLES10)
3525 * Recommended e2fsprogs version: 1.39.cfs2-0
3527 * The backwards-compatible /proc/sys/portals symlink has been removed
3528 in this release. Before upgrading, please ensure that you change
3529 any configuration scripts or /etc/sysctl.conf files that access
3530 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3531 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3532 in advance of the upgrade on any system running Lustre 1.4.6 or
3533 newer, since /proc/sys/lnet was added in that version.
3534 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3538 Frequency : only when quota is used
3540 Description: avoid scanning export list for quota master
3541 Details : Change the algorithms to avoid scanning export list in order
3542 to improve the efficiency.
3545 Frequency : MDS failover only, very rarely
3547 Description: "went back in time" messages on mds failover
3548 Details : The greatest transno may be lost when the current operation
3549 finishes with an error (transno==0) and the client's last_rcvd
3550 record is over-written. Save the greatest transno in the
3551 mds_last_transno for this case.
3554 Frequency : always for specific kernels and striping counts
3556 Description: client may get "Matching packet too big" without ACL support
3557 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3558 when trying to access files in certain configurations. The
3559 clients should in fact be denied when mounting because they do
3560 not understand ACLs.
3563 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3565 Description: many clients connecting with IO in progress causes connect timeouts
3566 Details : Avoid synchronous journal commits to avoid delays caused by many
3567 clients connecting/disconnecting when bulk IO is in progress.
3568 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3569 OST_IO_PORTAL to avoid delays behind potentially many pending
3573 Frequency : occasionally with multiple writers to a single file
3575 Description: shared writes to file may result in wrong size reported by stat()
3576 Details : Allow growing of kms when extent lock is cancelled
3579 Frequency : always with random mmap IO to multi-striped file
3581 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3582 Details : Only if the hole is at the end of OST object so that kms is too
3583 small. Fix is to increase kms accordingly in ll_nopage.
3586 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3588 Description: writes to a missing object would leak memory on the OST
3589 Details : If there is an inconsistency between the MDS and OST filesystems,
3590 such that the MDS references an object that doesn't exist, writes
3591 to that object will leak memory due to incorrect cleanup in the
3592 error handling path, eventually running out of memory on the OST.
3597 Description: Creating too long symlink causes lustre errors
3598 Details : Check symlink and name lengths before sending requests to MDS.
3601 Frequency : only if flock is enabled (not on by default)
3603 Description: posix locks not released on fd closure on 2.6.9+
3604 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3605 kernels, this caused such locks not to be released on fd close and
3606 then assertions on fs unmount about still used locks.
3609 Frequency : MDS failover only, very rarely
3611 Description: clients may get ASSERTION(granted_lock != NULL)
3612 Details : When request was taking a long time, and a client was resending
3613 a getattr by name lock request. The were multiple lock requests
3614 with the same client lock handle and
3615 mds_getattr_name->fixup_handle_for_resent_request found one of the
3616 lock handles but later failed with ASSERTION(granted_lock != NULL).
3621 Description: handle->h_buffer_credits > 0, assertion failure
3622 Details : h_buffer_credits is zero after truncate, causing assertion
3623 failure. This patch extends the transaction or creates a new
3627 Frequency : NFS re-export or patchless client
3628 Bugzilla : 11179, 10796
3629 Description: Crash on NFS re-export node (__d_move)
3630 Details : We do not want to hash the dentry if we don't have a lock.
3631 But if this dentry is later used in d_move, we'd hit uninitialised
3632 list head d_hash, so we just do this to init d_hash field but
3633 leave dentry unhashed.
3636 Frequency : NFS re-export or patchless client
3638 Description: NFS exports has problem with symbolic link
3639 Details : lustre client didn't properly install dentry when re-exported
3640 to NFS or running patchless client.
3643 Frequency : NFS re-export or patchless client
3645 Description: Various nfs/patchless fixes.
3646 Details : fixes reuse disconected alias for lookup process - this fixes
3647 warning "find_exported_dentry: npd != pd",
3648 fix permission error with open files at nfs.
3649 fix apply umask when do revalidate.
3652 Frequency : occasional
3654 Description: Crash on NFS re-export node
3655 Details : calling clear_page() on the wrong pointer triggered oops in
3656 generic_mapping_read().
3659 Frequency : rarely, using O_DIRECT IO
3661 Description: unaligned directio crashes client with LASSERT
3662 Details : check for unaligned buffers before trying any requests.
3665 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3667 Description: stale data returned from RAID cache
3668 Details : If only a small amount of IO is done to the RAID device before
3669 reading it again it is possible to get stale data from the RAID
3670 cache instead of reading it from disk.
3673 Frequency : always for sles10 kernel
3675 Description: sles10 support
3676 Details : ll_follow_link: compile fixes and using of nd_set_link
3677 under newer kernels.
3680 Frequency : depends on arch, kernel and compiler version, always on sles10
3683 Description: recursive or deep enough symlinks cause stack overflow
3684 Details : getting rid of large stack-allocated variable in
3688 Frequency : depends on hardware
3690 Description: lustre write performance loss in the SLES10 kernel
3691 Details : the performance loss is caused by using of write barriers in the
3692 ext3 code. The SLES10 kernel turns barrier support on by
3693 default. The fix is to undo that change for ldiskfs.
3695 ------------------------------------------------------------------------------
3697 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3699 * Support for kernels:
3700 2.6.9-42.0.3EL (RHEL 4)
3701 2.6.5-7.276 (SLES 9)
3702 2.4.21-47.0.1.EL (RHEL 3)
3703 2.6.12.6 vanilla (kernel.org)
3707 Frequency : quota enabled and large files being deleted
3709 Description: releasing more than 4GB of quota at once hangs OST
3710 Details : If a user deletes more than 4GB of files on a single OST it
3711 will cause the OST to spin in an infinite loop. Release
3712 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3717 Description: statfs data retrieved from /proc may be stale or zero
3718 Details : When reading per-device statfs data from /proc, in the
3719 {kbytes,files}_{total,free,avail} files, it may appear
3720 as zero or be out of date.
3723 Frequency : systems with MD RAID1 external journal devices
3725 Description: lconf's call to blkid is confused by RAID1 journal devices
3726 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3727 of returning all block devices that match the journal UUID.
3730 Frequency : always, for aggregate stripe size over 4GB
3732 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3733 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3734 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3737 Frequency : always if "lfs find" used on a local file/directory
3739 Description: "lfs find" segfaults if used on a local file/directory
3740 Details : The case where a directory component was not specified wasn't
3741 handled correctly. Handle this properly.
3744 Frequency : always on ppc64
3746 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3747 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3748 perform endianness conversion.
3751 Frequency : rarely (truncate to non-zero file size after write under load)
3752 Bugzilla : 10730, 10687
3753 Description: Files padded with zeros to next 4K multiple
3754 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3755 it is possible that files that are truncated to a non-zero size
3756 immediately after being written are filled with zero bytes beyond
3757 the truncated size. No file data is lost.
3759 Severity : enhancement
3761 Description: Allow recovery/failover for liblustre clients.
3762 Details : liblustre clients were unaware of failover configurations until
3765 Severity : enhancement
3767 Description: user file locks should fail when not mounting with flock option
3768 Details : Set up an error-returning stub in ll_file_operations.lock field
3769 to prevent incorrect behaviour when client is mounted without
3770 flock option. Also, set up properly f_op->flock field for
3774 Frequency : always on ia64
3776 Description: "lfs df" loops on printing out MDS statfs information
3777 Details : The obd_ioctl_data was not initialized and in some systems
3778 this caused a failure during the ioctl that did not return
3779 an error. Initialize the struct and return an error on failure.
3782 Frequency : SLES 9 only
3784 Description: Error of copying files with lustre special EAs as root
3785 Details : Client side always return success for setxattr call for lustre
3786 special xattr (currently only "trusted.lov").
3789 Frequency : rarely on clusters with both ia64+i386 clients
3791 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3792 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3793 clients are doing concurrent writes to the same file it is
3794 possible that the ia64 clients may LASSERT because the OST
3795 extent locks are not PAGE_SIZE aligned. Ensure that grown
3796 locks are always aligned on the request boundary.
3799 Frequency : specific use, occasional
3801 Description: Overwriting in use executable truncates on-disk binary image
3802 Details : If one node attempts to overwrite an executable in use by
3803 another node, we now correctly return ETXTBSY instead of
3804 truncating the file.
3806 Severity : enhancement
3808 Description: Async OSC create to avoid the blocking unnecessarily.
3809 Details : If a OST has no remain object, system will block on the creating
3810 when need to create a new object on this OST. Now, ways use
3811 pre-created objects when available, instead of blocking on an
3812 empty osc while others are not empty. If we must block, we block
3813 for the shortest possible period of time.
3818 Description: chmod on Lustre root is propagated to other clients
3819 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3820 invalid by the follow_mount time.
3823 Frequency : liblustre clients only
3825 Description: Race in 'instant cancel' lock handling could lead to such locks
3826 never to be granted in case of SMP MDS
3827 Details : Do not destroy not yet granted but cbpending locks in
3831 Frequency : replay/resend of open
3833 Description: non null lock assetion failure in mds_intent_policy
3834 Details : Trying to replay/resend lockless open requests resulted in
3835 mds_open() returning 0 with no lock. Now it sets a flag if
3836 a lock is going to be returned.
3838 Severity : enhancement
3840 Description: Checksum enhancements
3841 Details : New checksum enhancements allow for resending RPCs that failed
3844 Severity : enhancement
3846 Description: Tunables on number of dirty pages in cacche
3847 Details : Allow to set limit on number of dirty pages cached.
3852 Description: client crash on unmount - lock still has references
3853 Details : In some error handling cases it was possible to leak a lock
3854 reference on a client while accessing a file. This was not
3855 harmful to the client during operation, but would cause the
3856 client to crash when the filesystem is unmounted.
3859 Frequency : specific case, rare
3861 Description: ETXTBSY on mds though file not in use by client
3862 Details : ETXTBSY is no longer incorrectly returned when attempting to
3863 chmod or chown a directory that the user previously tried to
3864 execute or a currently-executing binary.
3867 Frequency : extremely rare except on liblustre-based clients
3869 Description: Lustre space not freed when files are deleted
3870 Details : Clean up open-unlinked files after client eviction. Previously
3871 the unlink was skipped and the files remained as orphans.
3876 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3877 Details : In some cases it was possible to send a blocking callback to
3878 a client doing a glimpse, even though that client didn't get
3879 a lock granted. When the glimpse lock is cancelled on the OST
3880 the freed lock is left on the waiting list and corrupted the list.
3883 Frequency : all core dumps
3885 Description: Broke core dumps to lustre
3886 Details : Negative dentry may be unhashed if parent does not have UPDATE
3887 lock, but some callers, e.g. do_coredump, expect dentry to be
3888 hashed after successful create, hash it in ll_create_it.
3890 ------------------------------------------------------------------------------
3892 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3894 * Support for kernels:
3895 2.6.9-42.0.2.EL (RHEL 4)
3896 2.6.5-7.276 (SLES 9)
3897 2.4.21-40.EL (RHEL 3)
3898 2.6.12.6 vanilla (kernel.org)
3902 Frequency : always on RHEL 3
3904 Description: Number of open files grows over time
3905 Details : The number of open files grows over time, whether or not
3906 Lustre is started. This was due to a filp leak introduced
3907 by one of our kernel patches.
3909 ------------------------------------------------------------------------------
3911 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3913 * Support for kernels:
3914 2.6.9-42.EL (RHEL 4)
3915 2.6.5-7.276 (SLES 9)
3916 2.4.21-40.EL (RHEL 3)
3917 2.6.12.6 vanilla (kernel.org)
3922 Bugzilla : 5719, 9635, 9792, 9684
3923 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3924 Details : If a server is under heavy load and cannot reply to new
3925 connection requests before the client resends the (re)connect,
3926 the connection handling code can behave badly if two service
3927 threads are concurrently handing separate (re)connections from
3928 the same client. Add better locking to the connection handling
3929 code, and ensure that only a single connection will be processed
3930 for a given client UUID, even if the lock is dropped.
3932 Severity : enhancement
3934 Description: add TCP zero-copy support to kernel
3935 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3936 sends if the hardware supports scatter-gather and checksumming.
3937 This allows socklnd to do client-write and server-read more
3938 efficiently and reduce CPU utilization from skbuf copying.
3941 Frequency : only if NFS exporting from client
3943 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3944 Details : If setting attributes on a file created under NFS that had
3945 never been opened it would be possible to oops the client
3946 if the file had no objects.
3949 Frequency : always for liblustre
3951 Description: liblustre client does MDS+OSTs setattr RPC for each write
3952 Details : When doing a write from a liblustre client, the client
3953 incorrectly issued an RPC to the MDS and each OST the file was
3954 striped over in order to update the timestamps. When writing
3955 with small chunks and many clients this could overwhelm the MDS
3956 with RPCs. In all cases it would slow down the write because
3957 these RPCs are unnecessary.
3959 Severity : enhancement
3961 Description: allow number of MDS service threads to be changed at module load
3962 Details : It is now possible to change the number of MDS service threads
3963 running. Adding "options mds mds_num_threads={N}" to the MDS's
3964 /etc/modprobe.conf will set the number of threads for the next
3965 time Lustre is restarted (assuming the "mds" module is also
3966 reloaded at that time). The default number of threads will
3967 stay the same, 32 for most systems.
3972 Description: OST crash if filesystem is unformatted or corrupt
3973 Details : If an OST is started on a device that has never been formatted
3974 or if the filesystem is corrupt and cannot even mount then the
3975 error handling cleanup routines would dereference a NULL pointer.
3980 Description: NULL pointer deref in llap_from_page.
3981 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3982 incorrect) mapping if the page was truncated/reclaimed while it was
3983 searched for. Check for this condition and skip such pages when
3984 doing readahead. Introduce extra check to llap_from_page() to
3985 verify page->mapping->host is non-NULL (so page is not anonymous).
3988 Frequency : Sometimes when using sys_sendfile
3990 Description: "page not covered by a lock" warnings from ll_readpage
3991 Details : sendfile called ll_readpage without right page locks present.
3992 Now we introduced ll_file_sendfile that does necessary locking
3993 around call to generic_file_sendfile() much like we do in
3997 Frequency : with certain MDS communication failures at client mount time
3999 Description: NULL pointer deref after failed client mount
4000 Details : a client connection request may delayed by the network layer
4001 and not be sent until after the PTLRPC layer has timed out the
4002 request. If the client fails the mount immediately it will try
4003 to clean up before the network times out the request. Add a
4004 reference from the request import to the obd device and delay
4005 the cleanup until the network drops the request.
4008 Frequency : occasionally during client (re)connect
4010 Description: assertion failure during client (re)connect
4011 Details : processing a client connection request may be delayed by the
4012 client or server longer than the client connect timeout. This
4013 causes the client to resend the connection request. If the
4014 original connection request is replied in this interval, the
4015 client may trip an assertion failure in ptlrpc_connect_interpret()
4016 which thought it would be the only running connect process.
4019 Frequency : only with obd_echo servers and clients that are rebooted
4021 Description: kernel BUG accessing uninitialized data structure
4022 Details : When running an obd_echo server it did not start the ping_evictor
4023 thread, and when a client was evicted an uninitialized data
4024 structure was accessed. Start the ping_evictor in the RPC
4025 service startup instead of the OBD startup.
4027 Severity : enhancement
4028 Bugzilla : 10193 (patchless)
4029 Description: Remove dependency on various unexported kernel interfaces.
4030 Details : No longer need reparent_to_init, exit_mm, exit_files,
4031 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4034 Frequency : rare (only users of deprecated and unsupported LDAP config)
4036 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4037 Details : LDAP apparently contains 'lustreName' attributes instead of
4038 'name'. A simple remapping of the name is sufficient.
4041 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4043 Description: waiting_locks_callback trips kernel BUG if client is evicted
4044 Details : Running with the dump_on_timeout debug flag turned on makes
4045 it possible that the waiting_locks_callback() can try to dump
4046 the Lustre kernel debug logs from an interrupt handler. Defer
4047 this log dumping to the expired_lock_main() thread.
4049 Severity : enhancement
4051 Description: Support NFS exporting on 2.6 kernels.
4052 Details : Implement non-rawops metadata methods for NFS server to use without
4053 changing NFS server code.
4056 Frequency : very rare (synthetic metadata workload only)
4058 Description: two racing renames might cause an MDS thread to deadlock
4059 Details : Running the "racer" program may cause one MDS thread to rename
4060 a file from being the source of a rename to being the target of
4061 a rename at exactly the same time that another thread is doing
4062 so, and the second thread has already enqueued these locks after
4063 doing a lookup of the target and is trying to relock them in
4064 order. Ensure that we don't try to re-lock the same resource.
4067 Frequency : only very large systems with liblustre clients
4069 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4070 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4071 all OSTs in parallel. This allows the eviction of stale liblustre
4072 clients to proceed much faster than if they were done in series,
4073 and also offers similar improvements for other set_info RPCs.
4078 Description: excessive CPU usage during initial read phase on client
4079 Details : During the initial read phase on a client, it would agressively
4080 retry readahead on the file, consuming too much CPU and impacting
4081 performance (since 1.4.5.8). Improve the readahead algorithm
4082 to avoid this, and also improve some other common cases (read
4083 of small files in particular, where "small" is files smaller than
4084 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4089 Description: MDS crash when receiving packet with unknown intent.
4090 Details : Do not LBUG in unknown intent case, just return -EFAULT
4092 Severity : enhancement
4093 Bugzilla : 9293, 9385
4094 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4095 Details : Do not serialize getattr (non-intent version) and statfs.
4098 Frequency : occasional, when OST network is overloaded/intermittent
4100 Description: client evicted by OST after bulk IO timeout
4101 Details : If a client sends a bulk IO request (read or write) the OST
4102 may evict the client if it is unresposive to its data GET/PUT
4103 request. This is incorrect if the network is overloaded (takes
4104 too long to transfer the RPC data) or dropped the OST GET/PUT
4105 request. There is no need to evict the client at all, since
4106 the pinger and/or lock callbacks will handle this, and the
4107 client can restart the bulk request.
4110 Frequency : Always when mmapping file with no objects
4112 Description: client crashes when mmapping file with no objects
4113 Details : Check that we actually have objects in a file before doing any
4114 operations on objects in ll_vm_open, ll_vm_close and
4120 Description: Request leak when working with deleted CWD
4121 Details : Introduce advanced request refcount tracking for requests
4122 referenced from lustre intent.
4124 Severity : Enhancement
4126 Description: Cache open file handles on client.
4127 Details : MDS now will return special lock along with openhandle, if
4128 requested and client is allowed to hold openhandle, even if unused,
4129 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4130 closing files for every read/write openration.
4132 Severity : Enhancement
4134 Description: Cache open negative dentries on client when possible.
4135 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4136 negative dentries on lock revocation.
4141 Description: Remounting a client read-only wasn't possible with a zconf mount
4142 Details : It wasn't possible to remount a client read-only with llmount.
4144 Severity : enhancement
4145 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4146 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4147 MPICH is included to add Lustre-specific ADIO interfaces.
4148 This is based closely on the UFS ADIO layer and only differs
4149 in file creation, in order to allow the OST striping to be set.
4150 This is user-contributed code and not supported by CFS.
4155 Description: extended inode attributes (immutable, append-only) work improperly
4156 when 2.4 and 2.6 kernels are used on client/server or vice versa
4157 Details : Introduce kernel-independent values for these flags.
4159 Severity : enhancement
4162 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4163 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4164 granularity. Now, Fractional MB granularity is be supported,
4165 this is very useful for low memory system.
4167 Severity : enhancement
4169 Description: Getattr by fid
4170 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4171 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4172 have locks and avoids getattr rpc requests for it on every stat.
4175 Frequency : Always, for filesystems larger than 2TB
4177 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4178 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4179 filesystems larger than 2TB to be mounted with mballoc enabled.
4182 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4184 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4185 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4186 filesystems larger than 2TB is mounted then the kernel will
4187 silently corrupt the start of the filesystem. CONFIG_LBD is
4188 enabled for all CFS-supported kernels, but the possibility of
4189 this happening with a modified kernel config exists.
4191 Severity : enhancement
4193 Description: add client O_DIRECT support for 2.6 kernels
4194 Details : It is now possible to do O_DIRECT reads and writes to files
4195 in the Lustre client mountpoint on 2.6 kernel clients.
4197 Severity : enhancement
4199 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4200 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4201 parallel, not waiting for response from every OST before sending
4202 a rpc to the next OST.
4207 Description: setattr vs write race when updating file timestamps
4208 Details : Client processes that update a file timestamp into the past
4209 right after writing to the file (e.g. tar) it is possible that
4210 the updated file modification time can be reset to the current
4211 time due to a race between processing the setattr and write RPC.
4213 Severity : enhancement
4215 Description: Bring 'lfs find' closer in line with regular Linux find.
4216 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4217 -print0 options and obtains all the needed info through the lustre
4220 Severity : enhancement
4222 Description: support up to 1024 configured devices on one node
4223 Details : change obd_dev array from statically allocated to dynamically
4224 allocated structs as they are first used to reduce memory usage
4229 Description: Flush dirty partially truncated pages during truncate
4230 Details : Immediatelly flush partially truncated pages in filter_setattr,
4231 this way we completely avoid having any pages in page cache on OST
4232 and can retire ugly workarounds during writes to flush such pages.
4237 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4238 Details : protect inode from truncation within vfs_unlink() context
4239 just take a reference before calling vfs_unlink() and release it
4240 when parent's i_sem is free.
4245 Description: last_id value checked outside lock on OST caused LASSERT failure
4246 Details : If there were multiple MDS->OST object precreate requests in
4247 flight, it was possible that the OST's last object id was checked
4248 outside a lock and incorrectly tripped an assertion. Move checks
4249 inside locks, and discard old precreate requests.
4252 Frequency : always, if extents are used on OSTs
4254 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4255 Details : OSTs using the extents format would not zero the high 16 bits of
4256 the index physical block number. This is not a problem for any
4257 OST filesystems smaller than 16TB, and no kernels support ext3
4258 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4259 new/modified files) and can be fixed for existing filesystems
4260 with e2fsprogs-1.39-cfs1.
4265 Description: import connection selection may be incorrect if timer wraps
4266 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4267 import connections to be ignored if the 32-bit jiffies counter
4268 wraps. Use a 64-bit jiffies counter.
4271 Frequency : very large clusters immediately after boot
4273 Description: LNET request buffers exhausted under heavy short-term load
4274 Details : If a large number of client requests are generated on a service
4275 that has previously never seen so many requests it is possible
4276 that the request buffer growth cannot keep up with the spike in
4277 demand. Instead of dropping incoming requests, they are held in
4278 the LND until the RPC service can accept more requests.
4281 Frequency : Sometimes during replay
4283 Description: Assertion failure in ll_local_open after replay.
4284 Details : If replay happened on an open request reply before we were able
4285 to set replay handler, reply will become not swabbed tripping the
4286 assertion in ll_local_open. Now we set the handler right after
4287 recognising of open request
4290 Frequency : very rare
4292 Description: kernel reports "badness in vsnprintf"
4293 Details : Reading from the "recovery_status" /proc file in small chunks
4294 may cause a negative length in lprocfs_obd_rd_recovery_status()
4295 call to vsnprintf() (which is otherwise harmless). Exit early
4296 if there is no more space in the output buffer.
4298 Severity : enhancement
4300 Description: clear OBD RPC statistics by writing to them
4301 Details : It is now possible to clear the OBD RPC statistics by writing
4302 to the "stats" file.
4307 Description: Client mtime is not the same on different clients after utimes
4308 Details : In some cases, the client was using the utimes() syscall on
4309 a file cached on another node. The clients now validate the
4310 ctime from the MDS + OSTs to determine which one is right.
4315 Description: Inability to activate failout mode
4316 Details : lconf script incorrectly assumed that in pythong string's numeric
4317 value is used in comparisons.
4320 Frequency : always with multiple stripes per file
4322 Description: Inefficient object allocation for mutli-stripe files
4323 Details : When selecting which OSTs to stripe files over, for files with
4324 a stripe count that divides evenly into the number of OSTs,
4325 the MDS is always picking the same starting OST for each file.
4326 Return the OST selection heuristic to the original design.
4331 Description: mount failures may take full timeout to return an error
4332 Details : Under some heavy load conditions it is possible that a
4333 failed mount can wait for the full obd_timeout interval,
4334 possibly several minutes, before reporting an error.
4335 Instead return an error as soon as the status is known.
4337 Frequency : quota enabled and large files being deleted
4339 Description: releasing more than 4GB of quota at once hangs OST
4340 Details : If a user deletes more than 4GB of files on a single OST it
4341 will cause the OST to spin in an infinite loop. Release
4342 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4347 Description: statfs data retrieved from /proc may be stale or zero
4348 Details : When reading per-device statfs data from /proc, in the
4349 {kbytes,files}_{total,free,avail} files, it may appear
4350 as zero or be out of date.
4353 Frequency : systems with MD RAID1 external journal devices
4355 Description: lconf's call to blkid is confused by RAID1 journal devices
4356 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4357 of returning all block devices that match the journal UUID.
4360 Frequency : always, for aggregate stripe size over 4GB
4362 Description: assertion fails when trying to use 4GB stripe size
4363 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4364 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4367 Frequency : always on ppc64
4369 Description: the first write on an ext3 filesystem with mballoc got stuck
4370 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4371 perform endianness conversion.
4373 ------------------------------------------------------------------------------
4375 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4377 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4378 INTEROPERATE with older versions automatically. Please read the
4379 user documentation before upgrading any part of a live system.
4380 * WARNING: Lustre networking configuration changes are required with
4381 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4384 * Support for kernels:
4385 2.6.9-22.0.2.EL (RHEL 4)
4386 2.6.5-7.244 (SLES 9)
4387 2.6.12.6 vanilla (kernel.org)
4390 Severity : enhancement
4391 Bugzilla : 7981/8208
4392 Description: Introduced Lustre Networking (LNET)
4393 Details : LNET is new networking infrastructure for Lustre, it includes
4394 a reorganized network configuration mode (see the user
4395 documentation for full details) as well as support for routing
4396 between different network fabrics. Lustre Networking Devices
4397 (LNDs) for the supported network fabrics have also been
4398 created for this new infrastructure.
4400 Severity : enhancement
4401 Description: Introduced Access control lists
4402 Details : clients can set ACLs on files and directories in order to have
4403 more fine-grained permissions than the standard Unix UGO+RWX.
4404 The MDS must be started with the "-o acl" mount option.
4406 Severity : enhancement
4407 Description: Introduced filesystem quotas
4408 Details : Administrators may now establish per-user quotas on the
4411 Severity : enhancement
4413 Description: Configuration change for the XT3
4414 The PTLLND is now used to run Lustre over Portals on the XT3
4415 The configure option(s) --with-cray-portals are no longer used.
4416 Rather --with-portals=<path-to-portals-includes> is used to
4417 enable building on the XT3. In addition to enable XT3 specific
4418 features the option --enable-cray-xt3 must be used.
4423 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4424 Details : A race between allocating a new llcd and re-getting the llcd_lock
4425 allowed another thread to grab newly-allocated llcd.
4427 Severity : enhancement
4429 Description: 2.6 OST async journal commit and locking fix to improve performance
4430 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4431 2.6 kernels are now async as they already were in 2.4 kernels so
4432 that they can commit concurrently with the network bulk transfer.
4433 For block-allocated files the filter allocation semaphore is held
4434 to avoid filesystem fragmentation during allocation. BKL lock
4435 removed for 2.6 xattr operations where it is no longer needed.
4440 Description: lconf incorrectly determined whether two IP networks could talk
4441 Details : In some more complicated routing and multiple-network
4442 configurations, lconf will avoid trying to make a network
4443 connection to a disjoint part of the IP space. It was doing the
4444 math incorrectly for one set of cases.
4449 Description: Fix for potential infinite loop processing records in an llog.
4450 Details : If an llog record is corrupted/zeroed, it is possible to loop
4451 forever in llog_process(). Validate the llog record length
4452 and skip the remainder of the block on error.
4455 Frequency : occasional (liblustre only)
4457 Description: liblustre could not open files whose last component is a symlink
4458 Details : sysio_path_walk() would incorrectly pass the open intent to
4459 intermediate path components.
4462 Frequency : rare (liblustre only with non-standard tuning)
4463 Bugzilla : 7201 (7350)
4464 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4465 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4466 after it has been released, only do this for referenced locks
4468 Severity : enhancement
4470 Description: specifying an (invalid) directory default stripe_size of -1
4471 would reset the directory default striping
4472 Details : stripe_size -1 was used internally to signal directory stripe
4473 removal, now use "all default" to signal dir stripe removal
4474 as a directory striping of "all default" is not useful
4477 Frequency : common for large clusters running liblustre clients
4479 Description: doing an ls when liblustre clients are running is slow
4480 Details : sending a glimpse AST to a liblustre client waits for every AST
4481 to time out, as liblustre clients will not respond. Since they
4482 cannot cache data we refresh the OST lock LVB from disk instead.
4484 Severity : enhancement
4486 Description: doing an ls at the same time as file IO can be slow
4487 Details : enqueue and other "small" requests can be blocked behind many
4488 large IO requests. Create a new OST IO portal for non-IO
4489 requests so they can be processed faster.
4492 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4494 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4495 created file with mode 000
4496 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4497 allow an ftruncate() as a non-root user to a file with mode 000.
4498 Lustre now respects this flag to disable mode checks when
4499 truncating a file owned by the user
4502 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4505 Description: Revoking locks from clients that went dead or catatonic might take
4507 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4508 cancellation of such locks instant on servers without waiting for
4509 any reply from clients. Clients drops these locks when cancel
4510 notification from server is received without replying.
4513 Frequency : liblustre-only, when liblustre client dies or becomes busy
4515 Description: Doing ls on Linux clients can take a long time with active
4517 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4518 granting such locks to it in the first place if possible. Locks
4519 are taken by proxy on the OST during the read or write and
4520 dropped immediately afterward. Add connect flags handling, do
4521 not grant locks to liblustre clients for glimpse ASTs.
4523 Severity : enhancement
4525 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4526 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4527 sequential reads and behaves badly with applications doing random
4528 reads. Improve it by reading ahead at least read region, and
4529 avoiding excessive large RPC for small reads.
4531 Severity : enhancement
4533 Description: Creating more than 1000 files for a single job may cause a load
4534 imbalance on the OSTs if there are also a large number of OSTs.
4535 Details : qos_prep_create() uses an OST index reseed value that is an
4536 even multiple of the number of available OSTs so that if the
4537 reseed happens in the middle of the object allocation it will
4538 still utilize the OSTs as uniformly as possible.
4543 Description: OST or MDS may oops in ping_evictor_main()
4544 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4545 but doesn't restart at beginning of obd_exports_timed list
4548 Severity : enhancement
4550 Description: improve by-nid export eviction on the MDS and OST
4551 Details : allow multiple exports with the same NID to be evicted at one
4552 time without re-searching the exports list.
4555 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4557 Description: MDS may oops in groups_free()
4558 Details : in rare race conditions a newly allocated group_info struct is
4559 freed again, and this can be NULL. The 2.4 compatibility code
4560 for groups_free() checked for a NULL pointer, but 2.6 did not.
4563 Frequency : common for liblustre clients doing little filesystem IO
4564 Bugzilla : 9352, 7313
4565 Description: server may evict liblustre clients accessing contended locks
4566 Details : if a client is granted a lock or receives a completion AST
4567 with a blocking AST already set it would not reply to the AST
4568 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4571 Frequency : lfs setstripe, only systems with more than 160 OSTs
4573 Description: unable to set striping with a starting offset beyond OST 160
4574 Details : llapi_create_file() incorrectly limited the starting stripe
4575 index to the maximum single-file stripe count.
4578 Frequency : LDAP users only
4580 Description: lconf did not handle in-kernel recovery with LDAP properly
4581 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4583 Severity : enhancement
4585 Description: bind OST threads to NUMA nodes to improve performance
4586 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4587 node and do their allocations there to localize memory access
4589 Severity : enhancement
4591 Description: llmount can determine client NID directly from Myrinet (GM)
4592 Details : the client NID code from gmnalnid was moved directly into
4593 llmount, removing the need to use this or specifying the
4594 client NID explicitly when mounting GM clients with zeroconf
4597 Frequency : if client is started with down MDS
4599 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4600 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4601 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4602 ptlrpc_import_delayed_req() because we didn't check if the
4603 request was interrupted, and we also didn't break out of the
4604 event loop if there was no timeout
4609 Description: data loss during non-page-aligned writes to a single file from
4610 both multiple nodes and multiple threads on one node at same time
4611 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4612 inconsistency led to false short-reads, that were cached and later
4613 used by ->prepare_write() to fill in partially written page,
4614 leading to data loss.
4617 Frequency : always, if lconf --abort_recovery used
4619 Description: lconf --abort_recovery fails with 'Operation not supported'
4620 Details : lconf was attempting to abort recovery on the MDT device and not
4623 Severity : enhancement
4625 Description: remove cleanup logs
4626 Details : replace lconf-generated cleanup logs with lustre internal
4627 cleanup routines. Eliminates the need for client-cleanup and
4630 Severity : enhancement
4632 Description: add support for EAs (user and system) on lustre filesystems
4633 Details : it is now possible to store extended attributes in the Lustre
4634 client filesystem, and with the user_xattr mount option it
4635 is possible to allow users to store EAs on their files also
4637 Severity : enhancement
4639 Description: Add possibility (config option) to show minimal available OST free
4641 Details : When compiled with --enable-mindf configure option, statfs(2)
4642 (and so, df) will return least minimal free space available from
4643 all OSTs as amount of free space on FS, instead of summary of
4644 free spaces of all OSTs.
4646 Severity : enhancement
4648 Description: do not expand extent locks acquired on OST-side
4649 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4650 by server: they are not cached anyway.
4653 Frequency : when mmap is used/binaries executed from Lustre
4655 Description: Unmmap pages before throwing them away from read cache.
4656 Details : llap_shrink cache now attempts to unmap pages before discarding
4657 them (if unmapping failed - do not discard). SLES9 kernel has
4658 extra checks that trigger if this unmapping is not done first.
4663 Description: lconf didn't resolve symlinks before checking to see whether a
4664 given mountpoint was already in use
4667 Frequency : when migrating failover services
4668 Bugzilla : 6395, 9514
4669 Description: When migrating a subset of services from a node (e.g. failback
4670 from a failover service node) the remaining services would
4671 time out and evict clients.
4672 Details : lconf --force (implied by --failover) sets the global obd_timeout
4673 to 5 seconds in order to quickly disconnect, but this caused
4674 other RPCs to time out too quickly. Do not change the global
4675 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4677 Severity : enhancement
4678 Frequency : if MDS is started with down OST
4679 Bugzilla : 9439,5706
4680 Description: Allow startup/shutdown of an MDS without depending on the
4681 availability of the OSTs.
4682 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4683 Add appropriate locking and lov-osc refcounts for safe
4684 cleaning. Add osc abort_inflight calls in case the
4685 synchronize never started.
4688 Frequency : occasional (Cray XT3 only)
4690 Description: root not authorized to access files in CRAY_PORTALS environment
4691 Details : The client process capabilities were not honoured on the MDS in
4692 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4693 been accessed by an authorized user then root was able to access
4694 the file on the local client also. The root user capabilities
4695 are now allowed on the MDS, as this environment has secure UID.
4698 Frequency : occasional
4700 Description: ldiskfs "too long searching" message happens too often
4701 Details : A debugging message (otherwise harmless) prints too often on
4702 the OST console. This has been reduced to only happen when
4703 there are fragmentation problems on the filesystem.
4708 Description: Division by zero in statfs when all OSCs are inactive
4709 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4710 lov_statfs divides by value returned by lov_get_stripecnt().
4714 Bugzilla : 9489, 3273
4715 Description: First write from each client to each OST was only 4kB in size,
4716 to initialize client writeback cache, which caused sub-optimal
4717 RPCs and poor layout on disk for the first writen file.
4718 Details : Clients now request an initial cache grant at (re)connect time
4719 and so that they can start streaming writes to the cache right
4720 away and always do full-sized RPCs if there is enough data.
4721 If the OST is rebooted the client also re-establishes its grant
4722 so that client cached writes will be honoured under the grant.
4727 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4728 Details : Now I/O RPCs go to different portal number and (presumably) fast
4729 lock requests (and glimses) and other RPCs get their own service
4730 threads pool that should be able to service those RPCs
4733 Severity : enhancement
4735 Description: Ability to exchange lustre version between client and servers and
4736 issue warnings at client side if client is too old. Also for
4737 liblustre clients there is ability to refuse connection of too old
4739 Details : New 'version' field is added to connect data structure that is
4740 filled with version info. That info is later checked by server and
4744 Frequency : rare, liblustre only.
4745 Bugzilla : 9296, 9581
4746 Description: Two simultaneous writes from liblustre at offset within same page
4747 might proceed at the same time overwriting eachother with stale
4749 Details : I/O lock withing llu_file_prwv was released too early, before data
4750 actually was hitting the wire. Extended lock-holding time until
4751 server acknowledges receiving data.
4754 Frequency : extremely rare. Never observed in practice.
4756 Description: avoid generating lustre_handle cookie of 0.
4757 Details : class_handle_hash() generates handle cookies by incrementing
4758 global counter, and can hit 0 occasionaly (this is unlikely, but
4759 not impossible, because initial value of cookie counter is
4760 selected randonly). Value of 0 is used as a sentinel meaning
4761 "unassigned handle" --- avoid it. Also coalesce two critical
4762 sections in this function into one.
4764 Severity : enhancement
4766 Description: allow liblustre clients to delegate truncate locking to OST
4767 Details : To avoid overhead of locking, liblustre client instructs OST to
4768 take extent lock in ost_punch() on client's behalf. New connection
4769 flag is added to handle backward compatibility.
4771 Severity : enhancement
4772 Bugzilla : 4928, 7341, 9758
4773 Description: allow number of OST service threads to be specified
4774 Details : a module parameter allows the number of OST service threads
4775 to be specified via "options ost ost_num_threads={N}" in the
4776 OSS's /etc/modules.conf or /etc/modprobe.conf.
4780 Bugzilla : 6146, 9635, 9895
4781 Description: servers crash with bad pointer in target_handle_connect()
4782 Details : In rare cases when a client is reconnecting it was possible that
4783 the connection request was the last reference for that export.
4784 We would temporarily drop the export reference and get a new
4785 one, but this may have been the last reference and the export
4786 was just destroyed. Get new reference before dropping old one.
4788 Severity : enhancement
4789 Frequency : if client is started with failover MDS
4791 Description: Allow multiple MDS hostnames in the mount command
4792 Details : Try to read the configuration from all specified MDS
4793 hostnames during a client mount in case the "primary"
4796 Severity : enhancement
4798 Description: Stop sending data to evicted clients as soon as possible.
4799 Details : Check if the client we are about to send or are sending data to
4800 was evicted already. (Check is done every second of waiting,
4801 for which l_wait_event interface was extended to allow checking
4802 of exit condition at specified intervals).
4805 Frequency : rare, normally only when NFS exporting is done from client
4807 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4809 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4812 Frequency : always, if srand() is called before liblustre initialization
4814 Description: Liblustre uses system PRNG disturbing its usage by user application
4815 Details : Introduce internal to lustre fast and high-quality PRNG for
4816 lustre usage and make liblustre and some other places in generic
4817 lustre code to use it.
4819 Severity : enhancement
4820 Bugzilla : 9477, 9557, 9870
4821 Description: Verify that the MDS configuration logs are updated when xml is
4822 Details : Check if the .xml configuration logs are newer than the config
4823 logs stored on the MDS and report an error if this is the case.
4824 Request --write-conf, or allow starting with --old_conf.
4826 Severity : enhancement
4828 Description: Handle symlinks in the path when checking if Lustre is mounted.
4829 Details : Resolve intermediate symlinks when checking if a client has
4830 mounted a filesystem to avoid duplicate client mounts.
4835 Description: lconf can hit an error exception but still return success.
4836 Details : The lconf command catches the Command error exception at the top
4837 level script context and will exit with the associated exit
4838 status, but doesn't ensure that this exit status is non-zero.
4843 Description: failure of ptlrpc thread startup can cause oops
4844 Details : Starting a ptlrpc service thread can fail if there are a large
4845 number of threads or the server memory is very fragmented.
4846 Handle this without oopsing.
4849 Frequency : always, only if liblustre and non-default acceptor port was used
4851 Description: liblustre cannot connect to servers with non-default acceptor port
4852 Details : tcpnal_set_default_params() was not called and was therefore
4853 ignoring the environment varaible TCPNAL_PORT, as well as other
4854 TCPNAL_ environment variables
4859 Description: two objects could be created on the same OST for a single file
4860 Details : If an OST is down, in some cases it was possible to create two
4861 objects on a single OST for a single file. No problems other
4862 than potential performance impact and spurious error messages.
4866 Bugzilla : 5681, 9562
4867 Description: Client may oops in ll_unhash_aliases
4868 Details : Client dcache may become inconsistent in race condition.
4869 In some cases "getcwd" can fail if the current directory is
4875 Description: Inode refcounting problems in NFS export code
4876 Details : link_raw functions used to call d_instantiate without obtaining
4877 extra inode reference first.
4881 Bugzilla : 9942, 9903
4882 Description: Referencing freed requests leading to crash, memleaks with NFS.
4883 Details : We used to require that call to ll_revalidate_it was always
4884 followed by ll_lookup_it. Also with revalidate_special() it is
4885 possible to call ll_revalidate_it() twice for the same dentry
4886 even if first occurence returned success. This fix changes semantic
4887 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4888 reference on a request referred from the intent.
4889 ll_intent_release() then releases such a request.
4892 Frequency : rare, normally benchmark loads only
4894 Description: unlinked inodes were kept in memory on the client
4895 Details : If a client is repeatedly creating and unlinking files it
4896 can accumulate a lot of stale inodes in the inode slab cache.
4897 If there is no other client load running this can cause the
4898 client node to run out of memory. Instead flush old inodes
4899 from client cache that have the same inode number as a new inode.
4902 Frequency : SLES9 2.6.5 kernel and long filenames only
4903 Bugzilla : 9969, 10379
4904 Description: utime reports stale NFS file handle
4905 Details : SLES9 uses out-of-dentry names in some cases, which confused
4906 the lustre dentry revalidation. Change it to always use the
4910 Frequency : rare, unless heavy write-truncate concurrency is continuous
4911 Bugzilla : 4180, 6984, 7171, 9963, 9331
4912 Description: OST becomes very slow and/or deadlocked during object unlink
4913 Details : filter_destroy() was holding onto the parent directory lock
4914 while truncating+unlinking objects. For very large objects this
4915 may block other threads for a long time and slow overall OST
4916 responsiveness. It may also be possible to get a lock ordering
4917 deadlock in this case, or run out of journal credits because of
4918 the combined truncate+unlink. Solution is to do object truncate
4919 first in one transaction without parent lock, and then do the
4920 final unlink in a new transaction with the parent lock. This
4921 reduces the lock hold time dramatically.
4924 Frequency : rare, 2.4 kernels only
4926 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4927 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4928 before unmounting their filesystem in order to avoid deadlock.
4929 The kernel_locked() function in 2.4 kernels only checks whether
4930 the kernel lock is held, not whether it is this process that is
4931 holding it as 2.6 kernels do.
4936 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4937 Details : The client ptlrpc code may be trying to reconnect to a down
4938 server before a previous connection attempt has timed out.
4939 Increase the reconnect interval to be longer than the connection
4940 timeout interval to avoid sending duplicate connections to
4944 Frequency : echo_client brw_test command
4946 Description: fix echo_client to work with OST preallocated code
4947 Details : OST preallocation code (5137) didn't take echo_client IO path
4948 into account: echo_client calls filter methods outside of any
4949 OST thread and, hence, there is no per-thread preallocated
4950 pages and buffers to use. Solution: hijack pga pages for IO. As
4951 a byproduct, this avoids unnecessary data copying.
4955 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4956 Description: Client can oops in mdc_commit_close() after open replay
4957 Details : It was possible for the MDS to return an open request with no
4958 transaction number in mds_finish_transno() if the client was
4959 evicted, but without actually returning an error. Clients
4960 would later try to replay that open and may trip an assertion
4961 Simplify the client close codepath, and always return an error
4962 from the MDS in case the open is not successful.
4965 Frequency : rare, 2.6 OSTs only
4967 Description: OST may deadlock under high load on fragmented files
4968 Details : If there was a heavy load and highly-fragmented OST filesystems
4969 it was possible to have all the OST threads deadlock waiting on
4970 allocation of biovecs, because the biovecs were not released
4971 until the entire RPC IO was completed. Instead, release biovecs
4972 as soon as they are complete to ensure forward IO progress.
4974 Severity : enhancement
4976 Description: Support for specifying external journal device at mount
4977 Details : If an OST or MDS device is formatted with an external journal
4978 device, this device major/minor is stored in the ext3 superblock
4979 and may not be valid for failover. Allow detecting and
4980 specifying the external journal at mount time.
4985 Description: Mounting an MDS with pending unlinked files may cause oops
4986 Details : target_finish_recovery() calls mds_postrecov() which returned
4987 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4988 considers this an error and immediately begins cleaning up the
4989 lov, just after starting the mds_lov process
4991 Severity : enhancement
4993 Description: Implement 'lfs df' to report actual free space on per-OST basis
4994 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4995 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4996 usage of objects; '-h' to report in human readable format.
4998 ------------------------------------------------------------------------------
5000 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5007 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5008 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5009 at startup during orphan unlinking. Instead, skip buddy update
5010 before setup, as it will be regenerated after recovery is complete.
5015 Description: If an OST is inactive, its locks might reference stale inodes.
5016 Details : lov_change_cbdata() must iterate over all namespaces, even if
5017 they are inactive to clear inode references from the lock.
5019 Severity : enhancement
5020 Frequency : occasional, if non-standard max_dirty_mb used
5022 Description: Client will block write RPCs if not enough grant
5023 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5024 then the client will block writes while waiting for another RPC
5025 to complete instead of consuming its dirty limit. With change
5026 we get improved performance when max_dirty_mb is small.
5028 Severity : enhancement
5029 Bugzilla : 3389, 6253
5030 Description: Add support for supplementary groups on the MDS.
5031 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5032 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5033 lookups for user supplementary groups into a cache.
5037 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5038 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5039 for liblustre O_EXCL is enforced by MDS.
5043 Description: readdir never returns NULL in liblustre.
5044 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5045 dentry in struct dirent.
5049 Frequency : liblustre only, depends on application IO pattern
5050 Description: liblustre clients evicted if not contacting servers
5051 Details : Don't put liblustre clients into the ping_evictor list, so
5052 they will not be evicted by the pinger ever.
5054 Severity : enhancement
5056 Description: Add ability to evict clients by NID from MDS.
5057 Details : By echoing "nid:$NID" string into
5058 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5059 $NID would be instantly evicted from this MDS and from all active
5060 OSTs connected to it.
5064 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5065 Details : lookup_it_finish() used to query file size from OSTs that was not
5070 Description: service threads change working directory to that of init
5071 Details : Starting lustre service threads may pin the working directory
5072 of the parent thread, making that filesystem busy. Threads
5073 now change to the working directory of init to avoid this.
5077 Frequency : during shutdown only
5078 Description: shutdown with a failed MDS or OST can cause unmount to hang
5079 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5084 Frequency : 2.6 only
5085 Description: chmod/chown may include an extra supplementary group
5086 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5087 supplementary group and if none is specified this is used.
5090 Bugzilla : 5479 (6816)
5092 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5093 Details : If lookup is in progress on a file that is unlinked we might try
5094 to revalidate the inode and fail in revalidate after lookup is
5095 complete and ll_file_open() enqueues the open again but
5096 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5099 Frequency : always, if lconf --abort_recovery used
5101 Description: lconf --abort_recovery fails with 'Operation not supported'
5102 Details : lconf was attempting to abort recovery on the MDT device and not
5105 ------------------------------------------------------------------------------
5107 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5112 Frequency : rare (only unsupported configurations with a node running as an
5114 Bugzilla : 6514, 5137
5115 Description: Mounting a Lustre file system on a node running as an OST could
5117 Details : OSTs now preallocates memory needed to write out data at
5118 startup, instead of when needed, to avoid having to
5119 allocate memory in possibly low memory situations.
5120 Specifically, if the file system is mounted on on OST,
5121 memory pressure could force it to try to write out data,
5122 which it needed to allocate memory to do. Due to the low
5123 memory, it would be unable to do so and the node would
5124 become unresponsive.
5126 Severity : enhancement
5128 Description: Addition of lconf --service command line option
5129 Details : lconf now accepts a '--service <arg>' option, which is
5130 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5132 Severity : enhancement
5134 Description: Failover mode is now the default for OSTs.
5135 Details : By default, OSTs will now run in failover mode. To return to
5136 the old behaviour, add '--failout' to the lmc line for OSTs.
5138 Severity : enhancement
5140 Description: Health checks are now provided for MDS and OSTs
5141 Details : Additional detailed health check information on MSD and OSTs
5142 is now provided through the procfs health_check value.
5145 Frequency : occasional, depends on IO load
5147 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5148 after numerous create/delete cycles
5149 Details : The ext3 inode allocation policy would not allocate new inodes
5150 very well on the OSTs because there are no new directories
5151 being created. Instead we look for groups with free space if
5152 the parent directories are nearly full.
5157 Description: Network or server problems during mount may cause partially
5158 mounted clients instead of returning an error.
5159 Details : The config llog parsing code may overwrite the error return
5160 code during mount error handling, returning success instead
5166 Description: MDS can fail to allocate large reply buffers
5167 Details : After long uptimes the MDS can fail to allocate large reply
5168 buffers (e.g. zconf client mount config records) due to memory
5169 fragmentation or consumption by the buffer cache. Preallocate
5170 some large reply buffers so that these replies can be sent even
5171 under memory pressure.
5175 Frequency : rare (liblustre)
5176 Description: fsx running with liblustre complained that using truncate() to
5177 extend the file doesn't work. This patch corrects that issue.
5178 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5179 ATTR_SIZE and lsm use in llu_setattr_raw.
5183 Frequency : rare, only 2.6 kernels
5184 Description: Unusual file access patterns on the MDS may result in inode
5185 data being lost in very rare circumstances.
5186 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5187 code in the 2.6 kernel caused the inode and/or EA data not to
5188 be read from disk, causing single-file corruption.
5192 Frequency : rare, only 2.6 filesystems using extents
5193 Description: Heavy concurrent write and delete load may cause data corruption.
5194 Details : It was possible under high-load situations to have an extent
5195 metadata block in the block device cache from a just-unlinked
5196 file overwrite a newly-allocated data block. We now unmap any
5197 metadata buffers that alias just-allocated data blocks.
5201 Frequency : filesystems with default stripe_count larger than 77
5202 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5203 Details : lconf specifies an inode size of 4096 bytes when the default
5204 stripe_count is larger than 77. This conflicts with the default
5205 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5206 this case to avoid pinning too much memory for large EAs.
5208 ------------------------------------------------------------------------------
5210 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5215 Frequency : rare (extremely heavy IO load with hundreds of clients)
5217 Description: Client is evicted, gets IO error writing to file
5218 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5219 caused clients to be evicted for AST timeouts. The fixes in
5220 bug 5192 mean we no longer need to have such short AST timeouts
5221 so ldlm_timeout has been increased.
5224 Frequency : occasional during --force or --failover shutdown under load
5225 Bugzilla : 5949, 4834
5226 Description: Server oops/LBUG if stopped with --force or --failover under load
5227 Details : a collection of import/export refcount and cleanup ordering
5228 issues fixed for safer force cleanup
5231 Frequency : only filesystems larger than 120 OSTs
5232 Bugzilla : 5990, 6223
5233 Description: lfs getstripe would oops on a very large filesystem
5234 Details : lov_getconfig used kfree on vmalloc'd memory
5237 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5238 Bugzilla : 6242, 6243
5239 Description: reading from files that had been truncated to a non-zero size
5240 but never opened returned no data
5241 Details : ll_file_read() reads zeros from no-object files to EOF
5246 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5247 Details : There's little harm in aborting MDS/OSS recovery and letting it
5248 try again, so I removed the LASSERT and return an error instead.
5250 Severity : enhancement
5252 Description: New debugging infrastructure for tracking down data corruption
5253 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5254 (b) cover more of the client-side code path, and (c) try to narrow
5255 down where problems occurred
5259 Bugzilla : 3819, 4364, 4397, 6313
5260 Description: Racing close and eviction MDS could cause assertion in mds_close
5261 Details : It was possible to get multiple mfd references during close and
5262 client eviction, leading to one thread referencing a freed mfd.
5264 Severity: : enhancement
5265 Bugzilla : 3262, 6359
5266 Description: Attempts to reconnect to servers are now more aggressive.
5267 Details : This builds on the enhanced upcall-less recovery that was added
5268 in 1.4.2. When trying to reconnect to servers, clients will
5269 now try each server in the failover group every 10 seconds. By
5270 default, clients would previously try one server every 25 seconds.
5275 Description: After recovery, certain operations trigger a failed
5276 assertion on a client.
5277 Details : Failing over an mds, using lconf -d --failover, while a
5278 client was doing a readdir() call would cause the client to
5279 LBUG after recovery completed and the readdir() was resent.
5281 Severity : enhancement
5283 Description: Default groups are now added by lconf
5284 Details : You can now run lconf --group <servicename> without having to
5285 manually add groups with lmc.
5288 Frequency : occasional
5290 Description: Nodes with an elan id of 0 trigger a failed assertion
5293 Frequency : always when accessing e.g. tty/console device nodes
5295 Description: tty and some other devices nodes cannot be used on lustre
5296 Details : file's private_data field is used by device data and lustre
5297 values in there got lost. New field was added to struct file to
5298 store fs-specific private data.
5301 Frequency : when exporting Lustre via NFS
5303 Description: NFSD failed occasionally when looking up a path component
5304 Details : NFSD is looking up ".." which was broken in ext3 directories
5305 that had grown large enough to become hashed.
5308 Frequency : Clusters with multiple interfaces not on the same subnet
5310 Description: Nodes will repeatedly try to reconnect to an interface which it
5311 cannot reach and report an error to the log.
5312 Details : Extra peer list entries will be created by lconf with some peers
5313 unreachable. lconf now validates the peer before adding it.
5316 Frequency : Only if a default stripe is set on the filesystem root.
5318 Description: Setting a default stripe on the filesystem root prevented the
5319 filesystem from being remounted.
5320 Details : The client was sending extra request flags in the root getattr
5321 request and did not allocate a reply buffer for the dir EA.
5324 Frequency : occasional, higher if lots of files are accessed by one client
5325 Bugzilla : 6159, 6097
5326 Description: Client trips assertion regarding lsm mismatch/magic
5327 Details : While revalidating inodes the VFS looks up inodes with ifind()
5328 and in rare cases can find an inode that is being freed.
5329 The ll_test_inode() code will free the lsm during ifind()
5330 when it finds an existing inode and then the VFS later attaches
5331 this free lsm to a new inode.
5335 Bugzilla : 6422, 7030
5336 Description: MDS deadlock between mkdir and client eviction
5337 Details : Creating a new file via mkdir or mknod (starting a transaction
5338 and getting the ns lock) can deadlock with client eviction
5339 (gets ns lock and trying to finish a synchronous transaction).
5342 Frequency : occasional
5343 Description: While starting a server, the fsfilt_ext3 module could not be
5345 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5346 kernels. Previously, lconf would still use the ext3 name
5347 when trying to load modules. Now, it will correctly use
5348 ext3 on 2.4 and ldiskfs on 2.6.
5350 Severity : enhancement
5351 Description: The default stripe count has been changed to 1
5352 Details : The interpretation of the default stripe count (0, to lfs
5353 or lmc) has been changed to mean striping across a single
5354 OST, rather than all available. For general usage we have
5355 found a stripe count of 1 or 2 works best.
5357 Severity : enhancement
5358 Description: Add support for compiling against Cray portals.
5359 Details : Conditional compiling for some areas that are different
5363 Frequency : occasional
5364 Bugzilla : 6409, 6834
5365 Description: Creating files with an explicit stripe count may lead to
5366 a failed assertion on the MDS
5367 Details : If some OSTs are full or unavailable, creating files may
5368 trigger a failed assertion on the MDS. Now, Lustre will
5369 try to use other servers or return an error to the
5373 Frequency : occasional
5375 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5376 caused lock timeout during readahead (since 1.4.2).
5377 Details : Processes doing readahead might match a lock that hasn't been
5378 granted yet if there are overlapping and conflicting lock
5379 requests. The readahead process waits on ungranted lock
5380 (original lock is CBPENDING), while OST waits for that process
5381 to cancel CBPENDING read lock and eventually evicts client.
5383 Severity : enhancement
5385 Description: Initial enabling of flock support for clients
5386 Details : Implements fcntl advisory locking and file status functions.
5387 This feature is provided as an optional mount flag (default
5388 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5389 locking are implemented yet, and those that are are not guaranteed
5390 to be completely correct in production environments.
5391 mount -t lustre -o [flock|noflock] ...
5394 Frequency : occasional
5396 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5397 assertion in the ext3 JBD (journaling) layer.
5398 Details : The b_committed_data struct is protected by the big kernel lock
5399 in 2.4 kernels, serializing journal_commit_transaction() and
5400 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5401 access to this struct. In 2.6 kernels there is finer grained
5402 locking to improve SMP performance of the JBD layer.
5406 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5411 Description: merge and cleanup kernel patches.
5414 -----------------------------------------------------------------------------
5416 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5418 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5419 versions of Lustre. Please update all servers and clients to
5420 version 1.4.2 or later at the same time. You must also run
5421 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5422 to update the configuration logs.
5424 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5425 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5426 - swab llog records of type '0' so we get proper header size/idx (5861)
5427 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5428 - fix rename of one directory over another leaking an inode (5953)
5429 - avoid SetPageDirty on 2.6 (5981)
5430 - don't re-add just-being-destroyed locks to the waiting list (5653)
5431 - when creating new directories, inherit the parent's custom
5432 striping settings if present parent (3048)
5433 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5434 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5435 - don't hold client locks on temporary worklist from l_lru (5666)
5436 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5437 - automatically evict dead clients (5921)
5438 - Update file size properly in create+truncate+fstat case (6196)
5439 - Do not unhash mountpoint dentries, do not allow removal of
5441 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5442 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5443 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5444 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5445 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5446 - Correctly handle reads of files with no objects (6243)
5447 - lctl recover will also mark a device active if deactivate used (5933)
5449 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5450 - allow --write-conf on an MDS with different nettype than client (5619)
5451 - don't write config llogs to MDS for mounts not from that MDS (5617)
5452 - lconf should create multiple TCP connections from a client (5201)
5453 - init scripts are now turned off by default; run chkconfig --on
5454 lustre and chkconfig --on lustrefs to use them
5455 - upcalls are no longer needed for clients to recover to failover
5457 - add --abort-recovery option to lconf to abort recovery on device
5459 - add support for an arbitrary number of OSTs (3026)
5460 - Quota support protocol changes.
5461 - forward compatibility changes to wire structs (6007)
5462 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5463 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5464 - improved "lustre status" script
5465 - initialize blocksize for non-regular files (6062)
5466 - added --disable-server and --disable-client configure options (5782)
5467 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5468 - Vanilla 2.4.29 support
5469 - increase maximum number of obd devices to 520 (6242)
5470 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5471 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5472 SLES 9 2.6.5 kernels
5473 - limit stripes per file to 160 (the maximum EA size) (6093)
5475 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5478 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5479 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5480 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5481 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5482 - small ext3 extents cleanups and fixes (5733)
5483 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5484 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5485 - avoid needless client->OST connect, fix handle mismatch (5317)
5486 - fix DLM error path that led to out-of-sync client, long delays (5779)
5487 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5488 - fix several locking issues related to i_size (5492,5624,5654,5672)
5489 - don't move pending lock onto export if it is already evicted (5683)
5490 - fix kernel oops when creating .foo in unlinked directory (5548)
5491 - fix deadlock in obdfilter statistics vs. object create (5811)
5492 - use time_{before,after} to avoid timer jiffies wrap (5882)
5493 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5494 - Do not leak request if server was not able to process it (5154)
5495 - If mds_open unable to find parent dir, make that negative lookup(5154)
5496 - don't create new directories with extent-mapping (5909, 5936)
5498 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5499 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5500 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5501 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5502 - support for RHEL4 kernel on i686 (5773)
5503 - provide error messages when incompatible logs are encountered (5898)
5505 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5506 * version 1.4.0.10 (1.4.1 release candidate 1)
5508 - don't keep a lock reference when lock is not granted (4238)
5509 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5510 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5511 - drop import inflight refcount on signal_completed_replay error (5255)
5512 - unlock page after async write error during send (3677)
5513 - handle missing objects in filter_preprw_read properly (5265)
5514 - no transno return for symlink open, don't save no-trasno open (3440)
5515 - don't try to complete elan receive that already failed (4012)
5516 - free RPC server reply state on error (5406)
5517 - clean up thread from ptlrpc_start_thread() on error (5160)
5518 - readahead could read extra page into cache that wasn't ejected (5388)
5519 - prevent races in class_attach/setup/cleanup/detach (5260)
5520 - don't dereference de->d_inode after l_dput of de (5458)
5521 - use "int" for stripe value returned from lock_to_stripe (5544)
5522 - mballoc allocation and error-checking fixes in 2.6 (5504)
5523 - block device patches to fix I/O request sizes in 2.6 (5482)
5524 - look up hostnames for IB nals (5602)
5525 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5526 - don't start multiple acceptors for the same port (5277)
5527 - fix incorrect LASSERT in mds_getattr_name (5635)
5528 - export a proc file for general "ping" checking (5628)
5529 - fix "lfs check" to not block when the MDS is down (5628)
5531 - service request history (4965)
5532 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5533 - create an "evict_client" /proc entry on OSTs, like the MDS has
5534 - fix mount usage message, return errors per mount(8) (5168)
5535 - change grep [] to grep "[]" in tests so they work in more UMLs
5536 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5537 - remove ancient LOV_MAGIC_V0 EA support (5047)
5538 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5539 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5540 - allow readahead allocations to fail when low on memory (5383)
5541 - mmap locking landed again, after considerable improvement (2828)
5542 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5544 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5547 - send OST transaction number in read/write reply to free req (4966)
5548 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5549 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5550 - the watchdog thread now runs as interruptible (5246)
5551 - flock/lockf fixes (but it's still disabled, pending 5135)
5552 - don't use EXT3 constants in llite code (5094)
5553 - memory shortage at startup could cause assertion (5176)
5555 - reorganization of lov code
5556 - single portals codebase
5558 - add extents/mballoc support (5025)
5559 - direct I/O reads in the obdfilter (4048)
5560 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5562 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5565 - send OST transaction number in read/write reply to free req (4966)
5566 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5567 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5568 - the watchdog thread now runs as interruptible (5246)
5569 - handle missing objects in filter_preprw_read properly (5265)
5570 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5571 - drop import inflight refcount on signal_completed_replay error (5255)
5572 - unlock page after async write error during send (3677)
5573 - return original error code on reconstructed replies (3761)
5574 - no transno return for symlink open, don't save no-trasno open (3440)
5576 - add pid to ldlm debugging output (4922)
5577 - bump the watchdog timeouts -- we can't handle 30sec yet
5578 - extra debugging for orphan dentry/inode bug (5259)
5580 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5583 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5584 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5585 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5586 - fix NULL dereference in /proc/sys/portals/routes (4827)
5587 - allow failed mdc_close() operations to be interrupted (4561)
5588 - stop precreate on OST before MDS would time out on it (4778)
5589 - don't send partial-page writes before EOF from client (4410)
5590 - discard client grant for sub-page writes on large-page clients (4520)
5591 - don't free dentries not owned by NFS code, check generation (4806)
5592 - fix lsm leak if mds_create_objects() fails (4801)
5593 - limit debug_daemon file size, always print CERROR messages (4789)
5594 - use transno after validating reply (3892)
5595 - process timed out requests if import state changes (3754)
5596 - update mtime on OST during writes, return in glimpse (4829)
5597 - add mkfsoptions to LDAP (4679)
5598 - use ->max_readahead method instead of zapping global ra (5039)
5599 - don't interrupt __l_wait_event() during strace
5601 - add software watchdogs to catch hung threads quickly (4941)
5602 - make lustrefs init script start after nfs is mounted
5603 - fix CWARN/ERROR duplication (4930)
5604 - return async write errors to application if possible (2248)
5605 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5606 - print NAL number in %x format (4645)
5607 - update barely-supported suse-2.4.21-171 series (4842)
5608 - support for sles 9 %post scripts
5609 - support for building 2.6 kernel-source packages
5610 - support for sles km_* packages
5612 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5615 - ignore -ENOENT errors in osc_destroy (3639)
5616 - notify osc create thread that OSC is being cleaned up (4600)
5617 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5618 - reconstruct ost_handle() like mds_handle() (4657)
5619 - create a new thread to do import eviction to avoid deadlock (3969)
5620 - let lconf resolve symlinked-to devices (4629)
5621 - don't unlink "objects" from directory with default EA (4554)
5622 - hold socknal file ref over connect in case target is down (4394)
5623 - allow more than 32000 subdirectories in a single directory (3244)
5624 - fix blocks count for O_DIRECT writes (3751)
5625 - OST returns ENOSPC from object create when no space left (4539)
5626 - don't send truncate RPC if file size isn't changing (4410)
5627 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5628 - bind to privileged port in socknal and tcpnal (3689)
5630 - rate limit CERROR/CWARN console message to avoid overload (4519)
5631 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5632 - basic mmap support (3918)
5633 - kernel patch series update from b1_4 (4711)
5635 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5638 - avoid crash during MDS cleanup with OST shut down (2775)
5639 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5640 - don't use bad inodes on the MDS (3744)
5641 - dynamic object preallocation to improve recovery speed (4236)
5642 - don't hold spinlock over lock dumping or change debug flags (4401)
5643 - don't zero obd_dev when it is force cleaned (3651)
5644 - print grants to console if they go negative (4431)
5645 - "lctl deactivate" will stop automatic recovery attempts (3406)
5646 - look for existing locks in ldlm_handle_enqueue() (3764)
5647 - don't resolve lock handle twice in recovery avoiding race (4401)
5648 - revalidate should check working dir is a directory (4134)
5650 - don't always mark "slow" obdfilter messages as errors (4418)
5652 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5655 - don't close LustreDB during write_conf until it is done (3860)
5656 - fix typo in lconf for_each_profile (3821)
5657 - allow dumping logs from multiple threads at one time (3820)
5658 - don't allow multiple threads in OSC recovery (3812)
5659 - fix debug_size parameters (3864)
5660 - fix mds_postrecov to initialize import for llog ctxt (3121)
5661 - replace config semaphore with spinlock (3306)
5662 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5663 - don't allow enqueue to complete on a destroyed export (3822)
5664 - down write_lock before checking llog header bitmap (3825)
5665 - recover from lock replay timeout (3764)
5666 - up llog sem before sending rpc (3652)
5667 - reduce ns lock hold times when setting kms (3267)
5668 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5669 - fix NULL deref and obd_dev leak on setup error (3312)
5670 - replace some LBUG about llog ops with error handling (3841)
5671 - don't match INVALID dentries from d_lookup and spin (3784)
5672 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5673 - fix invalid assertion in ptlrpc_set_wait (3880)
5675 - add libwrap support for the TCP acceptor (3996)
5676 - add /proc/sys/portals/routes for non-root route listing (3994)
5677 - allow setting MDS UUID in .xml (2580)
5678 - print the stack of a process that LBUGs (4228)
5680 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5683 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5684 - make sure to unset replay flag from failed open requests (3440)
5685 - if default stripe count is 0, use OST count for inode size (3636)
5686 - update parent mtime/ctime on client for create/unlink (2611)
5687 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5688 - free recovery state on server during a forced cleanup (3571)
5689 - unregister_reply for resent reqs (3063)
5690 - loop back devices mounting and status check on 2.6 (3563)
5691 - fix resource-creation race that can provoke i_size == 0 (3513)
5692 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5693 - more debugging for page-accounting assertion (3746)
5694 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5695 - avoid deadlock after precreation failure (3758)
5696 - fix race and lock order deadlock in orphan handling (3450, 3750)
5697 - add validity checks when grabbing inodes from l_ast_data (3599)
5699 - add /proc/.../recovery_status to obdfilter (3428)
5700 - lightweight CDEBUG infrastructure, debug daemon (3668)
5701 - change default OSC RPC parameters to be better on small clusters
5702 - turn off OST read cache for files smaller than 32MB
5703 - install man pages and include them in rpms (3100)
5704 - add new init script for (un)mounting lustre filesystems (2593)
5705 - run chkconfig in %post for init scripts (3701)
5706 - drop scimac NAL (unmaintained)
5708 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5711 - clean kiobufs before and after use (3485)
5712 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5713 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5714 - change init script to fail more clearly if not run as root (1528)
5715 - allow clients to reconnect during replay (1742)
5716 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5717 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5718 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5719 - fix cond_resched (3554)
5720 - only evict unfinished clients after recovery (3515)
5721 - allow bulk resend, prevent data loss (3570)
5722 - dynamic ptlrpc request buffer allocation (2102)
5723 - don't allow unlinking open directory if it isn't empty (2904)
5724 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5725 - remove extraneous obd dereference causing LASSERT failure (3334)
5726 - don't use get_cycles() when creating temp. files on the mds (3156)
5727 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5728 - handle EEXIST for set-stripe, set proper directory name (3336)
5730 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5731 - fix ksocknal_fmb_callback() error messages (2918)
5733 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5736 - don't copy lvb into (possibly NULL) reply on error (2983)
5737 - don't deref dentry after dput, don't free lvb on error (2922)
5738 - use the kms to determine writeback rpc length (2947)
5739 - increment oti_logcookies when osc is inactive (2948)
5740 - update client's i_blocks count via lvb messages (2543)
5741 - handle intent open/close of special files properly (1557)
5742 - mount MDS with errors=remount-ro, like obdfilter (2009)
5743 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5744 - don't use cancelling-locks' kms values (2947)
5745 - use highest lock extent for kms, not last one (2925)
5746 - don't dereference ERR_PTR() dentry in error handling path (3107)
5747 - fix thread race in portals_debug_dumplog() (3122)
5748 - create lprocfs device entries at setup instead of at attach (1519)
5749 - common AST error handler, don't evict client on completion race (3145)
5750 - zero nameidata in detach_mnt in 2.6 (3118)
5751 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5752 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5753 - initialize RPC timeout timer earlier for 2.6 (3219)
5754 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5755 - print nal/nid for unknown nid (3258)
5756 - additional checks for oscc recovery before doing precreate (3284)
5757 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5758 - don't crash in mdc_close for bad permissions on open (3285)
5759 - zero i_rdev for non-device files (3147)
5760 - clear page->private before handing to FS, better assertion (3119)
5761 - tune the read pipeline (3236)
5762 - fix incorrect decref of invalidated dentry (2350)
5763 - provide read-ahead stats and refine rpc in flight stats (3328)
5764 - don't hold journal transaction open across create RPC (3313)
5765 - update atime on MDS at close time (3265)
5766 - close LDAP connection when recovering to avoid server load (3315)
5767 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5768 - don't leak open file on MDS after open resend (3325)
5769 - serialize filter_precreate and filter_destroy_precreated (3329)
5770 - loop device shouldn't call sync_dev() for nul device (3092)
5771 - clear page cache after eviction (2766)
5772 - resynchronize MDS->OST in background (2824)
5773 - refuse to mount the same filesystem twice on same mountpoint (3394)
5774 - allow llmount to create routes for mounting behind routers (3320)
5775 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5776 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5777 - fix rare problem with rename on htree directories (3417)
5779 - allow default OST striping configuration per directory (1414)
5780 - fix compilation for qswnal for 2.6 kernels (3125)
5781 - increase maximum number of MDS request buffers for large systems
5782 - change liblustreapi to be useful for external progs like lfsck (3098)
5783 - increase local configuration timeout for slow disks (3353)
5784 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5786 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5789 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5790 - don't overwrite extent policy data in reply if lock was blocked (2901)
5791 - drop filter export grants atomically with removal from device (2663)
5792 - del obd_self_export from work_list in class_disconnect_exports (2908)
5793 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5794 - swab reply message in mdc_close, other PPC fixes (2464)
5795 - fix destroying of named logs (2325)
5796 - overwrite old logs when running lconf --write_conf (2264)
5797 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5798 - fix race in target_handle_connect (2898)
5799 - mds_reint_create() should take same inode create lock (2926)
5800 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5801 - don't close files for self_export to avoid uninitialized obd (2936)
5802 - allow MDS with the same name as client node (2939)
5803 - hold dentry reference for closed log files for unlink (2325)
5804 - reserve space for all logs during transactions (2059)
5805 - don't evict page beyond end of stripe extent (2925)
5806 - don't oops on a deleted current working directory (2399)
5807 - handle hard links to targets without a parent properly (2517)
5808 - don't dereference NULL lock when racing during eviction (2867)
5809 - don't grow lock extents when lots of conflicting locks (2919)
5811 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5814 - account for cache space usage on clients to avoid data loss (974)
5815 - lfsck support in lustre kernel code (2349)
5816 - reduce journal credits needed for BRW writes (2370)
5817 - orphan handling to avoid losing space on client/server crashes
5818 - ptlrpcd can be blocked, stopping ALL progress (2477)
5819 - use lock value blocks to assist in proper KMS, faster stat (1021)
5820 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5821 - recovery for initial connections (2355)
5822 - fixes for mds_cleanup_orphans (1934)
5823 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5824 - block all file creations until orphan recovery completes (1901)
5825 - client remove rq_connection from request struct (2423)
5826 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5827 - recovery timer race (2670)
5828 - mdc_close recovey bug (2532)
5829 - ptlrpc cleanup bug (2710)
5830 - mds timeout on local locks (2588)
5831 - namespace lock held during RPCs (2431)
5832 - handle interrupted sync write properly (2503)
5833 - don't try to handle a message that hasn't been replied to (2699)
5834 - client assert failure during cleanup after abort recovery (2701)
5835 - leak mdc device after failed mount (2712)
5836 - ptlrpc_check_set allows timedout requests to complete (2714)
5837 - wait for inflight reqs when ptlrpcd finishes (2710)
5838 - make sure unregistered services are removed from the srv_list
5839 - reset bulk XID's when resending them (caught by 1138 test)
5840 - unregister_bulk after timeout
5841 - fix lconf error (2694)
5842 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5843 - readahead locks pages, leaves pending causing memory pressure (2673)
5844 - increase OST request buffers to 4096 on large machines (2729)
5845 - fix up permission of existing directories in simple_mkdir (2661)
5846 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5847 - don't assign transno to errored transactions (2742)
5848 - don't delete objects on OST if given a bogus objid from MDS (2751)
5849 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5850 - if rq_no_resend, then timeout request after recovery (2432)
5851 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5852 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5853 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5854 - MDS recovery shouldn't delete objects that it has given out (2730)
5855 - if enqueue arrives after completion, don't clobber LVB (2819)
5856 - don't unlock pages twice when trigger_group_io returns error (2814)
5857 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5858 - don't write pages to disk if there was an error (1450)
5859 - don't ping imports that have recovery disabled (2676)
5860 - take buffered bytes into account when balancing socknal conn (2817)
5861 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5862 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5863 - remove little-endian swabbing of llog records (1987)
5864 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5865 - retry reposting request buffers if they fail (1191)
5866 - grow extent at grant time to avoid granting a revoked lock (2809)
5867 - lock revoke doesn't evict page if covered by a second lock (2765)
5868 - disable VM readahead to avoid reading outside lock extents (2805)
5870 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5871 - updated kernel patches for hp-2.4.20 kernel (2681)
5873 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5876 - fix truncated write corruption (2366)
5877 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5879 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5880 - ensure len(uuid) < 37 in lmc (1171)
5881 - fix ia64 OOPS in llog_test (2255)
5882 - zero end of page at obdfilter for partial page writes (2648)
5883 - don't leave stale dentries around after renames (bug 2428)
5884 - fix timeouts when evicting a client with a single lock held (2642)
5885 - set deadline for the initial HELLO message to drain (2634)
5886 - print out dotted-quad IP addresses in the socknal (2302)
5888 - additional debugging for MDS client eviction problem (2443)
5889 - fix mkfsoptions support for osts (2603, 2604)
5891 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5894 - add series for the vanilla 2.6.0 kernel
5895 - add series for the vanilla 2.4.24 kernel
5896 - add series for a cray x86/64 UL kernel drop
5897 - fix xattr patches for the vanilla 2.4.19 series
5899 - generate true UUIDs in lmc (1171)
5900 - have portals stack dumping break in UML (2466)
5901 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5902 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5903 - wake callback waiting threads on client eviction (2460)
5904 - Add --ptldebug and --subsystem to lmc (1719)
5905 - update assertion to allow safe interrupt allocation
5906 - set rq_no_resend for cancel requests (2432)
5907 - recalculate ptlrpcd timeout after resend (2494)
5908 - call vfs_rmdir when removing pending directories (2368)
5909 - fix renaming a file to itself (2429)
5910 - lmc creates a default one-stripe lov (2454)
5911 - expand procfs space to handle large clusters (2326)
5912 - increase UML stack to avoid overflow
5913 - update lconf's list of debug and subsystem masks
5914 - fix lfs find --obd (2510)
5915 - /proc tunable for disabling filter read caching (2591)
5916 - stop rpm packages from altering slapd.conf (2301)
5917 - disable nagle in the socknal under 0conf (2578)
5918 - choose mds inode size based on stripe count (2572)
5919 - fix kernel-source rpm problems (2516)
5921 - add --disable-doc to avoid pdf generation (2421)
5922 - update documentation, tests, type-os, comments
5923 - avoid format warnings on ia64
5924 - remove the TOE NAL
5925 - tiny code cleanups by removing unused fields
5927 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5930 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5931 - remove the most blatant lies from BUILDING file (2371)
5932 - change default debug level to reasonable production setting
5933 - reduce client side cache size to reduce cache flush time
5934 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5935 - make TCP zerocopy and pinger support enabled by default (2476)
5936 - sync writes completed after process exits caused crashes (2319)
5937 - maintain correct mount count on the MDS (2356)
5938 - backout 1557, because 2316 wasn't really fixed
5939 - better file I/O statistics gathering in /proc
5940 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5941 - another kernel patch to fix zero-copy TCP function export
5942 - don't take duplicate lock when processing re-sent getattr (2420)
5943 - lctl uses obd_self_export instead of creating new conn (2353)
5944 - MDS/OST recovery case which requires object creation asserted (2425)
5945 - move lfs from /usr/sbin to /usr/bin in packages
5946 - fix race between mds_client_add and mds_client_free (2417)
5947 - use kmalloc instead of slabs in portals (2430)
5948 - don't create duplicate records when a failover MDS is present (2442)
5949 - remove unnecessary mount age check (2332)
5950 - don't remove directory inodes from locks prematurely (2451)
5951 - don't break if MDS service name is the same as hostname (2103)
5952 - fix races in client write RPC generation when cache full (2482)
5954 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5957 - remove now-unused request->rq_obd (278)
5958 - if an allocation fails, print out how much memory we've used (1933)
5959 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5960 - add the "configurable stack size" patch to most series files (1256)
5961 - ability to write large log records, for 100+ OST configs (2306)
5962 - fix NULL deref when filter_prep fails (2314)
5963 - fix operator precedence error in filter_sync
5964 - dynamic allocation of socknal TX descriptors (2315)
5965 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5966 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5967 - ensure that CWARN messages in Portals always get to the syslog
5968 - __init/__exit are not for prototype decls (ldlm_init/exit)
5969 - x86-64 compile warning fixes
5970 - fix gateway LMC keyword conflict (2318)
5971 - fix MDS lock inversions in getattr/reint paths (1844)
5972 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5973 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5974 - DLM race condition prevented some lock evictions (2328)
5975 - ENOMEM detection and retry on socknal sends (2230)
5976 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5977 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5978 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5979 - better error messages when a client is rejected during recovery (1505)
5980 - avoid cancelling locks which were never granted, after failure (2330)
5981 - fix i_sem/journal inversion in mds_client_add (2333)
5982 - fix truncate/getattr lock cycle deadlock (2334)
5983 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5984 - fix two rare exit paths which could leak an l_lock() ref (2321)
5985 - fencepost error in MDS/OST orphan recovery (2226)
5986 - make log record alignment 8 bytes (1988)
5987 - lstripe now fails when requested offset > ost_count (2237)
5988 - ensure that all kernel series have a complete list.h (1607)
5989 - fix crashes in special-file operations (2316)
5990 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5992 - allow configurable automake binary, for testing new versions
5993 - small update to the lfs documentation
5995 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5997 * fix negative export reference count in fsfilt_sync (2312)
5999 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6000 * release candidate 0.9.1
6002 - orphans are moved into the PENDING directory for possible recovery
6003 - replayed opens now open by fid for orphan/rename safety (1042)
6004 - last close of an orphan inode generates a transno (683)
6005 - chdir() and mount() now pin the directory entry (1020)
6006 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6007 - discard very old requests without processing them (1502)
6008 - remove some common, well-understood CERRORs (1505)
6009 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6010 - clear "grant" flags in OST replies until OST grant code lands (1644)
6011 - fix read performance by not clobbering i_blksize on client (1598)
6012 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6013 - make LRU size a /proc tunable, clears locks when reduced (707)
6014 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6015 - server locks take references on exports now (1558)
6016 - build fixes for 2.4.20-rh trees (1663)
6017 - return an error from lov_create if all OSCs are inactive (1751)
6018 - fix import levels when a reconnect happens without a timeout (1597)
6019 - exit early from mds_open if we get a lookup error (1749)
6020 - partial page read at EOF wouldn't wait for disk before sending (1642)
6021 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6022 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6023 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6024 - write/truncate lock inversion (1639)
6025 - Don't auto-load obdclass, portals modules during cleanup (1495)
6026 - fix timestamps from jumping to "now" (1763)
6027 - extra journal assertions (1648)
6028 - add an extra multiunlink test (1771)
6029 - fix read_record/write_record API (1776)
6030 - fix leak of offset_extent, possible incorrect i_size later (1772)
6031 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6032 - Debugging for the kqswnal_get_idle_tx problems (1820)
6033 - Allow recovery to be attempted multiple times (1536)
6034 - Write out MDS last_rcvd file after it is first created (1600)
6035 - Fix tx_descriptor leak in failed transmit situations (1827)
6036 - ext3 journaling fixes for assertion failure after IO error (1871)
6037 - class_export_put() on freed export after completion AST error (1896)
6038 - Fix revalidate looping in VFS (1322)
6039 - Don't access a freed export during MDS_REINT timeout (1521)
6040 - Add open-unlink recovery support on the MDS (1673,1764)
6041 - Return an error if no MDS data was read from last_rcvd (1946)
6042 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6043 - Don't setup a disk device that doesn't match exported UUID (317)
6044 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6045 - avoid committing NULL handle in force close
6046 - local.sh is now a one-stripe LOV configuration
6047 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6048 - fix ext3 htree duplicate directory entry corruption (1516)
6049 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6050 - update to new LOV EA format (2097)
6051 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6053 2003-06-15 Phil Schwan <phil@clusterfs.com>
6056 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6057 - per-import/export recovery handling (958, 931, 959)
6058 - multiple last-rcvd slots, for serving multiple FSes (949)
6059 - connections are again shared between multiple imp/exports (963, 964)
6060 - "umount -f" would hang if any requests needed to be sent (393, 978)
6061 - avoid pinning large req buffer by copying for queued messages (989)
6062 - add "uuid" to "lctl device" command to help upcalls (991)
6063 - "open" RPCs with transnos would confuse recovery counters (1037)
6064 - do proper endian conversion of all wire messages (288, 340, 891)
6065 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6066 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6067 - fix ptlrpc_connection leak in target_handle_connect (1174)
6068 - fix import refcounting bug in OST and MDS cleanup (1134)
6069 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6070 - fix very unlikely obd_types race condition (501)
6071 - remove osc_open hack for echo_client (1187)
6072 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6073 - a failure in read_inode2 leads to deadlock (1139)
6074 - cancel ack-locks as soon as transaction is committed (1072)
6075 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6076 - make sure to commitrw after any preprw to avoid deadlock (1162)
6077 - failing to execute a file in a lustre FS would lock inode (1203)
6078 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6079 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6080 - obd_uuid2tgt would walk off the end of the list (1255)
6081 - on IA64 the osc would give portals incorrect bulk size (1258)
6082 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6083 - fix lock inversion caused by new llite matching code (1282)
6084 - limit the number of dirty pages on a client to 10MB (1286)
6085 - timed out locks were not being corrected cancelled (1289)
6086 - fix O_DIRECT above 4GB on IA-32 (1292)
6087 * major user-visible changes
6088 - fail out/fail over policy now controlled by the upcall (993)
6090 - add OBD_PING to check server availability and failure (954)
6091 - lustre messages are now sent in sending host order (288, 340, 891)
6092 - add eadatalen to MDS getattr reply (340)
6093 - OST read replies may contain second buffer, with per-page status (593)
6095 2003-03-11 Phil Schwan <phil@clusterfs.com>
6098 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6099 - failed open()s could cause deadlock; fixed (867, 869)
6100 - stop cancelling OST locks when files are closed (481)
6101 - overlapping XID spaces caused network corruption (851, 853)
6102 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6103 - setattr_raw added, to do single-RPC, server-side setattrs
6104 - lmc/lconf syntax change for OST UUIDs
6105 - fix crashy race condition between ptlrpc_free_req and osc_close
6106 - don't use request in mdc_enqueue if we hit a timeout (889)
6107 - don't set the inode i_size for regular files from the MDS (896)
6108 - handle out of order completion AST (842)
6109 - don't LBUG if a lock request times out after receiving AST (913)
6110 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6111 - if a bad lock AST arrives, send an error instead of dropping entirely
6112 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6113 - fix leak in bulk IO when only partially completed (899, 900, 926)
6114 - fix O_DIRECT for ia64 (55)
6115 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6116 - C-z after timeout could hang a process forever; fixed (977)
6118 - client-side I/O cache (678, 924, 929, 941, 970)
6120 - READPAGE and SETATTRs which don't take server-side locks get
6123 2003-02-11 Phil Schwan <phil@clusterfs.com>
6126 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6127 - Fix fsfilt_extN_readpage() to read a full page of directory
6128 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6129 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6130 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6131 - Most intent operations were removed, in favour of a new RPC mode
6132 that does a single RPC to the server and bypasses most of the VFS
6133 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6134 - Aggressively cancel local locks on DLM servers
6135 - mds_reint_unlink sends EA to the client if it's the last nlink.
6136 client uses that EA to unlink OST objects.
6137 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6138 - recursive symlinks were fixed (439)
6139 - fixed NULL deref in DEBUG_REQ
6140 - filter_update_lastobjid no longer calls sync, which annoyed extN
6141 - fixed multi-client small-writes to a single file problem (445)
6142 - fixed mtime updates during file writes (607)
6143 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6144 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6145 - fixed Linux OST size reporting problem (444, 656)
6146 - OST now updates object mtime with writes or setattr (607, 619)
6147 - client verifies file size before zeroing page past EOF (445)
6148 - OST now writes last allocated objid to disk with allocation (108)
6149 - LOV on echo now works (409)
6151 - mds_reint_unlink sends a new buffer, with the EA included. this
6152 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6153 set if a regular file was being unlinked, and it was the last link
6154 - use PtlGet from the target for bulk writes (315)
6155 - OST now updates object mtime with writes or setattr (607, 619)
6156 - LDLM now has a grant-time callback to revalidate locked items, if
6158 - Many MDS operations were reorganized to combat race conditions
6160 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6161 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6163 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6166 - Fully reactivate OST imports after reconnection (512, others)
6167 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6168 - More graceful error handling for truncating on dead OST (515)
6169 - Don't error out unless we're actually accessing dead stripes (474)
6170 - Fix garbage sizes when stripes are missing (410)
6171 - LRU counters were broken, causing constant lock purge (433, 432)
6172 - garbage on read from stripes with failed OSTs (441)
6173 - mark OSCs as active before reconnecting during recovery (438)
6174 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6175 - lfind did not preserve OST order in output (443)
6176 - symlinks cause hung clients, incorrect data (439)
6177 - stop dereferencing request after dropping refcount (457)
6178 - don't LASSERT(spin_is_locked) on non-SMP (455)
6179 - fixes for many rename() bugs
6180 - fstat didn't correctly synchronize attributes (399)
6181 - server must handle lock cancellation during blocking AST prep (487)
6182 - bulk descriptors were free()d too soon (511)
6183 - fix paths in lconf, which would load incorrect modules (451, 507)
6184 - fix confusing lconf 'host not found' error message (386)
6185 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6186 - fix race condition in mdc_blocking_ast() for inode access (526)
6187 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6188 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6189 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6190 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6191 - fix lstripe in several ways (kernel side) (527)
6192 - fix request leak in ldlm_cli_enqueue (262)
6193 - incorrect OSC was marked inactive after OST failure
6194 - call mds_fs_cleanup before unmounting filesystem (524)
6195 - fix races between taking ns_lock and ldlm_lock_change_resource
6196 - fix races updating LOV export open file list
6197 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6198 - fix recovery NULL deref in ldlm_cli_cancel_unused
6199 - fix some DLM races by using new hash table for lock handles (419)
6200 - permit the client to specify desired inodes, at replay
6201 - duplicate requests when we queue them for replay reintegration
6202 - fix last_rcvd offset calculation
6203 - sync after each recovered transaction, so we always make progress
6204 - never, not always, ERESTART requests without transnos
6205 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6207 - skip replay if the MDS says that the client is already connected
6208 - don't check for a recovery-enabled export to match lctl's UUID
6209 - don't INC_USE_COUNT for phantom exports
6210 - don't crash when cleaning up phantom exports (567)
6211 - don't double-finish or set replay data for errored mdc_open requests
6212 - abort requests when they time out, so we don't get old replies
6213 - send/receive replies for AST messages again
6214 - if the client says that it doesn't have the lock, cancel it on the
6216 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6217 mark it as destroyed, it will all work out when decref is called
6218 - fix module use counts (22, 581)
6220 - ASTs now expect a reply (server cancels lock on error reply)
6222 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6225 - fix many simultaneous client startup (392)
6226 - fix dentry->d_it clobbering
6227 - credentials weren't being shipped for readdir/getattr operations
6228 - remove invalid assertions triggered during some concurrent MD
6230 - proper Lustre versions added (336, 389)
6231 - fix memory leak for create error case (398)
6232 - fix LOV locking bug that would get cli/srv out of sync
6233 - fix echo client over LOV (409)
6234 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6235 - fix double-O_EXCL intent crash (424)
6236 - avoid sending multiple lock CANCELs (352)
6238 - MDS can do multi-client recovery (modulo bugs in new code)
6240 - many updates, edits, cleanups
6242 2002-11-18 Phil Schwan <phil@clusterfs.com>
6245 - fix null d_it dereference (346)
6246 - fix full OST/dbench hang (333)
6247 - fix permission problem with file removal (286)
6248 - fix removal of OSCs from LOV when they fail
6249 - fix NULL deref during bulk timeout (214)
6250 - fix problems related to multiple filesystems on one MDS (241)
6251 - fixed serious subtle metadata locking bugs
6252 - free locks on clients when inodes are removed due to memory
6254 - fix inode pointer in lock data (285)
6255 - partial support for multiple MDS on a single host (241)
6256 - data locks weren't cancelled at clear_inode time (290, 311)
6257 - intent locks could lead to unbounded lock growth (205)
6258 - added a maximum lock count, an LRU list, and a flusher
6259 - fix multiple rename (365)
6260 - properly abstracted the echo client
6261 - OSC locked 1 byte too many; fixed
6262 - rewrote brw callback code:
6263 - fixed recovery bugs related to LOVs (306)
6264 - fixed too-many-pages-in-one-write crash (191)
6265 - fixed (again) crash in sync_io_timeout (214)
6266 - probably fixed callback-related race (385)
6268 - Add capability to MDS protocol
6269 - LDLM cancellations and callbacks on different portals
6271 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6274 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6275 - defer open object destruction to close time (601981, 138)
6276 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6277 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6278 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6279 - don't call obd_setattr in ll_file_release for destroyed objects
6280 * protocol change to lustre_msg: move |version| and add |flags|
6281 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6282 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6283 * added replay of create, unlink, link and rename operations during
6284 MDS failover; recovery should be much more robust now
6285 * remove failed OSCs from LOVs (only lov_create uses this so far)
6286 * the lustre-HOWTO was brought (more) up to date (582544)
6288 2002-10-23 Phil Schwan <phil@clusterfs.com>
6291 - in-use dentries weren't being reused properly (617851)
6292 - prevent multiple LDLM setup (599178)
6293 - fix LOV size calculations for truncate (617853)
6294 - fix client handling of MDS intent errors (POSIX)
6295 - fix permission bug in lovstripe.c test (624321)
6296 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6297 - truncate past end of file could corrupt data
6298 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6299 - a race in recovery could return ETIMEDOUT to apps (623947)
6300 - building outside the source directory was fixed
6301 * the lustre-HOWTO was brought (more) up to date (582544)
6302 * major progress was made on recovery functionality
6304 2002-10-10 Phil Schwan <phil@clusterfs.com>
6307 - recovery deadlock fix
6308 - rm -rf causes LBUG fix (617817)
6309 - file open by multiple tasks fix (618962)
6310 - directory permissions bugs (602707 and 620007)
6311 - journal_stop fixed with locking (611313)
6312 - O_APPEND failures resolved (618273, perhaps 614459)
6313 - lconf PATH fix (619770)
6314 - IA64 build fix (621450)
6315 - RPC buffer sizes scale with amount of memory
6317 2002-10-01 Phil Schwan <phil@clusterfs.com>
6320 - locks would be cancelled without throwing away data pages,
6321 resulting in inconsistent data (605627)
6322 - inode attributes were not always being refreshed (605627, 612449)
6323 - lconf now continues to cleanup after lctl reports an error
6324 - MDS now enforces user permissions (602707)
6325 - lprocfs cleanup fixed, but not yet enabled (614157)
6326 - fixed infinite server hang, should a client not respond to an AST
6327 - avoid going into recovery if user calls readlink() with a buffer
6328 that's too small (613941)
6329 - AST RPCs no longer require replies (614867) -- this may be changed
6330 - don't crash server if client sends an IOV that's too big (611336)
6331 - fixed lock conversion deadlock (611892)
6332 - fixed the following of symlinks (614622)
6333 * recovery: the server can remove locks from a client that dies, other
6334 clients can make progress
6335 * more extN patch fixes
6336 * compile-time configurable ptlrpc buffer allocations
6338 - collaborative read cache document
6339 - Lustre Lite Performance CDR document-in-progress
6341 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6344 - fix typo in patch-2.4.18
6346 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6349 - clear ptlrpc request each time in handle_incoming_request()
6350 - unlink of files now destroys the object on the OST
6352 2002-09-19 Peter Braam <braam@clusterfs.com>
6354 * add hard link support
6355 * change obdfile creation method
6356 * kernel patch changed
6358 2002-09-19 Peter Braam <braam@clusterfs.com>
6361 - stack overflow bug in extN fixed
6363 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6365 * documentation updates
6366 - add man pages for config tools
6367 - update tests/README to describe testing with new config tools
6368 - finish metadata API descriptions
6369 * bug fixes and cleanups
6370 - statfs workaround for 16TB limit
6371 - LOV stripe allocation improved, can stripe on subset of OSTs
6372 - LOV file size/IO offset was wrong for files > 4GB in size
6373 - object EA data was being dropped, caused files to be unreadable
6374 - memory overflow with non-LOV OST caused memory corruption
6375 - fixed regression tests to work with new config tools, obdfilter
6376 - fixed bug when directory size became larger than 1 block
6377 - fixed bug (for single client case) when PWD was deleted
6378 - invalidate local directory pages when doing intent-based ops
6379 - avoid LDLM oops when lock callback contained bad data
6381 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6383 * documentation updates
6384 * bug fixes and cleanups
6385 - configuration tools
6388 - 64-bit compile warnings
6389 - 64-bit internal statfs data
6391 * test_brw on persistent OST devices
6393 * lprocfs (disabled)
6395 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6397 * documentation updates
6398 * bug fixes and cleanups
6399 * configuration tools
6401 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6404 * many small fixes to 0_5_4
6405 * io/network handling
6406 * thinkos in MDS operations
6408 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6411 * crucial basic fixes to 0.5.3
6412 * IOR, Iozone work over Elan
6415 2002-08-07 Phil Schwan <phil@clusterfs.com>
6416 * version 0_5_3, our first alpha
6417 * we use the new Portals iovs
6418 * documentation updates
6419 * bug fixes and cleanups
6420 * small changes in the DLM wire protocol
6422 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6423 * version 0_5_1 with some initial stability,
6424 * locking on MD and file I/O.
6425 * documentation updates
6426 * several bug fixes since 0.5.0
6427 * small changes in wire protocol
6429 2002-07-18 Phil Schwan <phil@clusterfs.com>
6431 * delivered as Lustre Light Alpha
6432 * fixed a crash after handling invalid MDS requests
6433 * fixed directory pages for architectures with non-4k pages sizes
6435 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6436 * release version v0_4_4
6437 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6439 * Fixes a number of interruption problems with OST operations.
6440 * Update documentation for portals header changes
6441 * Move all wire protocol structs/defines to lustre_idl.h
6442 * Fixes symlink length bug.
6443 * Add tcpdump to repository.
6445 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6446 * release version v0_4_3
6447 * Fixes statfs for inodes on extN.
6448 * Fixes bug in runtests which would delete /etc/hosts.
6449 * Use 64-bit object IDs wherever possible (not into VFS though)
6450 Remove ost_get_info, which is unused by lustre, and out of date.
6452 2002-07-03 Peter Braam <braam@clusterfs.com>
6453 * release version v0_4_2 Fixes a lookup error (type not passed)
6454 * move forward to head of Portals
6455 * move forward to latest Lustre kernel
6457 2002-06-25 Peter Braam <braam@clusterfs.com>
6458 * release version v0_4_1. Hopefully stable on single node use.