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 2.6.32-504.12.2.el6 (RHEL6.6)
8 3.0.101-0.47.52 (SLES11 SP3)
9 * Client known to build on unpatched kernels:
10 2.6.32-431.29.2.el6 (RHEL6.5)
11 2.6.32-504.12.2.el6 (RHEL6.6)
12 3.10.0-229.1.2 (RHEL7.1)
13 3.0.101-0.47.52 (SLES11 SP3)
14 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
15 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
16 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
17 stack overflow. For more information, please refer to bugzilla 17630.
18 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
19 the re-exporting nodes to be mounted with "32bitapi" mount option
21 --------------------------------------------------------------------------------
23 03-10-2015 Intel Corporation
25 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
26 for currently supported client and server kernel versions.
27 * Server known to build on patched kernels:
28 2.6.32-431.29.2.el6 (RHEL6.5)
29 2.6.32-504.8.1.el6 (RHEL6.6)
30 3.0.101-0.46 (SLES11 SP3)
31 * Client known to build on unpatched kernels:
32 2.6.32-431.29.2.el6 (RHEL6.5)
33 2.6.32-504.8.1.el6 (RHEL6.6)
34 3.10.0-123.20.1.el7 (RHEL7)
35 3.0.101-0.46 (SLES11 SP3)
36 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
37 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
38 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
39 stack overflow. For more information, please refer to bugzilla 17630.
40 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
41 the re-exporting nodes to be mounted with "32bitapi" mount option
43 Severity : enhancement
45 Description: control OST-index in IDIF via ROCOMPAT flag.
46 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
47 last_rcvd file. For new formatted OST device, it will be auto set;
48 for the case of upgrading from old OST device, you can enable it
49 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
50 enabled, for new created OST-object, its IDIF-in-LMA will contain
51 the OST-index; for the existing OST-object, the OSD will convert
52 old format IDIF as new format IDIF with OST-index stored in the
53 LMA EA when accessing such OST-object or via OI scrub. Once such
54 flag is enabled, it cannot be reverted back, so the system cannot
55 be downgraded to the orignal incompatible version.
57 --------------------------------------------------------------------------------
59 07-30-2014 Intel Corporation
61 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
62 for currently supported client and server kernel versions.
63 * Server known to build on patched kernels:
64 2.6.32-431.20.3.el6 (RHEL6.5)
65 3.0.101-0.31 (SLES11 SP3)
66 * Client known to build on unpatched kernels:
67 2.6.32-431.20.3.el6 (RHEL6.5)
68 3.0.101-0.31 (SLES11 SP3)
71 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
72 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
73 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
74 stack overflow. For more information, please refer to bugzilla 17630.
75 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
76 the re-exporting nodes to be mounted with "32bitapi" mount option
78 --------------------------------------------------------------------------------
80 10-31-2013 Intel Corporation
82 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
83 * Server builds for kernels:
84 2.6.32-358.18.1.el6 (RHEL6.4)
85 * Client builds for unpatched kernels:
86 2.6.32-358.18.1.el6 (RHEL6.4)
87 3.0.92-0.8 (SLES11 SP3)
89 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
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 --------------------------------------------------------------------------------
98 05-30-2013 Intel Corporation
100 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
101 * Server builds for kernels:
102 2.6.32-358.6.2.el6 (RHEL6.4)
103 * Client builds for unpatched kernels:
104 2.6.32-358.6.2.el6 (RHEL6.4)
105 3.0.74-0.6.6 (SLES11 SP2)
107 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
108 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
109 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
110 stack overflow. For more information, please refer to bugzilla 17630.
111 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
112 the re-exporting nodes to be mounted with "32bitapi" mount option
114 --------------------------------------------------------------------------------
116 9-30-2012 Whamcloud, Inc.
118 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
119 * Server support for kernels:
120 2.6.32-279.5.1.el6 (RHEL6)
121 * Client support for unpatched kernels:
122 2.6.32-279.5.1.el6 (RHEL6)
124 2.6.18-238.19.1.el5 (RHEL5)
125 2.6.16.60-0.69.1 (SLES10)
126 * Recommended e2fsprogs version: 1.41.90.wc4
127 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
128 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
129 stack overflow. For more information, please refer to bugzilla 17630.
130 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
131 the re-exporting nodes to be mounted with "32bitapi" mount option
133 Severity : enhancement
135 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
137 Severity : enhancement
139 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
141 Severity : enhancement
143 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
145 Severity : enhancement
147 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
149 Severity : enhancement
151 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
153 --------------------------------------------------------------------------------
155 04-30-2012 Whamcloud, Inc.
157 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
158 * Server support for kernels:
159 2.6.32-220.4.2.el6 (RHEL6)
160 * Client support for unpatched kernels:
161 2.6.32-220.4.2.el6 (RHEL6)
163 2.6.18-238.19.1.el5 (RHEL5)
164 2.6.16.60-0.69.1 (SLES10)
165 * Recommended e2fsprogs version: 1.41.90.wc4
166 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
167 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
168 stack overflow. For more information, please refer to bugzilla 17630.
169 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
170 the re-exporting nodes to be mounted with "32bitapi" mount option
172 Severity : enhancement
174 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
176 Severity : enhancement
178 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
180 Severity : enhancement
182 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
184 Severity : enhancement
186 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
188 --------------------------------------------------------------------------------
190 09-30-2011 Whamcloud, Inc.
192 * Server support for kernels:
193 2.6.18-238.19.1.el5 (RHEL5)
194 2.6.32-131.6.1.el6 (RHEL6)
195 * Client support for unpatched kernels:
196 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
197 2.6.18-238.19.1.el5 (RHEL5)
198 2.6.32-131.6.1.el6 (RHEL6)
199 2.6.29.4-167.fc11 (FC11)
200 2.6.16.60-0.69.1 (SLES10)
202 * Recommended e2fsprogs version: 1.41.90.wc3
203 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
204 Lustre filesystem with 4K stack may cause a stack overflow. For more
205 information, please refer to bugzilla 17630.
206 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
207 mounted with "32bitapi" mount option
209 Severity : enhancement
210 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
212 Severity : enhancement
213 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
215 Severity : enhancement
216 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
219 Frequency : only with SLES10
221 Description: Use OFED "KMP" provided by Novell
222 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
223 Lustre is now built against this package. That means you need to
224 install the ofed-kmp package from Novell for the patchless client.
225 Note that the ofed-kmp that Novell ships may not exactly match the
226 kernel version but should still be compatible.
228 Severity : enhancement
230 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
231 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
233 --------------------------------------------------------------------------------
235 2010-07-15 Oracle, Inc.
237 * Server support for kernels:
238 2.6.18-164.11.1.el5 (RHEL 5)
239 2.6.18-164.11.1.0.1.el5 (OEL 5)
240 * Client support for unpatched kernels:
241 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
242 2.6.18-164.11.1.el5 (RHEL 5),
243 2.6.18-164.11.1.0.1.el5 (OEL 5)
244 2.6.16.60-0.42.8 (SLES 10),
246 2.6.29.4-167.fc11 (FC11)
247 * Recommended e2fsprogs version: 1.41.10-sun2
249 Severity : enhancement
251 Description: Distribute lfs_migrate script with lustre
252 Details : The lfs_migrate mentioned in the manual is now included in the
253 lustre tree in lustre/scripts and will distributed in lustre RPMs.
255 Severity : enhancement
257 Description: Failover nid list added to import proc file.
259 Severity : enhancement
261 Description: "som_preview" mount option required to use SOM
262 Details : The clients will need to use som_preview mount option
266 Description: Lustre does not do 1MB IOs to HW RAID
267 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
269 Severity : enhancement
271 Description: lfs quota output is not very convenient for awk/sed-parsing
272 Details : Some positions in lfs quota output table could be empty or
273 non-empty which made it hard to parse it with scripts, now
274 a dash is put instead of space where there is not supposed
279 Description: Update to OFED 1.5.1
280 Details : OFED 1.5.1 will be used in environments that don't use
281 vendor supplied OFED.
285 Description: Remove lnet. prefix from lctl params display
286 Details : Also renamed lnet.memused to lnet_memused to
291 Description: Use conf_param -d to remove permanant settings.
295 Description: Add list_param -R
296 Details : lctl list_param now supports recursive mode.
299 Bugzilla : 21829, 22312
300 Description: Fix llobdstat.
301 Details : 1. Now updating the write_bytes counter
302 2. Port script fixes from 1.8.
306 Description: decrease the usage of memory on clients.
307 Details : 1. On clients, recycle dentries and inodes unused.
308 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
313 Description: llobdstat fix and enhancement
314 Details : add a counter to set a limit to how many samples will be returned
315 fix a wildcard in the path to limit to obdfilter stats only
319 Description: lfs check only allowed for root
320 Details : Code cleanups and added a regresion test of non-root use of
325 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
326 Details : Fix hr->hr_index overload.
330 Description: Fix FIEMAP functionality.
334 Description: FC11 patchless client support added
338 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
339 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
343 Description: New acc-sm script and various test-framework enhancments.
347 Description: Previously landed patch was reverted.
351 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
355 Description: Add ldiskfs maxdirsize mount option.
359 Description: Fix mkfs.lustre support for 16TB LUNs.
363 Description: Print mmp_check_interval
364 Details : print mmp_check_interval and make it possible to abort
365 mount operation in case it takes too long.
369 Description: Add cascading_rw to lustre mpi tests.
373 Description: Store SOM data in EA.
374 Details : Also several SOM related cleanups.
378 Description: Add refcount to md_open_data
382 Description: Use refcount on lov_request_set to prevent memory corruption.
386 Description: Statahead: use parent dir's i_mutex for synchronization.
390 Description: loadgen improvements.
394 Description: Enable peer health detection in ksocklnd and ko2iblnd.
398 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
402 Description: Fix race on obd_devcie:md_stats
406 Description: Fix SLES 11 client builds.
407 Details : Resolved uninitalized pointer gcc warnings.
411 Description: Allow non-root access for "lfs check".
412 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
416 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
418 Severity : enhancement
420 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
421 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
425 Description: Sanity test_27p fails.
426 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
427 Sanity tests 27 were cleaned up and improved.
431 Description: Unify req format on client/servers
432 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
433 buffers and swabbers approach..
438 Description: drop join file stuff.
440 Severity : enhancement
442 Description: Add 2.6.27 and SLES11 (patchless client) support
446 Description: Fix in ptlrpc_expire_one_request() to print the signed time
449 Severity : enhancement
451 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
454 Severity : enhancement
456 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
458 Severity : enhancement
460 Description: correctly handle big reply message.
461 Details : send LNet event if reply is bigger then buffer and adjust this buffer
466 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
468 Severity : enhancement
470 Description: Add support for OEL5.
474 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
476 Severity : enhancement
478 Description: "weak-modules" support
479 Details : Implement "weak-modules" support which enables kernel modules
480 to be used with any kernel that implements the same kABI. In
481 order to achieve this modules are now installed in
482 /lib/modules/$(uname -r)/updates/kernel on all distributions.
486 Description: client could not reconnect to OST because of an active request.
487 Details : The client now retries to reconnect to the same server, if a
488 connect request failed with EBUSY or -EAGAIN. Test case 35b
489 conf-sanity.sh has been created.
493 Description: Don't allow make backward step on assignin osc next id.
494 Details : race between allocation next id and ll_sync thread can be cause
495 of set wrong osc next id and can be kill valid ost objects.
499 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
503 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
507 Description: Error handling in osc_statfs_interpret() has been improved.
508 Details : Check in osc_statfs_interpret() for EBADR.
512 Description: Do not send statfs() requests to OSTs disabled by administrator.
513 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
515 Severity : enhancement
517 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
519 Severity : enhancement
521 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
523 Severity : enhancement
525 Description: Enhancement for lfs(1) command to use numeric uid/gid.
529 Description: Improved cleanup procedure for conf-sanity test_46a.
533 Description: Do not start bulk transfer, if client tried to reconnect.
534 Details : Do not start bulk transfer, if client tried to reconnect,
535 but failed due to active rpcs. Check for exp_abort_active_req
536 has been implemented.
540 Description: Prevent inconsistences between linux and lustre mount structures.
541 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
542 Make the sleep interruptible.
544 Severity : enhancement
546 Description: provide server to client comms path
547 Details : server to client communications path via new LDLM_SET_INFO rpc
551 Description: truncate starts GFP_FS allocation under transaction causing deadlock
552 Details : ldiskfs_truncate calls grab_cache_page which may start page
553 allocation under an open transaction. This may lead to
554 calling prune_icache with consequent lustre reentrance.
558 Description: send correctly lsm on open replay
559 Details : MDS is trust to LSM size on replay open, but client can set wrong size
564 Description: enable client interface failover
565 Details : When a child reconnects from another NID, properly update export
566 nid hash position and ldlm reverse import.
568 Severity : enhancement
570 Description: Communicate OST degraded/readonly state via statfs to MDS
571 Details : Flags in the statfs returned from OSTs indicate whether the
572 OST is in a degraded RAID state, or if the filesystem has
573 turned read-only after a filesystem error is detected.
578 Description: don't panic if EPROTO was hit when reading symlink
579 Details : correctly handling request reference in error cases.
581 Severity : enhancement
583 Description: Add LustreNetLink, a kernel-userspace communcation path.
585 Severity : enhancement
587 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
592 Description: access to llog context before init.
593 Details : move handling CATALOGS file at osc layer and forbid access to llog
597 Frequency : rare, only when client is writing to OST on local node
599 Description: Avoid deadlock for local client writes
600 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
601 memory freeing context. This allows OST threads to set the
602 PF_MEMALLOC flag on task structures in order to allocate memory
603 from reserved pools and complete IO.
604 Use GFP_HIGHUSER for OST allocations for non-local client writes,
605 so that the OST threads generate memory pressure and allow
606 inactive pages to be reclaimed.
608 Severity : enhancement
610 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
615 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
616 Details : move ldlm namespace creation in setup phase to avoid grab
617 _lprocfs_lock with cli_sem held.
621 Description: Temporarily disable grant shrink.
622 Details : Disable the feature for debugging.
625 Frequency : only when formatting test filesystems
627 Description: Unable to run several mkfs.lustre on loop devices at the same time
628 Details : mkfs.lustre returns error 256 on the concurrent loop devices
629 formatting. The solution is to proper handle the error.
631 Severity : enhancement
633 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
635 Severity : enhancement
637 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
642 Description: resolve race between obd_disconnect and class_disconnect_exports
643 Details : if obd_disconnect will be called to already disconnected export he
644 forget release one reference and osc module can't unloaded.
646 Severity : enhancement
648 Description: Allow tuning service thread via /proc
649 Details : For each service a new
650 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
651 created that can be used to set min/max thread counts, and get the
652 current number of running threads.
657 Descriptoin: don't return error if have partially created objects for file.
658 Details : lov_update_create_set uses set->set_success as index for created
659 objects, so if some requests will be failed, they will have hole
660 at end of array and qos_shrink_lsm for allocate correct lsm.
662 Severity : enhancement
664 Description: Update OFED support to 1.4.1
669 Description: Reduce small size read RPC
670 Details : Set read-ahead limit for every file and only do read-ahead when
671 available read-ahead pages are bigger than 1M to avoid small size
674 Severity : enhancement
676 Description: move AT tunable parameters for more consistent usage
677 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
679 Severity : enhancement
681 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
682 Details : allow skip disconnected ost for send statfs request and hide error
688 Description: fix racy locking of mballoc block bitmaps causing BUG
689 Details : The locking of the mballoc buddy bitmap and the in-memory
690 block bitmap was using two different spin locks in some
691 cases. This made it possible to incorrectly access the
692 mballoc bitmap while another process was modifying it,
693 causing a sanity assertion to fail. While no on-disk corruption
694 was reported, there was some risk of this happening.
697 Frequency : rare, on llog test 6
699 Descriptoin: don't allow connect to already connected import
700 Details : allowing connect to already connected import is hide connecting problem.
705 Descriptoin: allow kill process which wait statahead result
706 Details : for some reasons 'ls' can stick in waiting result from statahead,
707 in this case need way for kill this process.
709 Severity : enhancement
711 Description: Add state history info file, enhance import info file
712 Details : Track import connection state changes in a new osc/mdc proc file;
713 add overview-type data to the osc/mdc import proc file.
715 Severity : enhancement
717 Description: MDS create should not wait for statfs RPC while holding DLM lock.
719 Severity : enhancement
721 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
726 Descriptoin: OST grant too much space to client even there are not enough space.
727 Details : Client will shrink its grant cache to OST if there are no write
728 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
729 this grant cache if there are already not enough avaible space
730 (left_space < total_clients * 32M).
733 Frequency : start MDS on uncleanly shutdowned MDS device
735 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
736 Details : stay in waiting mds<>ost recovery finished produce random bugs
737 due race between two ll_sync thread for one lov target. send
738 ACTIVATE event only if connect realy finished and import have
742 Frequency : rare, connect and disconnect target at same time
744 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
745 Details : don't call obd_disconnect under lov_lock. this long time
746 operation and can block ptlrpcd which answer to connect request.
751 Descriptoin: don't lose wakeup for imp_recovery_waitq
752 Details : recover_import_no_retry or invalidate_import and import_close can
753 both sleep on imp_recovery_waitq, but we was send only one wakeup
757 Frequency : always with long access acl
759 Descriptoin: mds can't pack reply with long acl.
760 Details : mds don't control size of acl but they limited by reint/getattr
763 Severity : enhancement
765 Description: Update to SLES10 kernel-2.6.16.60-0.33.
767 Severity : enhancement
769 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
772 Frequency : start MDS on uncleanly shutdowned MDS device
774 Descriptoin: aborting recovery hang on MDS
775 Details : don't throttle destroy RPCs for the MDT.
778 Frequency : on remount
780 Description: external journal device not working after the remount
781 Details : clear dev_rdonly flag for external journal devices in
787 Description: shutdown vs evict race
788 Details : client_disconnect_export vs connect request race.
789 if client will evicted at this time - we start invalidate
790 thread without referece to import and import can be freed
794 Frequency : rare, need acl's on inode.
796 Description: client can't handle ost additional correctly
797 Details : if ost was added after client connected to mds client can have
798 hit lnet_try_match_md ... to big messages to wide striped files.
799 in this case need teach client to handle config events about add
800 lov target and update client max ea size at that event.
802 Severity : enhancement
804 Description: Changelogs
805 Details : Changelogs are a lightweight mechanism to track filesystem
806 metadata and namespace changes. The changelog is recorded
807 permanently on the MDTs, and is periodically "consumed" / purged
808 when records are no longer needed.
810 Severity : enhancement
812 Description: compact fld format with extents
813 Details : Store range of seq rather than every seq in FLD. Seq
814 controller update FLD rather than clients. In Case of CMD, mdt0
815 has FLD, all other metadata server act as non persistent proxy
816 for FLD queries and cache fld entries in fld cache.
821 Description: don't skip ost target if they assigned to file
822 Details : Drop slow OSCs if we can, but not for requested start idx.
823 This means "if OSC is slow and it is not the requested
824 start OST, then it can be skipped, otherwise skip it only
825 if it is inactive/recovering/out-of-space.
829 Description: more cleanup in mds_lov
830 Details : not send LOV EA under replay, we can't know about they size at this
831 time. Don't allow client connect to mds before any ost connected,
832 for avoid problems with LOV EA size and returning EIO to client.
834 Severity : enhancement
836 Description: Interoperability at server side (Disk interoperability)
838 Severity : enhancement
840 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
842 Severity : enhancement
844 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
846 Severity : enhancement
848 Description: New client IO stack (CLIO).
850 Severity : enhancement
852 Description: Commit on sharing. Eliminate inter-client dependencies between
853 uncommitted transactions by doing transaction commits.
854 Thereby clients may recovery independently.
857 Frequency : Create a symlink file with a very long name
859 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
860 Details : If there is no extra space in the request for early cancels,
861 ldlm_req_handles_avail() returns 0 instead of a negative value.
863 Severity : enhancement
865 Description: Add /proc entry for import status
866 Details : The mdc, osc, and mgc import directories now have
867 an import directory that contains useful import data for debugging
870 Severity : enhancement
872 Description: Re-disable certain /proc logging
873 Details : Enable and disable client's offset_stats, extents_stats and
874 extents_stats_per_process stats logging on the fly.
877 Frequency : Only on FC kernels 2.6.22+
879 Description: oops in statahead
880 Details : Do not drop reference count for the dentry from VFS when lookup,
881 VFS will do that by itself.
883 Severity : enhancement
885 Description: Generic /proc file permissions
886 Details : Set /Proc file permissions in a more generic way to enable non-
887 root users operate on some /proc files.
891 Description: Hitting mdc_commit_close() ASSERTION
892 Details : Properly handle request reference release in
893 ll_release_openhandle().
897 Description: quota recovery deadlock during mds failover
898 Details : This patch includes att18982, att18236, att18237 in bz14840.
900 1. fix osts hang when mds does failover with quotaon
901 2. prevent watchdog storm when osts threads wait for the
906 Frequency : only patchless client
907 Description: add workaround for race between add/remove dentry from hash
909 Severity : enhancement
911 Description: Allow OST glimpses to return PW locks
915 Description: LBUG when llog conf file is full
916 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
920 Description: Prevent import from entering FULL state when server in recovery
924 Description: service mount cannot take device name with ":"
925 Details : Only when device name contains ":/" will mount treat it as
931 Description: replace ptlrpcd with the statahead thread to interpret the async
932 statahead RPC callback
936 Frequency : on recovery
937 Description: I/O failures after umount during fail back
938 Details : if client reconnected to restarted server we need join to recovery
939 instead of find server handler is changed and process self
940 eviction with cancel all locks.
942 Severity : enhancement
944 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
946 Severity : enhancement
948 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
950 Severity : enhancement
952 Description: Upcall on Lustre log has been dumped
953 Details : Allow for a user mode script to be called once a Lustre log has
954 been dumped. It passes the filename of the dumped log to the
955 script, the location of the script can be specified via
956 /proc/sys/lnet/debug_log_upcall.
961 Description: avoid idr_remove called for id which is not allocated.
962 Details : Move assigment s_dev for clustered nfs to end of initialization,
963 to avoid problem with error handling.
968 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
969 Details : When connection is reused this not moved from CONN_UNUSED_HASH
970 into CONN_USED_HASH and this prodice warning when put connection
971 again in unused hash.
974 Severity : enhancement
976 Description: File striping can now be set to use an arbitrary pool of OSTs.
978 Severity : enhancement
980 Description: Export bytes_read/bytes_write count on OSC/OST.
984 Description: Early reply size mismatch, MGC loses connection
985 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
986 the connect flags are properly negotiated.
991 Description: quotas are not honored with O_DIRECT
992 Details : all writes with the flag O_DIRECT will use grants which leads to
993 this problem. Now using OBD_BRW_SYNC to guard this.
997 Description: add quota statistics
998 Details : 1. sort out quota proc entries and proc code.
999 2. add quota statistics
1001 Severity : enhancement
1003 Description: enable quota support for HEAD.
1007 Description: Properly propagate oinfo flags from lov to osc for statfs
1008 Details : restore missing copy oi_flags to lov requests.
1010 Severity : enhancement
1012 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1013 : l_getgroups(8), lst(8), routerstat(8)
1014 Details : included man pages for llobdstat(8), llstat(8),
1015 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1017 Severity : enhancement
1019 Description: configure's --enable-quota should check the
1020 : kernel .config for CONFIG_QUOTA
1021 Details : configure is terminated if --enable-quota is passed but
1022 : no quota support is in kernel
1026 Description: Remove portals compatibility
1027 Details : Remove portals compatibility, not interoperable with releases
1032 Description: Resolve device initialization race
1033 Details : Prevent proc handler from accessing devices added to the
1034 obd_devs array but yet be intialized.
1036 Severity : enhancement
1038 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1040 Severity : enhancement
1042 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1047 Description: Using wrong pointer in osc_brw_prep_request
1048 Details : Access to array[-1] can produce panic if kernel compiled with
1049 CONFIG_PAGE_ALLOC enabled
1053 Description: Client runs out of low memory
1054 Details : Consider only lowmem when counting initial number of llap pages
1058 Description: *optional* service tags registration
1059 Details : if the "service tags" package is installed on a Lustre node
1060 When the filesystem is mounted, a local-node service tag will
1061 be created. See http://inventory.sun.com/ for more information
1062 about the Service Tags asset management system.
1066 Description: Kernel BUG tries to release flock
1067 Details : Lustre does not destroy flock lock before last reference goes
1068 away. So always drop flock locks when client is evicted and
1069 perform unlock regardless of successfulness of speaking to MDS.
1073 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1077 Description: sanity test 65a fails if stripecount of -1 is set
1078 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1083 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1084 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1087 Severity : enhancement
1089 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1093 Description: do not process already freed flock
1094 Details : flock can possibly be freed by another thread before it reaches
1095 to ldlm_flock_completion_ast.
1099 Description: LBUG during stress test
1100 Details : Need properly lock accesses the flock deadlock detection list.
1104 Description: oops in page fault handler
1105 Details : kernel page fault handler can return two special 'pages' in error
1106 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1110 Description: timeout with invalidate import.
1111 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1112 should be handled by ptlrpcd. This produce long age waiting and
1113 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1115 Severity : enhancement
1117 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1121 Description: enable MGS and MDT services start separately
1122 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1123 a co-located MGS without starting the MGS, which is a complement
1124 to 'nosvc' mount option.
1128 Frequency : after recovery
1129 Description: precreate to many object's after del orphan.
1130 Details : del orphan st in oscc last_id == next_id and this triger growing
1131 count of precreated objects. Set flag LOW to skip increase count
1132 of precreated objects.
1136 Frequency : rare, on clear nid stats
1137 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1138 Details : when clean nid stats sometimes try destroy live entry,
1139 and this produce panic in free.
1143 Description: Stack overflow during MDS log replay
1144 ease stack pressure by using a thread dealing llog_process.
1148 Description: wait until IO finished before start new when do lock cancel.
1149 Details : VM protocol want old IO finished before start new, in this case
1150 need wait until PG_writeback is cleared until check dirty flag and
1151 call writepages in lock cancel callback.
1153 Severity : enhancement
1155 Description: using special macro for print time and cleanup in includes.
1159 Description: mds_mfd_close() ASSERTION(rc == 0)
1160 Details : In mds_mfd_close(), we need protect inode's writecount change
1161 within its orphan write semaphore to prevent possible races.
1165 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1169 Frequency : rare, on shutdown ost
1170 Description: don't hit live lock with umount ost.
1171 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1172 use shrink_dcache_sb instead.
1176 Description: don't panic with use echo client
1177 Details : echo client pass NULL as client nid pointer and this produce null
1178 pointer dereference.
1182 Description: fix build on ppc32
1183 Details : compile code with -m64 flag produce wrong object file for ppc32.
1187 Description: add message levels for liblustreapi
1191 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1192 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1193 will result in an -EAGAIN and in the exit call path we call
1194 alloc_rr() which will with increasing aggressiveness attempt to
1195 aquire precreated objects on the minimum number of required OSCs.
1199 Description: Use old size assignment to avoid deadlock
1200 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1201 the scheduling while holding a spinlock. We do not need locking
1202 for size in ll_update_inode() because size is only updated from
1203 the MDS for directories or files without objects, so there is no
1204 other place to do the update, and concurrent access to such inodes
1205 are protected by the inode lock.
1209 Description: resolve "_IOWR redefined" build error on SLES10
1213 Description: dump the memory debugging after all modules are unloaded to
1214 suppress false negative in conf_sanity test 39
1216 Severity : enhancement
1218 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1223 Description: fix SLES kernel versioning
1224 Details : the kernel version for our SLES 10 kernel did not include a "-"
1225 before the "smp" at the end. while this was not a problem in
1226 general, it did mean that software trying to use the kernel
1227 version to try to detect a vendor specific kernel would fail.
1228 this was most evident by the OFED build scripts.
1232 Description: Don't update lov_desc members until making sure they are valid
1233 Details : When updating lov_desc members via proc fs, need fix their
1234 validities before doing the real update.
1238 Description: don't put request into delay list while invalidate in flight.
1239 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1240 import in flight. this produce timeout for invalidate and sometimes
1241 can cause stale data.
1245 Frequency : on ppc only
1246 Description: not convert ost objects for directory because it's not exist.
1247 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1251 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1253 Severity : enhancement
1255 Description: Update to support 2.6.22.14 vanilla kernel.
1259 Frequency : rare, on recovery
1260 Description: read procfs can produce deadlock in some situation
1261 Details : Holding lprocfs lock which send rpc can produce block for destroy
1262 obd objects and this also block reconnect with -EALREADY. This
1263 isn't fix all lprocfs bugs - but make it rare.
1265 Severity : enhancement
1267 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1270 Frequency : frequent on X2 node
1272 Description: mdc_set_open_replay_data LBUG
1273 Details : Set replay data for requests that are eligible for replay.
1277 Description: lustre_mgs: operation 101 on unconnected MGS
1278 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1279 MGC, and late on MGC cannot successfully connect to MGS and a lot
1280 of the error messages complaining that MGS is not connected.
1284 Frequency : on network error
1285 Description: panic with double free request if network error
1286 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1287 but it's true only for synchronus enqueue, for async enqueue
1288 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1291 Severity : enhancement
1293 Description: client-side metadata stat-ahead during readdir(directory readahead)
1294 Details : perform client-side metadata stat-ahead when the client detects
1295 readdir and sequential stat of dir entries therein
1298 Frequency : on start mds
1300 Description: Implement get_info(last_id) in obdfilter.
1303 Frequency : occasional
1305 Description: Correctly check stale fid, not start epoch if ost not support SOM
1306 Details : open with flag O_CREATE need set old fid in op_fid3 because
1307 op_fid2 was overwritten with new generated fid, but mds can answer
1308 with one of these two fids and both is not stale. Setattr
1309 incorrectly started an epoch and assume will be called
1310 done_writing, but without SOM done_writing ever being called.
1313 Frequency : rare, depends on device drivers and load
1315 Description: MDS or OSS nodes crash due to stack overflow
1316 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1317 In some cases, in conjunction with device drivers that use a lot
1318 of stack the MDS (or possibly OSS) service threads could overflow
1319 the stack. One change which was identified to consume additional
1320 stack has been reworked to avoid the extra stack usage.
1323 Frequency : occasional
1325 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1326 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1327 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1328 oscc_has_objects and see OSC has no objests and call
1329 oscc_internal_create to resend create request.
1331 Severity : enhancement
1333 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1335 Severity : enhancement
1337 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1340 Frequency : very rare
1342 Description: Fix replay if there is an un-replied request and open
1343 Details : In some cases, older replay request will revert the
1344 mcd->mcd_last_xid on MDS which is used to record the client's
1345 latest sent request.
1347 Severity : enhancement
1349 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1351 Severity : enhancement
1353 Description: Add rhel5 support to HEAD.
1355 Serverity : enhancement
1357 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1362 Description: Don't allow skipping OSTs if index has been specified.
1363 Details : Don't allow skipping OSTs if index has been specified, make
1364 locking in internal create lots better.
1368 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1369 Details : access to bitfield in structure is always rounded to long
1370 and this produce problem with not atomic change any bit.
1374 Description: Lustre make rpms failed.
1375 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1376 builds Lustre rpms from tarball.
1379 Frequency : rare on shutdown ost
1381 Description: If llog cancel was not send before clean_exports phase, this can
1382 produce deadlock in llog code.
1383 Details : If llog thread has last reference to obd and call class_import_put
1384 this produce deadlock because llog_cleanup_commit_master wait when
1385 last llog_commit_thread exited, but this never success because was
1386 called from llog_commit_thread.
1390 Description: allow userland application know is lost one of stripes.
1391 Details : fill lvb_blocks with error code on ost and return it to
1392 application if error flag found.
1396 Description: NULL lov_tgts causing MDS oops
1397 Details : more safe checks for NULL lov_tgts for avoid oops.
1399 Severity : enhancement
1401 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1405 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1406 Details : If an OST is down the MDS will hang indefinitely in
1407 obd_statfs_rqset() waiting for the statfs data. While for
1408 MDS QOS usage of statfs, it should not stuck in waiting.
1410 Severity : enhancement
1412 Description: remote_acl support
1413 Details : Support ACL-based permission check for remote user.
1414 Support setfacl/getfacl for remote user with the utils
1415 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1416 the system "{s,g}etfacl" utils.
1418 Severity : enhancement
1420 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1422 Severity : enhancement
1424 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1428 Description: make mgs_setparam() handle fsname containing dash
1429 Details : fsname containing a dash does not work with lctl conf_param
1431 Severity : enhancement
1433 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1435 Severity : enhancement
1437 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1441 Frequency : rare, at shutdown
1442 Description: access already free / zero obd_namespace.
1443 Details : if client_disconnect_export was called without force flag set,
1444 and exist connect request in flight, this can produce access to
1445 NULL pointer (or already free pointer) when connect_interpret
1446 store ocd flags in obd_namespace.
1450 Frequency : only at startup
1451 Description: not alloc memory with spinlock held.
1452 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1453 if any spinlock held.
1455 Severity : enhancement
1457 Description: make lustre randomly fail allocating memory
1458 Details : Make lustre randomly failed allocating memory for testing purpose.
1460 Severity : enhancement
1462 Description: lost problems with lov objid file
1463 Details : Fixes some scability and access to not inited memory problems
1464 in work with lov objdid file.
1469 Description: lfs find does not continue on file error
1470 Details : Continue other files processing when a file/dir is absent.
1474 Description: Inconsistent usage of lustre_pack_reply()
1475 Details : Standardize the usage of lustre_pack_reply() such that it
1476 always generate a CERROR on failure.
1479 Frequency : occasional
1481 Description: MDS hang or stay in waiting lock
1482 Details : If client receive lock with CBPENDING flag ldlm need send lock
1483 cancel as separate rpc, to avoid situation when cancel request
1484 can't processed due all i/o threads stay in wait lock.
1488 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1492 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1496 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1500 Description: Update to RHEL4 latest kernel.
1502 Severity : enhancement
1504 Description: Build SLES10 patchless client fails
1505 Details : The configure was broken by run ./configure with
1506 --with-linux-obj=.... argument for patchless client. When the
1507 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1508 can't search header adequately. Use absolute path such as
1509 -I($LINUX)/include instead.
1513 Description: interrupt oig_wait produce painc on resend.
1514 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1515 private set, and this produce situation when rq_ptlrpcd_data not
1516 copyed to new allocated request and triggered LBUG on assert
1517 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1518 ptlrpcd set if request is changed and can be safety changed to use
1523 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1524 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1525 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1526 these macros and replace them with their difinitions instead.
1530 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1531 Details : Update which_patch & target file for SLES10 latest kernel.
1534 Description: Remove client patches from SLES 10 kernel.
1535 Details : This causes SLES 10 clients to behave as patchless clients
1536 even on a Lustre-patched (server) kernel.
1538 Severity : enhancement
1540 Description: self-adjustable client's lru lists
1541 Details : use adaptive algorithm for managing client cached locks lru
1542 lists according to current server load, other client's work
1543 pattern, memory activities, etc. Both, server and client
1544 side namespaces provide number of proc tunables for controlling
1547 Severity : enhancement
1549 Description: light-weight GSS support
1550 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1551 and reduce performance overhead.
1553 Severity : enhancement
1555 Description: Linux keyring support
1556 Details : Support using service of Linux keyring for Lustre GSS internal
1557 context refresh/cache mechanism.
1561 Description: Fix errors in lfs documentation
1562 Details : Fixes man pages
1566 Description: don't use GFP_* in generic Lustre code.
1567 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1571 Description: obdclass is limited by single OBD_ALLOC(idarray)
1572 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1576 Description: warnings with build patchless client with vanila 2.6.19 and up
1577 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1578 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1582 Frequency : Only for SLES
1583 Description: sanity_quota fail test_1
1584 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1585 file, which makes TSTID[2] inunique.
1589 Frequency : Only for SLES9
1590 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1591 Details : Update target/ChangeLog/which_patch .
1595 Frequency : Only for test_10 in sanity-quota.sh
1596 Description: a bug in quota test code
1597 Details : checking a test flag in a wrong place
1601 Frequency : Only for test_13 in sanity-quota.sh
1602 Description: a bug in quota test script
1603 Details : shouldn't check overfull allocated quota
1607 Description: jbd statistics
1608 Details : Port older jbd statistics patch for sles10
1610 Severity : enhancement
1612 Description: Add printing inode info into message about error in writepage.
1617 Description: reply_lock_interpret crash due to race with it and lock cancel.
1618 Details : Do not replay locks that are being cancelled. Do not reference
1619 locks by their address during replay, just by their handle.
1623 Frequency : When flocks are used.
1624 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1625 Details : Flock locks might destroy just granted lock if it could be merged
1626 with another existing flock, this is done in completion handler,
1627 so teach ldlm_cli_enquque_fini that this is a valid case for
1633 Description: Oops in read and write path when failing to allocate lock.
1634 Details : Check if lock allocation failed and return error back.
1638 Description: lstripe command fails for valid OST index
1639 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1640 instead of lov->desc.ld_active_tgt_count.
1644 Description: sanity.sh failed test 103
1645 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1650 Description: df doesn't work properly if diskfs blocksize != 4K
1651 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1655 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1656 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1658 Severity : enhancement
1660 Description: Adaptive timeouts
1661 Details : RPC timeouts adapt to changing server load and network
1662 conditions to reduce resend attempts and improve recovery time.
1666 Description: llapi_file_create() does not allow some changes
1667 Details : add llapi_file_open() that allows specifying the mode and
1668 open flags, and also returns an open file handle.
1672 Description: merge and cleanup kernel patches.
1673 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1677 Description: Add journal checksum support.(Kernel part)
1678 Details : The journal checksum feature adds two new flags i.e
1679 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1680 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1681 indicates that the commit block contains the checksum for
1682 the blocks described by the descriptor blocks. Now commit
1683 record can be sent to disk without waiting for descriptor
1684 blocks to be written to disk. This behavior is controlled
1685 using JBD2_FEATURE_ASYNC_COMMIT flag.
1689 Description: OSS needs mutliple precreate threads
1690 Details : Add ability to start more than one create thread per OSS.
1694 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1695 Details : execute lfs setstripe on client
1699 Description: mds_obd_create error creating tmp object
1700 Details : When the user sets quota on root, llog will be affected and can't
1701 create files and write files.
1704 Frequency : Always on ia64 patchless client, and possibly others.
1706 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1707 Details : This allows the patchless client to be loaded on architectures
1708 without this export.
1712 Description: RedHat Update kernel for RHEL5
1713 Details : Add the kernel config file for RHEL5.
1717 Description: RedHat Update kernel for RHEL5
1718 Details : Modify the kernel config file more closer RHEL5.
1722 Description: Build failure against Centos5 (RHEL5)
1723 Details : Define PAGE_SIZE when it isn't present.
1727 Description: test_7 on sanity-quota.sh can't be used on separate machines
1728 Details : add facet to handle it
1732 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1733 Details : with NFS, the anon dentry's parent was set to itself in
1734 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1735 corresponding dentry other than use rec->ur_name.
1737 Severity : enhancement
1739 Description: lfs setstripe enhancement
1740 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1744 Description: enable data checksumming by default
1745 Details : enable checksum by default, allow --disable-checksum
1746 configure option and "-o nochecksum" mount option. Checksums
1747 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1751 Description: lustre support for RHEL5
1752 Details : Add support for RHEL5.
1756 Description: Client eviction due to failover config
1757 Details : after a connection loss, the lustre client should attempt to
1758 reconnect to the last active server first before trying the
1759 other potential connections.
1763 Description: when mds and osts use different quota unit(32bit and 64bit),
1764 quota will be released repeatly.
1765 Details : void sending multiple quota reqs to mds, which will keep the
1766 status between the reqs.
1770 Description: osts not allocated evenly to files
1771 Details : change the condition to increase offset_idx
1774 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1776 Description: Data corruption for OSTs that are formatted larger than 2TB
1778 Details : When generating the bio request for lustre file writes the
1779 sector number would overflow a temporary variable before being
1780 used for the IO. The data reads correctly from Lustre (which
1781 will overflow in a similar manner) but other file data or
1782 filesystem metadata may be corrupted in some cases.
1786 Description: Tune the kernel for good SCSI performance.
1787 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1788 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1793 Description: rewrite ext2-derived code in obdclass/uuid.c
1794 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1795 from scratch preserving functionality.
1799 Description: "lfs find -obd UUID" prints directories
1800 Details : "lfs find -obd UUID" will return all directory names instead
1801 of just file names. It is incorrect because the directories
1802 do not reside on the OSTs.
1806 Description: use i_size_read and i_size_write in 2.6 port
1807 Details : replace inode->i_size access with i_size_read/write()
1810 Frequency : when removing large files
1812 Description: scheduling issue during removal of large Lustre files
1813 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1814 It causes scheduling issues when removing large files (17TB in the
1818 Frequency : only with liblustre clients on XT3
1820 Description: evictions taking too long
1821 Details : allow llrd to evict clients directly on OSTs
1824 Frequency : only on ppc
1826 Description: /proc/fs/lustre/devices broken on ppc
1827 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1828 We should make sure the type of 'index' is loff_t and then cast
1829 explicitly as needed below. Do not assign an explicitly cast
1833 Frequency : only for rhel5
1835 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1836 Details : Modify the target file & which_kernel.
1838 Severity : enhancement
1840 Description: omit set fsid for export NFS
1841 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1842 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1846 Frequency : Always, for kernels after 2.6.16
1847 Description: Fix warning idr_remove called for id=.. which is not allocated.
1848 Details : Last kernels save old s_dev before kill super and not allow
1849 to restore from callback - restore it before call kill_anon_super.
1853 Description: buffer overruns could theoretically occur
1854 Details : llapi_semantic_traverse() modifies the "path" argument by
1855 appending values to the end of the origin string, and a
1856 overrun may occur. Adding buffer overrun check in liblustreapi.
1860 Description: Fix error on 'ls .' at the top of the Lustre mount.
1861 Details : Don't revalidate dentry if it is a root dentry.
1865 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1866 Details : Modify vm-tunables-rhel4.patch.
1870 Description: Kernel config for 2.6.18-vanilla.
1871 Details : Modify targets/2.6-vanilla.target.in.
1872 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1873 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1874 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1875 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1879 Description: improve handling recoverable errors
1880 Details : if request processig with error which can be recoverable on server
1881 request should be resend, otherwise page released from cache and
1886 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1887 Details : Modify target file & which_patch.
1888 A flaw was found in the IA32 system call emulation provided
1889 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1890 value could be stored in the %RAX register, which could trigger an
1891 out-of-bounds system call table access. An untrusted local user
1892 could exploit this flaw to run code in the kernel
1893 (ie a root privilege escalation). (CVE-2007-4573).
1897 Description: O_DIRECT bypasses client statistics.
1898 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1899 read_ahead_stats were updated. copied the stats section from
1900 osc_send_oap_rpc() into async_internal().
1904 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1908 Description: change order of libsysio includes
1909 Details : '#include sysio.h' should always come before '#include xtio.h'
1913 Description: handle "serious error: objid * already exists" more gracefully
1914 Details : If LAST_ID value on disk is smaller than the objects existing in
1915 the O/0/d* directories, it indicates disk corruption and causes an
1916 LBUG(). If the object is 0-length, then we should use the existing
1917 object. This will help to avoid a full fsck in most cases.
1919 Severity : enhancement
1921 Description: adapt the lustre_config script to support the upgrade case
1922 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1923 server targets to 1.6 in parallel.
1927 Description: To avoid grant space > avaible space when the disk is almost
1928 full. Without this patch you might see the error "grant XXXX >
1929 available" or some LBUG about grant, when the disk is almost
1931 Details : In filter_check_grant, for non_grant cache write, we should
1932 check the left space by if (*left > ungranted + bytes), instead
1933 of (*left > ungranted), because only we are sure the left space
1934 is enough for another "bytes", then the ungrant space should be
1935 increase. In client, we should update cl_avail_grant only there
1936 is OBD_MD_FLGRANT in the reply.
1940 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1941 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1942 described in CVE-2007-4573. This problem could allow untrusted
1943 local users to gain root access.
1946 Frequency : when using O_DIRECT and quotas
1948 Description: Incorrect file ownership on O_DIRECT output files
1949 Details : block usage reported by 'lfs quota' does not take into account
1950 files that have been written with O_DIRECT.
1955 Description: touch file failed when fs is not full
1956 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1957 otherwise we can get ENOSP while fs is not full.
1961 Description: parallel lock callbacks
1962 Details : Instead of sending blocking and completion callbacks as separated
1963 requests, adding them to a set and sending in parallel.
1966 Frequency : only for Cray XT3
1967 Bugzilla : 12829/13455
1968 Description: Changing primary group doesn't change the group lustre assigns to
1970 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1971 overridden with the primary group provided by the group upcall,
1972 whereas the supplied fsgid can be trusted if it is in the list of
1973 supplementary groups returned by the group upcall.
1975 Severity : enhancement
1977 Description: Allow masking D_WARNING, D_ERROR messages from console
1978 Details : Console messages can now be disabled via lnet.printk.
1982 Description: User code with malformed file open parameter crashes client node
1983 Details : Before packing join_file req, all the related reference should be
1984 checked carefully in case some malformed flags cause fake
1985 join_file req on client.
1989 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1990 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1991 mode is equal to the granted mode. However, it should also check
1992 that LDLM_FL_CP_REQD is not set.
1996 Description: Heavy nfs access might result in deadlocks
1997 Details : After ELC code landed, it is now improper to enqueue any mds
1998 locks under och_sem, because enqueue might want to decide to
1999 cancel open locks for same inode we are holding och_sem for.
2003 Description: Client eviction while running blogbench
2004 Details : A lot of unlink operations with concurrent I/O can lead to a
2005 deadlock causing evictions. To address the problem, the number of
2006 oustanding OST_DESTROY requests is now throttled to
2007 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2008 callbacks are processed in priority.
2012 Description: enable ACLs on MDS by default
2013 Details : ACLs must be enabled on MDS by default.
2016 Frequency : PPC/PPC64 only
2018 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2019 Details : fix duplicated definitions between asm-ppc64/types.h and
2020 lustre_types.h on PPC.
2023 Frequency : PPC/PPC64 only
2025 Description: asm-ppc/segment.h does not exist
2026 Details : fix compile issue on PPC.
2030 Description: better handle error messages in extents code
2033 Frequency : RHEL4 only
2035 Description: mkfs is very slow on IA64/RHEL4
2036 Details : A performance regression has been discovered in the MPT Fusion
2037 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2038 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2039 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2041 Severity : enhancement
2043 Description: SNMP support enhancement
2044 Details : Adding total number of sampled request for an MDS node in snmp
2047 Severity : enhancement
2049 Description: Optimize ldlm waiting list processing for PR extent locks
2050 Details : When processing waiting list for read extent lock and meeting read
2051 lock that is same or wider to it that is not contended, skip
2052 processing rest of the list and immediatelly return current
2053 status of conflictness, since we are guaranteed there are no
2054 conflicting locks in the rest of the list.
2058 Description: Time out and refuse to reconnect
2059 Details : When the failover node is the primary node, it is possible
2060 to have two identical connections in imp_conn_list. We must
2061 compare not conn's pointers but NIDs, otherwise we can defeat
2062 connection throttling.
2066 Description: port llog fixes from b1_6 into HEAD
2067 Details : Port llog reference couting and some llog cleanups from b1_6
2068 (bug 10800) into HEAD, for protect from panic and access to already
2069 free llog structures.
2073 Description: Detect stride IO mode in read-ahead
2074 Details : When a client does stride read, read-ahead should detect that and
2075 read-ahead pages according to the detected stride pattern.
2079 Description: data checksumming impacts single node performance
2080 Details : add support for several checksum algorithm. Currently, only CRC32
2081 and Adler-32 are supported. The checksum type can be changed on
2082 the fly via /proc/fs/lustre/osc/*/checksum_type.
2086 Description: use adler32 for page checksums
2087 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2092 Description: build for x2 fails
2093 Details : fix compile issue on Cray systems.
2097 Description: Properly match for duplicate locks
2098 Details : Due to different lock order from skiplists code, we need to
2099 traverse entire list for now
2102 Frequency : only on PPC/SLES10
2104 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2105 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2106 be defined. Add a hack in mkfs_lustre.c to work around this compile
2111 Description: LASSERT on MDS when client holding flock lock dies
2112 Details : ldlm pool logic depends on number of granted locks equal to
2113 number of released locks which is not true for flock locks, so
2114 just exclude such locks from consideration.
2118 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2119 Details : Use fsfilt_sync() for both the whole filesystem sync and
2120 individual file sync to eliminate dangerous inode locking
2121 with I_LOCK that can lead to a deadlock.
2125 Description: performance in 1.6.3
2126 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2127 sure that 1MB requests are not fragmented by the block layer.
2129 Severity : enhancement
2131 Description: organize the server-side client stats on per-nid basis
2132 Details : Change the structure of stats under obdfilter and mds to
2140 The "uuid"s file would list the uuids of _active_ exports.
2141 And the clear entry is to clear all stats and stale nids.
2143 Severity : enhancement
2145 Description: eliminate client locks in face of contention
2146 Details : file contention detection and lockless i/o implementation
2147 for contended files.
2151 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2154 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2156 Description: MDS may lose file striping (and hence file data) in some cases
2157 Details : If there are additional extended attributes stored on the MDS,
2158 in particular ACLs, SELinux, or user attributes (if user_xattr
2159 is specified for the client mount options) then there is a risk
2160 of attribute loss. Additionally, the Lustre file striping
2161 needs to be larger than default (e.g. striped over all OSTs),
2162 and an additional attribute must be stored initially in the
2163 inode and then increase in size enough to be moved to the
2164 external attribute block (e.g. ACL growing in size) for file
2169 Description: skiplist implementation simplification
2170 Details : skiplists are used to group compatible locks on granted list
2171 that was implemented as tracking first and last lock of each
2172 lock group the patch changes that to using doubly linked lists
2176 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2177 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2178 because LASSERT on both the data supplied by a client, and the
2179 data on disk is dangerous and incorrect.
2181 Severity : enhancement
2183 Description: Slow truncate/writes to huge files at high offsets.
2184 Details : Directly associate cached pages to lock that protect those pages,
2185 this allows us to quickly find what pages to write and remove
2186 once lock callback is received.
2190 Description: more ldlm soft lockups
2191 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2192 starve other threads from the resource lock for a long time in
2193 case of long waiting queue, so change the debug level from
2194 D_OTHER to the less frequently used D_INFO.
2196 Severity : enhancement
2198 Description: add -gid, -group, -uid, -user options to lfs find
2202 Description: Hung threads in invalidate_inode_pages2_range
2203 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2204 once one is completed. As a result, some RPCs are stuck in the
2205 queue and are never sent.
2209 Description: filter threads hungs on waiting journal commit
2210 Details : Cleanup filter group llog code, then only filter group llog will
2211 be only created in the MDS/OST syncing process.
2215 Description: Procfs and llog threads access destoryed import sometimes.
2216 Details : Sync the import destoryed process with procfs and llog threads by
2217 the import refcount and semaphore.
2219 Severity : enhancement
2221 Description: openlock cache of b1_6 port to HEAD
2226 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2227 Details : If insertion of an extent fails, then discard the inode
2228 preallocation and free data blocks else it can lead to duplicate
2233 Description: don't always update ctime in ext3_xattr_set_handle()
2234 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2235 In some cases the ctime should not be updated, for example for
2236 2.0->1.8 compatibility it is necessary to delete an xattr and it
2237 should not update the ctime.
2241 Bugzilla : 15713/16362
2242 Description: Assertion in iopen_connect_dentry in 1.6.3
2243 Details : looking up an inode via iopen with the wrong generation number can
2244 populate the dcache with a disconneced dentry while the inode
2245 number is in the process of being reallocated. This causes an
2246 assertion failure in iopen since the inode's dentry list contains
2247 both a connected and disconnected dentry.
2251 Description: assertion failure in ldlm_handle2lock()
2252 Details : fix a race between class_handle_unhash() and class_handle2object()
2253 introduced in lustre 1.6.5 by bug 13622.
2258 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2259 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2264 Description: oops in sd_iostats_seq_show()
2265 Details : unloading/reloading the scsi low level driver triggers a kernel
2266 bug when trying to access the sd iostat file.
2271 Description: Kernel panics during QLogic driver reload
2272 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2273 patch, causing memory corruption.
2278 Description: journal_dev option does not work in b1_6
2279 Details : pass mount option during pre-mount.
2281 Severity : enhancement
2283 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2284 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2285 extent information of a file. It can be used to map logical blocks
2286 in a file to physical blocks in the block device.
2290 Description: LDLM soft lockups - improvement
2291 Details : It is be possible to send the lock handle along with each read
2292 or write request because the client is already doing a lock match
2293 itself so there isn't any reason the OST should have to re-do that
2297 Frequency : only on Cray X2
2299 Description: X2 build failures
2300 Details : fix build failures on Cray X2.
2304 Description: xid & resent requests
2305 Details : Initialize RPC XID from clock at startup (randomly if clock is
2308 Severity : enhancement
2310 Description: Add lustre_start utility to start or stop multiple Lustre servers
2315 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2316 Details : In case of memory pressure, list_del() can be called twice on
2317 req->rq_history_list, causing a kernel oops.
2321 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2322 Details : fix stack overflow in the distributed lock manager by defering
2323 export eviction after a failed AST to the elt thread instead of
2324 handling it in the dlm interpret routine.
2328 Description: Convert some comments to new format.
2329 Details : Update documenting comments to match doxygen conventions.
2333 Description: Grammar fixes.
2334 Details : A couple of trivial spelling fixes.
2338 Description: OSD_COUNTERS-mandatory
2339 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2340 to simplify the code.
2344 Description: simplify cmm_device freeing logic.
2345 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2349 Description: Add lockdep support to dt_object_operations locking interface.
2350 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2351 parameter indicating lock ordering. Update mdd code to use new
2356 Description: Introduce failloc constants for lockless IO tests.
2357 Details : Add two new failloc constants to test lockless IO. Only one of
2358 them in implemented---another is checked in yet to be landed
2363 Description: Add lockdep support for inode mutex.
2364 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2365 in the situations where Lustre has to lock more than one inode
2370 Description: Add optional invariants checking support.
2371 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2372 configure switch. This macro is to be used for consistency and
2373 sanity checks that are too expensive to be left in `production'
2378 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2379 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2384 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2385 Details : Use spin_lock_nested() in (the only) situation where more than
2386 one ldlm_lock is locked simultaneously. Also, fix possible
2387 dead-lock in ldlm_lock_change_resource() by enforcing particular
2392 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2393 Details : Instead of specifying each ldlm_lock call-back through separate
2394 parameter, wrap them into struct ldlm_callback_suite.
2398 Description: Kill join_lru obd method and its callers.
2399 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2400 in memory---a requirement implemented through ->o_join_lru() obd
2401 method in HEAD. Remove this method and its users.
2405 Description: Add asynchronous ldlm ENQUEUE completion handler.
2406 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2407 case that stock ldlm_completion_ast() cannot handle as it waits
2408 until lock is granted. Introduce new ldlm_completion_ast_async()
2409 for this. Also comment ldlm_completion_ast().
2413 Description: ldlm_error <-> errno conversion.
2414 Details : Add functions to map (rather arbitrary) between LDLM error codes
2415 and standard errno values. CLIO needs this to prevent LDLM specific
2416 constants from escaping ldlm and osc.
2420 Description: Kill unused ldlm_handle2lock_ns() function.
2421 Details : Kill unused ldlm_handle2lock_ns() function.
2425 Description: Add lu_ref support to ldlm_lock
2426 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2427 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2428 and ldlm_resource. LDLM interface has to be changed, because code that
2429 releases a reference on a lock, has to "know" what reference this is.
2430 In the most frequent case
2432 lock = ldlm_handle2lock(handle);
2434 LDLM_LOCK_PUT(lock);
2436 no changes are required. When any other reference (received _not_ from
2437 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2438 instead of LDLM_LOCK_PUT().
2440 Arguably, changes are pervasive, and interface requires some discipline
2441 for proper use. On the other hand, it was very instrumental in finding
2442 a few leaked lock references.
2446 Description: Add ldlm_lock_addref_try().
2447 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2448 attempts to addref a lock that might be being canceled
2453 Description: Add ldlm_weigh_callback().
2454 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2455 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2456 hard-coded `number of pages' logic.
2460 Description: Add lockdep annotations to llog code.
2461 Details : Use appropriately tagged _nested() locking calls in the places
2462 where llog takes more than one ->lgh_lock lock.
2466 Description: Add loi_kms_set().
2467 Details : Wrap kms updates into a helper function.
2471 Description: Constify instances of struct lsm_operations.
2472 Details : Constify instances of struct lsm_operations.
2476 Description: lu_conf support.
2477 Details : On a server, a file system object is uniquely identified
2478 by a fid, which is sufficient to locate and load all object
2479 state (inode). On a client, on the other hand, more data are
2480 necessary instantiate an object. Change lu_object_find() and
2481 friends to take additional `lu_conf' argument describing object.
2482 Typically this includes layout information.
2486 Description: lu_context fixes.
2487 Details : Introduce new lu_context functions that are needed on the client
2488 side, where some system threads (ptlrpcd) are shared by multiple
2489 modules, and so cannot be stopped during module shutdown.
2493 Description: Add start and stop methods to lu_device_type_operations.
2494 Details : Introduce two new methods in lu_device_type_operations, that are
2495 invoked when first instance of a given type is created and last one
2496 is destroyed respectively. This is need by CLIO.
2500 Description: Add lu_ref support to struct lu_device.
2501 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2502 track leaked references.
2506 Description: Introduce lu_kmem_descr.
2507 Details : lu_kmem_descr and its companion interface allow to create
2508 and destroy a number of kmem caches at once.
2512 Description: Fix lu_object finalization race.
2513 Details : Fix a race between lu_object_find() finding an object and its
2514 concurrent finalization. This race is (most likely) not possible
2515 on the server, but might happen on the client.
2519 Description: Introduce lu_ref interface.
2520 Details : lu_ref is a debugging module allowing to track references to
2521 a given object. It is quite cpu expensive, and has to be
2522 explicitly enabled with --enable-lu_ref. See usage description
2527 Description: Factor lu_site procfs stats into a separate function.
2528 Details : Separate lu_site stats printing code into a separate function
2529 that can be reused on a client.
2533 Description: Constify instances of struct {lu,dt,md}_device_operations.
2534 Details : Constify instances of struct {lu,dt,md}_device_operations.
2538 Description: Introduce struct md_site and move meta-data specific parts of
2539 struct lu_site here.
2540 Details : Move md-specific fields out of struct lu_site into special struct
2541 md_site, so that lu_site can be used on a client.
2545 Description: Kill mdd_lov_destroy().
2546 Details : Remove unused mdd code.
2550 Description: Add st_block checking to multistat.c.
2551 Details : Add st_block checking to multistat.c.
2555 Description: Add lu_ref support to struct obd_device.
2556 Details : Add lu_ref tracking to obd_device.
2560 Description: Kill obd_set_fail_loc().
2561 Details : Remove unused code.
2565 Description: Add special type for ptlrpc_request interpret functions.
2566 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2567 there. Actual usage will be in CLIO.
2571 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2572 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2573 doesn't work with lockdep.
2577 Description: Add rwv.c test program.
2578 Details : New testing program exercising readv(2) and writev(2) (Qian).
2582 Description: Add sendfile.c test program.
2583 Details : New testing program exercising sendfile(2) (Jay).
2587 Description: Ratelimit a message that can be very frequent.
2588 Details : Ratelimit a memory allocation failure message that can
2593 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2594 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2598 Description: Liblustre build fixes.
2599 Details : Liblustre build fixes.
2603 Description: libcfs: add cfs_{need,cond}_resched() interface.
2604 Details : libcfs: add cfs_{need,cond}_resched() definition and
2605 implementations for Linux, NT, and liblustre.
2607 Severity : enhancement
2609 Description: More exported tunables for mballoc
2610 Details : Add support for tunable preallocation window and new tunables for
2611 large/small requests
2615 Description: Detect corruption of block bitmap and checking for preallocations
2616 Details : Checks validity of on-disk block bitmap. Also it does better
2617 checking of number of applied preallocations. When corruption is
2618 found, it turns filesystem readonly to prevent further corruptions.
2622 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2623 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2624 window to 0. If it does hit cache miss, and read-pattern is still
2625 stride-io mode, does not reset the stride window, but also does not
2626 increase the stride window length in this case.
2630 Frequency : only for big-endian servers
2631 Description: Check if system is big-endian while mounting fs with extents feature
2632 Details : Mounting a filesystem with extents feature will fail on big-endian
2633 systems since ext3-based ldiskfs is not supported on big-endian
2634 systems. This can be over-riden with "bigendian_extents" mount option.
2636 Severity : enhancement
2638 Description: The root squash functionality
2639 Details : A security feature, which is to prevent users from being able
2640 to mount lustre on their desktop, run as root, and delete
2641 all of the files in the filesystem. The goal is accomplished by
2642 remapping user id (UID) and group id (GID) of the root user to
2643 a UID and GID specified by the system administartor via Lustre
2644 configuration management server (MGS). The functionality also
2645 allows to specify sets of clients for which the remapping does
2650 Description: Excessive recovery window
2651 Details : With AT enabled, the recovery window can be excessively long (6000+
2652 seconds). To address this problem, we no longer use
2653 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2654 timeout no longer depends on the service time, it is set to
2655 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2656 time via pb_service_time.
2660 Description: Watchdog triggered on MDS failover
2661 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2662 the OSTs know that the MDS "UUID" can be reused for the same export
2663 from a different NID, so we do not need to wait for the export to be
2667 Frequency : rare, only if using MMP with Linux RAID
2669 Description: MMP doesn't work with Linux RAID
2670 Details : While using HA for Lustre servers with Linux RAID, it is possible
2671 that MMP will not detect multiple mounts. To make this work we
2672 need to unplug the device queue in RAID when the MMP block is being
2673 written. Also while reading the MMP block, we should read it from
2674 disk and not the cached one.
2676 Severity : enhancement
2678 Description: open file using fid
2679 Details : A file can be opened using just its fid, like
2680 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2683 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2686 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2687 Details : add proper locking for creation of jbd_4k slab cache
2691 Description: MMP check in ext3_remount() fails without displaying any error
2692 Details : When multiple mount protection fails during remount, proper error
2695 Severity : enhancement
2697 Description: Allow stripe size to be up to 4G-64k
2698 Details : Fix math logic to allow large stripe sizes.
2702 Description: add check for >8TB ldiskfs filesystems
2703 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2704 Don't allow >8TB ldiskfs filesystems to be mounted without
2705 force_over_8tb mount option
2710 Descriptoin: OSS complains "too many blocks" and crashes
2711 Details : In filter_commitrw_write():
2712 1. lquota_chkquota();
2713 2. obdo_from_inode();
2714 3. lquota_pending_commit();
2715 "2." has changed uid and gid in this case, then caused
2716 lquota_chkquota() and lquota_pending_commit() to use
2717 different uid and gid. That is the root of the bug.
2721 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2722 Details : Detect and print out the 'stripe count = -1' case properly.
2726 Description: Out or order replies might be lost on replay
2727 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2728 request with tid smaller then one currently being added, add it
2729 to the start, not end of the list.
2731 --------------------------------------------------------------------------------
2733 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2735 * Support for kernels:
2736 2.6.5-7.283 (SLES 9),
2737 2.6.9-55.EL (RHEL 4),
2738 2.6.16.46-0.14 (SLES 10),
2739 2.6.18.8 vanilla (kernel.org)
2740 * Client support for unpatched kernels:
2741 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2742 2.6.16 - 2.6.22 vanilla (kernel.org)
2743 * Due to recently discovered recovery problems, we do not recommend
2744 using patchless RHEL 4 clients with this or any earlier release.
2745 * Recommended e2fsprogs version: 1.39.cfs8
2746 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2747 * Starting with this release, the ldiskfs backing filesystem required
2748 by Lustre is now in its own package, lustre-ldiskfs. This package
2749 should be installed. It is versioned separately from Lustre and
2750 may be released separately in future.
2754 Description: block reactivating mgc import until all deactivates complete
2755 Details : Fix race when failing back MDT/MGS to itself (testing)
2758 Frequency : at statup only
2760 Description: mds_lov_synchronize race leads to various problems
2761 Details : simultaneous MDT->OST connections at startup can cause the
2762 sync to abort, leaving the OSC in a bad state.
2764 Severity : enhancement
2766 Description: add optional extra BUILD_VERSION info
2767 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2768 to override the lustre version.
2771 Frequency : 2.6.18 servers only
2773 Description: ll_kern_mount() doesn't release the module reference
2774 Details : The ldiskfs module reference count never drops down to 0
2775 because ll_kern_mount() doesn't release the module reference.
2780 Description: server LBUG when using old ost_num_threads parameter
2781 Details : Accept the old ost_num_threads parameter but warn that it
2782 is deprecated, and fix an off-by-one error that caused an LBUG.
2787 Description: Transient SCSI error results in persistent IO issue
2788 Details : iobuf->dr_error is not reinitialized to 0 between two
2792 Frequency : sometimes when underlying device returns I/O errors
2794 Description: OSTs not going read-only during write failures
2795 Details : OSTs are not remounted read-only when the journal commit threads
2796 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2797 instead of the ext3 wrappers.
2801 Description: poor connect scaling with increasing client count
2802 Details : Don't run filter_grant_sanity_check for more than 100 exports
2803 to improve scaling for large numbers of clients.
2806 Frequency : SLES10 only
2808 Description: sanity-quota.sh quotacheck failed: rc = -22
2809 Details : Quotas cannot be enabled on SLES10.
2812 Frequency : liblustre clients only
2814 Description: getdirentries does not give error when run on compute nodes
2815 Details : getdirentries does not fail when the size specified as an argument
2816 is too small to contain at least one entry
2818 Severity : enhancement
2820 Description: Add LNET router traceability for debug purposes
2821 Details : If a checksum failure occurs with a router as part of the
2822 IO path, the NID of the last router that forwarded the bulk data
2823 is printed so it can be identified.
2828 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2829 Details : Due to a race condition, liblustre clients were occasionally
2830 evicted incorrectly.
2832 Severity : enhancement
2834 Description: lfs setstripe use optional parameters instead of postional
2837 Severity : enhancement
2839 Description: Nanosecond timestamp support for ldiskfs
2840 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2841 resolution timestamps. There is not yet support for this at
2842 the Lustre filesystem level.
2845 Frequency : during server recovery
2847 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2848 Details : request with rq_no_resend flag not awake l_wait_event if they get
2852 Frequency : nfs export on patchless client
2854 Description: connectathon hang when test nfs export over patchless client
2855 Details : Disconnected dentry cannot be found with lookup, so we do not need
2856 to unhash it or make it invalid
2859 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2860 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2861 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2866 Description: open req refcounting wrong on reconnect
2867 Details : If reconnect happened between getting open reply from server and
2868 call to mdc_set_replay_data in ll_file_open, we will schedule
2869 replay for unreferenced request that we are about to free.
2870 Subsequent close will crash in variety of ways.
2871 Check that request is still eligible for replay in
2872 mdc_set_replay_data().
2877 Description: disable writes to filesystem when reading health_check file
2878 Details : the default for reading the health_check proc file has changed
2879 to NOT do a journal transaction and write to disk, because this
2880 can cause reads of the /proc file to hang and block HA state
2881 checking on a healthy but otherwise heavily loaded system. It
2882 is possible to return to the previous behaviour during configure
2883 with --enable-health-write.
2885 Severity : enhancement
2887 Description: 64-bit inode version
2888 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2889 to the inode. This will be required for version-based recovery.
2894 Description: MDS fails to start if a duplicate client export is detected
2895 Details : in some rare cases it was possible for a client to connect to
2896 an MDS multiple times. Upon recovery the MDS would detect this
2897 and fail during startup. Handle this more gracefully.
2899 Severity : enhancement
2901 Description: Add -o localflock option to simulate old noflock
2903 Details : This will achieve local-only flock/fcntl locks
2909 Description: log_commit_thread vs filter_destroy race leads to crash
2910 Details : Take import reference before releasing llog record semaphore
2915 Description: Wrong request locking in request set processing
2916 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2917 imp_delayed_list, in this place should be used imp_lock.
2920 Frequency : when reconnection
2922 Description: Grant Leak when osc reconnect to OST
2923 Details : When osc reconnect ost, OST(filter) should check whether it
2924 should grant more space to client by comparing fed_grant and
2925 cl_avail_grant, and return the granted space to client instead
2926 of "new granted" space, because client will call osc_init_grant
2927 to update the client grant space info.
2930 Frequency : when client reconnect to OST
2932 Description: Grant Leak when osc do resend and replay bulk write
2933 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2934 bulk write request, because the grant info while be sync between
2935 OSC and OST when reconnect, and we should ignore the grant info
2936 these of resend/replay write req.
2941 Description: Grant space more than available left space sometimes.
2942 Details : When then OST is about to be full, if two bulk writing from
2943 different clients came to OST. Accord the avaliable space of the
2944 OST, the first req should be permitted, and the second one
2945 should be denied by ENOSPC. But if the seconde arrived before
2946 the first one is commited. The OST might wrongly permit second
2947 writing, which will cause grant space > available space.
2950 Frequency : when client is evicted
2952 Description: Grant might be wrongly erased when osc is evicted by OST
2953 Details : when the import is evicted by server, it will fork another
2954 thread ptlrpc_invalidate_import_thread to invalidate the
2955 import, where the grant will be set to 0. While the original
2956 thread will update the grant it got when connecting. So if
2957 the former happened latter, the grant will be wrongly errased
2958 because of this race.
2963 Description: Checking Stale with correct fid
2964 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2965 to check whether it is stale, because sometimes, we want the
2966 enqueue happened anyway, and op_data.fid2 will not be initialized.
2968 Severity : enhancement
2970 Description: update patchless client
2971 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2974 Frequency : only with 2.4 kernel
2976 Description: random memory corruption
2977 Details : size of struct ll_inode_info is to big for union inode.u and this
2978 can be cause of random memory corruption.
2983 Description: Memory leak in recovery
2984 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2985 It should also check obd_fail before fsfilt_start, otherwise if
2986 fsfilt_start return -EROFS,(failover mds during mds recovery).
2987 then the req will return with repmsg->transno = 0 and rc = EROFS.
2988 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2989 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2990 be freed no matter whether fsfilt_commit success or not.
2993 Frequency : only with huge count clients
2995 Description: Prevents from taking the superblock lock in llap_from_page for
2997 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2998 for llap_from_page call in ll_removepage prevents from taking the
2999 superblock lock for a soon died page.
3004 Description: Not check open intent error before release open handle
3005 Details : in some rare cases, the open intent error is not checked before
3006 release open handle, which may cause
3007 ASSERTION(open_req->rq_transno != 0), because it tries to release
3008 the failed open handle.
3013 Description: Set cat log bitmap only after create log success.
3014 Details : in some rare cases, the cat log bitmap is set too early. and it
3015 should be set only after create log success.
3019 Description: Accessing a block bevice can re-enable I/O when Lustre is
3020 tearing down a device.
3021 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3026 Description: service threads may hog cpus when there are a lot of requests
3027 Details : Insert cond_resched to give other threads a chance to use some CPU
3032 Description: the cat log was not initialized in recovery
3033 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3034 unlink log on mds will not be initialized until mds post
3035 recovery. And also in mds post recovery, the unlink log will
3036 initialization will be done asynchronausly, so there will be race
3037 between add unlink log and unlink log initialization.
3041 Description: brw_stats were being printed incorrectly
3042 Details : brw_stats were being printed as log2 but all of them were not
3043 recorded as log2. Also remove some code duplication arising from
3044 filter_tally_{read,write}.
3048 Frequency : rare, only in recovery.
3049 Description: ASSERTION(req->rq_type != LI_POISON) failed
3050 Details : imp_lock should be held while iterating over imp_sending_list for
3051 prevent destroy request after get timeout in ptlrpc_queue_wait.
3055 Description: replay-single.sh test 52 fails
3056 Details : A lock's skiplist need to be cleanup when it being unlinked
3057 from its resource list.
3061 Description: Short directio read returns full requested size rather than
3063 Details : Direct I/O operations should return actual amount of bytes
3064 transferred rather than requested size.
3066 Severity : enhancement
3068 Description: metadata RPC reduction (e.g. for rm performance)
3069 Details : decrease the amount of synchronous RPC between clients and servers
3070 by canceling conflicing lock before the operation on the client
3071 and packing thier handles into the main operation RPC to server.
3073 Severity : enhancement
3075 Description: Async OSC create to avoid the blocking unnecessarily.
3076 Details : If a OST has no remain object, system will block on the creating
3077 when need to create a new object on this OST. Now, ways use
3078 pre-created objects when available, instead of blocking on an
3079 empty osc while others are not empty. If we must block, we block
3080 for the shortest possible period of time.
3084 Description: improve handling recoverable errors
3085 Details : if request processig with error which can be recoverable on server
3086 request should be resend, otherwise page released from cache and
3089 Severity : enhancement
3091 Description: refine locking for avoid write wrong info into lov_objid file
3092 Details : fix possible races with add new target and write/update data in
3095 --------------------------------------------------------------------------------
3097 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3102 Frequency : on some architectures
3104 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3105 Details : Uninitialized flags sometimes cause configuration commands to
3109 Frequency : patchless clients only
3111 Description: missing __iget() symbol export
3112 Details : The __iget() symbol export is missing. To avoid the need for
3113 this on patchless clients the deathrow inode reaper is turned
3114 off, and we depend on the VM to clean up old inodes. This
3115 dependency was during via the fix for bug 12181.
3117 --------------------------------------------------------------------------------
3119 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3121 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3122 INTEROPERATE with older versions automatically. In many cases a
3123 special upgrade step is needed. Please read the
3124 user documentation before upgrading any part of a live system.
3125 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3126 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3127 * WARNING: Lustre configuration and startup changes are required with
3128 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3132 Severity : enhancement
3134 Description: MountConf
3135 Details : Lustre configuration is now managed via mkfs and mount
3136 commands instead of lmc and lconf. New obd types (MGS, MGC)
3137 are added for dynamic configuration management. See
3138 https://mail.clusterfs.com/wikis/lustre/MountConf for
3141 Severity : enhancement
3143 Description: dynamic OST addition
3144 Details : OSTs can now be added to a live filesystem
3146 Severity : enhancement
3148 Description: startup order invariance
3149 Details : MDTs and OSTs can be started in any order. Clients only
3150 require the MDT to complete startup.
3152 Severity : enhancement
3154 Description: parallel, asynchronous orphan cleanup
3155 Details : orphan cleanup is now performed in separate threads for each
3156 OST, allowing parallel non-blocking operation.
3158 Severity : enhancement
3160 Description: optimized stripe assignment
3161 Details : stripe assignments are now made based on ost space available,
3162 ost previous usage, and OSS previous usage, in order to try
3163 to optimize storage space and networking resources.
3165 Severity : enhancement
3167 Description: Permanently set tunables
3168 Details : All writable /proc/fs/lustre tunables can now be permanently
3169 set on a per-server basis, at mkfs time or on a live system.
3171 Severity : enhancement
3173 Description: Lustre message v2
3174 Details : Add lustre message format v2.
3176 Severity : enhancement
3178 Description: client OST exclusion list
3179 Details : Clients can be started with a list of OSTs that should be
3180 declared "inactive" for known non-responsive OSTs.
3184 Description: SPEC SFS validation failure on NFS v2 over lustre.
3185 Details : Changes the blocksize for regular files to be 2x RPC size,
3186 and not depend on stripe size.
3188 Severity : enhancement
3190 Description: Multiple MD RPCs in flight.
3191 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3192 To avoid overly-overloading MDS, introduce a limit on number of
3193 MDS RPCs in flight for a single client and add /proc controls
3194 to adjust this limit.
3196 Severity : enhancement
3198 Description: client read/write statistics
3199 Details : Add client read/write call usage stats for performance
3200 analysis of user processes.
3201 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3202 file access. extents_stats shows chunk size distribution.
3203 extents_stats_per_process show chunk size distribution per
3206 Severity : enhancement
3208 Description: mds statistics
3209 Details : Add detailed mds operations statistics in
3210 /proc/fs/lustre/mds/*/stats.
3214 Description: Failure of copying files with lustre special EAs.
3215 Details : Client side always return success for setxattr call for lustre
3216 special xattr (currently only "trusted.lov").
3221 Description: Refcount LNET uuids
3222 Details : The global LNET uuid list grew linearly with every startup;
3223 refcount repeated list entries instead of always adding to
3226 Severity : enhancement
3228 Description: Dynamic service threads
3229 Details : Within a small range, start extra service threads
3230 automatically when the request queue builds up.
3233 Frequency : mixed-endian client/server environments
3235 Description: mixed-endian crashes
3236 Details : The new msg_v2 system had some failures in mixed-endian
3239 Severity : enhancement
3241 Description: Easy OST removal
3242 Details : OSTs can be permanently deactivated with e.g. 'lctl
3243 conf_param lustre-OST0001.osc.active=0'
3245 Severity : enhancement
3247 Description: MGS proc entries
3248 Details : Added basic proc entries for the MGS showing what filesystems
3251 Severity : enhancement
3253 Description: provide MGS failover
3254 Details : Added config lock reacquisition after MGS server failover.
3256 Severity : enhancement
3258 Description: add Linux 2.4 support
3259 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3263 Description: a large application tries to do I/O to the same resource and dies
3264 in the middle of it.
3265 Details : Check the req->rq_arrival time after the call to
3266 ost_brw_lock_get(), but before we do anything about
3267 processing it & sending the BULK transfer request. This
3268 should help move old stale pending locks off the queue as
3269 quickly as obd_timeout.
3272 Frequency : when an incorrect nid is specified during startup
3274 Description: ptlrpc connect to non-existant node causes kernel crash
3275 Details : LNET can't be re-entered from an event callback, which
3276 happened when we expire a message after the export has been
3277 cleaned up. Instead, hand the zombie cleanup off to another
3280 Severity : enhancement
3282 Description: plain/inodebits lock performance improvement
3283 Details : Grouping plain/inodebits in granted list by their request modes
3284 and bits policy, thus improving the performance of search through
3288 Frequency : only if OST filesystem is corrupted
3290 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3291 Details : for a short time RPCs with bulk IO are in the replay list,
3292 but replay of bulk IOs is unimplemented. If the OST filesystem
3293 is corrupted due to disk cache incoherency and then replay is
3294 started it is possible to trip an assertion. Avoid putting
3295 committed RPCs into the replay list at all to avoid this issue.
3298 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3300 Description: System hang on startup
3301 Details : This bug allowed the liblustre (e.g. catamount) client to
3302 return to the app before handling all startup RPCs. This
3303 could leave the node unresponsive to lustre network traffic
3304 and manifested as a server ptllnd timeout.
3306 Severity : enhancement
3308 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3309 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3310 boolean to control whether to print peer debug info when a
3311 client's RPC times out.
3314 Frequency : only for kernels with patches from Lustre below 1.4.3
3316 Description: Remove old rdonly API
3317 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3320 Frequency : only for devices with external journals
3322 Description: Set external device read-only also
3323 Details : During a commanded failover stop, we set the disk device
3324 read-only while the server shuts down. We now also set any
3325 external journal device read-only at the same time.
3328 Frequency : when upgrading from 1.4 while trying to change parameters
3330 Description: The wrong (new) MDC name was used when setting parameters for
3331 upgraded MDT's. Also allows changing of OSC (and MDC)
3332 parameters if --writeconf is specified at tunefs upgrade time.
3335 Frequency : when setting specific ost indicies
3337 Description: QOS code breaks on skipped indicies
3338 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3339 created with --index need not be sequential.
3341 Severity : enhancement
3343 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3344 Details : The uninit_groups feature works in conjunction with the kernel
3345 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3346 up the pass1 processing of e2fsck. This is a read-only feature
3347 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3348 mount filesystems that have had this feature enabled.
3350 Severity : enhancement
3352 Description: Improve multi-block allocation algorithm to avoid fragmentation
3353 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3354 allocation locality and avoid filesystem fragmentation.
3356 ------------------------------------------------------------------------------
3358 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3360 * Support for kernels:
3361 2.4.21-47.0.1.EL (RHEL 3)
3362 2.6.5-7.283 (SLES 9)
3363 2.6.9-42.0.10.EL (RHEL 4)
3364 2.6.12.6 vanilla (kernel.org)
3365 2.6.16.27-0.9 (SLES 10)
3366 * Recommended e2fsprogs version: 1.39.cfs5
3368 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3372 Frequency : occasional, depends on client load and configuration
3373 Bugzilla : 12181, 12203
3374 Description: data loss for recently-modified files
3376 Details : In some cases it is possible that recently written or created
3377 files may not be written to disk in a timely manner (this should
3378 normally be within 30s unless client IO load is very high).
3379 The problem appears as zero-length files or files that are a
3380 multiple of 1MB in size after a client crash or client eviction
3381 that are missing data at the end of the file.
3383 This problem is more likely to be hit on clients where files are
3384 repeatedly created and unlinked in the same directory, clients
3385 have a large amount of RAM, have many CPUs, the filesystem has
3386 many OSTs, the clients are rebooted frequently, and/or the files
3387 are not accessed by other nodes after being written.
3389 The presence of the problem can be detected by looking at
3390 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3391 smaller than the second (nr_unused) then dirty files will not
3392 be flushed automatically to disk. "sync; sleep 10" should be
3393 run several times on the node before unmounting it to update
3394 Lustre (this is also safe to run on nodes without this problem).
3396 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3397 that can cause this same problem, so customers using that kernel
3398 also need to update the kernel in addition to Lustre. In order
3399 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3401 It is normal that files written just before a client crash (less
3402 than 30s) may not yet have been flushed to disk, even for local
3406 Frequency : frequent on thin XT3 nodes
3408 Description: UUID collision on thin XT3 Linux nodes
3409 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3410 randomly, since we relied on an insufficiently-seeded PRNG.
3415 Description: OSS hangs after "All ost request buffers busy"
3416 Details : A deadlock between quota and journal operations caused OSS
3417 hangs after printing "All ost request buffers busy."
3420 Frequency : always on liblustre builds
3422 Description: Cleanup compiler warnings on liblustre
3425 Frequency : always on liblustre builds on XT3
3427 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3431 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3432 > 32000 subdirectories
3433 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3434 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3435 subdirectory count crosses 32000. This will aid e2fsck to
3436 correctly handle more than 32000 subdirectories.
3439 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3441 Description: System hang on startup
3442 Details : This bug allowed the liblustre (e.g. catamount) client to
3443 return to the app before handling all startup RPCs. This
3444 could leave the node unresponsive to lustre network traffic
3445 and manifested as a server ptllnd timeout.
3447 Severity : enhancement
3449 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3450 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3451 boolean to control whether to print peer debug info when a
3452 client's RPC times out.
3457 Description: make O_SYNC working on 2.6 kernels
3458 Details : 2.6 kernels use different method for mark pages for write,
3459 so need add a code to lustre for O_SYNC work.
3464 Description: Failure to close file and release space on NFS
3465 Details : Put inode details into lock acquired in ll_intent_file_open.
3466 Use mdc_intent_lock in ll_intent_open to properly
3467 detect all kind of errors unhandled by mdc_enqueue.
3472 Description: proc file read during shutdown sometimes raced obd removal,
3474 Details : Add lock to prevent obd access after proc file removal.
3477 Frequency : Only for files larger than 4GB on 32-bit clients.
3479 Description: improperly doing page alignment of locks
3480 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3481 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3486 Description: RPCs being resent when they shouldn't be
3487 Details : Some RPCs that should not be resent are being resent. This
3488 can cause inconsistencies in the RPC state machine. Do not
3489 resend such requests.
3492 Frequency : rare, only with NFS export
3494 Description: Crash on NFS re-export node
3495 Details : under very unusual load conditions an assertion is hit in
3496 ll_intent_file_open()
3499 Frequency : only if OST filesystem is corrupted
3501 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3502 Details : for a short time RPCs with bulk IO are in the replay list,
3503 but replay of bulk IOs is unimplemented. If the OST filesystem
3504 is corrupted due to disk cache incoherency and then replay is
3505 started it is possible to trip an assertion. Avoid putting
3506 committed RPCs into the replay list at all to avoid this issue.
3511 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3512 Details : Large single O_DIRECT read and write calls can fail to allocate
3513 a sufficiently large buffer to process the request. In case of
3514 allocation failure the allocation is retried with a smaller
3515 buffer and broken into smaller requests.
3517 Severity : enhancement
3519 Description: Add -o localflock option to simulate old noflock behaviour.
3520 Details : This will achieve local-only flock/fcntl locks coherentness.
3525 Description: versioning check is incomplete
3526 Details : Checking the version difference of client vs. server, report
3527 error if the gap is too big.
3532 Description: add support PG_writeback bit
3533 Details : add support for PG_writeback bit for Lustre, for more carefull
3534 work with page cache in 2.6 kernel. This also fix some deadlocks
3535 and remove hack for work O_SYNC with 2.6 kernel.
3537 Severity : enhancement
3539 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3540 Details : The uninit_groups feature works in conjunction with the kernel
3541 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3542 up the pass1 processing of e2fsck. This is a read-only feature
3543 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3544 mount filesystems that have had this feature enabled.
3546 Severity : enhancement
3548 Description: Improve multi-block allocation algorithm to avoid fragmentation
3549 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3550 allocation locality and avoid filesystem fragmentation.
3552 ------------------------------------------------------------------------------
3554 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3556 * Support for kernels:
3557 2.6.9-42.0.3.EL (RHEL 4)
3558 2.6.5-7.276 (SLES 9)
3559 2.4.21-47.0.1.EL (RHEL 3)
3560 2.6.12.6 vanilla (kernel.org)
3561 2.6.16.21-0.8 (SLES10)
3562 * Recommended e2fsprogs version: 1.39.cfs2-0
3564 * The backwards-compatible /proc/sys/portals symlink has been removed
3565 in this release. Before upgrading, please ensure that you change
3566 any configuration scripts or /etc/sysctl.conf files that access
3567 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3568 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3569 in advance of the upgrade on any system running Lustre 1.4.6 or
3570 newer, since /proc/sys/lnet was added in that version.
3571 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3575 Frequency : only when quota is used
3577 Description: avoid scanning export list for quota master
3578 Details : Change the algorithms to avoid scanning export list in order
3579 to improve the efficiency.
3582 Frequency : MDS failover only, very rarely
3584 Description: "went back in time" messages on mds failover
3585 Details : The greatest transno may be lost when the current operation
3586 finishes with an error (transno==0) and the client's last_rcvd
3587 record is over-written. Save the greatest transno in the
3588 mds_last_transno for this case.
3591 Frequency : always for specific kernels and striping counts
3593 Description: client may get "Matching packet too big" without ACL support
3594 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3595 when trying to access files in certain configurations. The
3596 clients should in fact be denied when mounting because they do
3597 not understand ACLs.
3600 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3602 Description: many clients connecting with IO in progress causes connect timeouts
3603 Details : Avoid synchronous journal commits to avoid delays caused by many
3604 clients connecting/disconnecting when bulk IO is in progress.
3605 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3606 OST_IO_PORTAL to avoid delays behind potentially many pending
3610 Frequency : occasionally with multiple writers to a single file
3612 Description: shared writes to file may result in wrong size reported by stat()
3613 Details : Allow growing of kms when extent lock is cancelled
3616 Frequency : always with random mmap IO to multi-striped file
3618 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3619 Details : Only if the hole is at the end of OST object so that kms is too
3620 small. Fix is to increase kms accordingly in ll_nopage.
3623 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3625 Description: writes to a missing object would leak memory on the OST
3626 Details : If there is an inconsistency between the MDS and OST filesystems,
3627 such that the MDS references an object that doesn't exist, writes
3628 to that object will leak memory due to incorrect cleanup in the
3629 error handling path, eventually running out of memory on the OST.
3634 Description: Creating too long symlink causes lustre errors
3635 Details : Check symlink and name lengths before sending requests to MDS.
3638 Frequency : only if flock is enabled (not on by default)
3640 Description: posix locks not released on fd closure on 2.6.9+
3641 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3642 kernels, this caused such locks not to be released on fd close and
3643 then assertions on fs unmount about still used locks.
3646 Frequency : MDS failover only, very rarely
3648 Description: clients may get ASSERTION(granted_lock != NULL)
3649 Details : When request was taking a long time, and a client was resending
3650 a getattr by name lock request. The were multiple lock requests
3651 with the same client lock handle and
3652 mds_getattr_name->fixup_handle_for_resent_request found one of the
3653 lock handles but later failed with ASSERTION(granted_lock != NULL).
3658 Description: handle->h_buffer_credits > 0, assertion failure
3659 Details : h_buffer_credits is zero after truncate, causing assertion
3660 failure. This patch extends the transaction or creates a new
3664 Frequency : NFS re-export or patchless client
3665 Bugzilla : 11179, 10796
3666 Description: Crash on NFS re-export node (__d_move)
3667 Details : We do not want to hash the dentry if we don't have a lock.
3668 But if this dentry is later used in d_move, we'd hit uninitialised
3669 list head d_hash, so we just do this to init d_hash field but
3670 leave dentry unhashed.
3673 Frequency : NFS re-export or patchless client
3675 Description: NFS exports has problem with symbolic link
3676 Details : lustre client didn't properly install dentry when re-exported
3677 to NFS or running patchless client.
3680 Frequency : NFS re-export or patchless client
3682 Description: Various nfs/patchless fixes.
3683 Details : fixes reuse disconected alias for lookup process - this fixes
3684 warning "find_exported_dentry: npd != pd",
3685 fix permission error with open files at nfs.
3686 fix apply umask when do revalidate.
3689 Frequency : occasional
3691 Description: Crash on NFS re-export node
3692 Details : calling clear_page() on the wrong pointer triggered oops in
3693 generic_mapping_read().
3696 Frequency : rarely, using O_DIRECT IO
3698 Description: unaligned directio crashes client with LASSERT
3699 Details : check for unaligned buffers before trying any requests.
3702 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3704 Description: stale data returned from RAID cache
3705 Details : If only a small amount of IO is done to the RAID device before
3706 reading it again it is possible to get stale data from the RAID
3707 cache instead of reading it from disk.
3710 Frequency : always for sles10 kernel
3712 Description: sles10 support
3713 Details : ll_follow_link: compile fixes and using of nd_set_link
3714 under newer kernels.
3717 Frequency : depends on arch, kernel and compiler version, always on sles10
3720 Description: recursive or deep enough symlinks cause stack overflow
3721 Details : getting rid of large stack-allocated variable in
3725 Frequency : depends on hardware
3727 Description: lustre write performance loss in the SLES10 kernel
3728 Details : the performance loss is caused by using of write barriers in the
3729 ext3 code. The SLES10 kernel turns barrier support on by
3730 default. The fix is to undo that change for ldiskfs.
3732 ------------------------------------------------------------------------------
3734 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3736 * Support for kernels:
3737 2.6.9-42.0.3EL (RHEL 4)
3738 2.6.5-7.276 (SLES 9)
3739 2.4.21-47.0.1.EL (RHEL 3)
3740 2.6.12.6 vanilla (kernel.org)
3744 Frequency : quota enabled and large files being deleted
3746 Description: releasing more than 4GB of quota at once hangs OST
3747 Details : If a user deletes more than 4GB of files on a single OST it
3748 will cause the OST to spin in an infinite loop. Release
3749 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3754 Description: statfs data retrieved from /proc may be stale or zero
3755 Details : When reading per-device statfs data from /proc, in the
3756 {kbytes,files}_{total,free,avail} files, it may appear
3757 as zero or be out of date.
3760 Frequency : systems with MD RAID1 external journal devices
3762 Description: lconf's call to blkid is confused by RAID1 journal devices
3763 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3764 of returning all block devices that match the journal UUID.
3767 Frequency : always, for aggregate stripe size over 4GB
3769 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3770 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3771 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3774 Frequency : always if "lfs find" used on a local file/directory
3776 Description: "lfs find" segfaults if used on a local file/directory
3777 Details : The case where a directory component was not specified wasn't
3778 handled correctly. Handle this properly.
3781 Frequency : always on ppc64
3783 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3784 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3785 perform endianness conversion.
3788 Frequency : rarely (truncate to non-zero file size after write under load)
3789 Bugzilla : 10730, 10687
3790 Description: Files padded with zeros to next 4K multiple
3791 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3792 it is possible that files that are truncated to a non-zero size
3793 immediately after being written are filled with zero bytes beyond
3794 the truncated size. No file data is lost.
3796 Severity : enhancement
3798 Description: Allow recovery/failover for liblustre clients.
3799 Details : liblustre clients were unaware of failover configurations until
3802 Severity : enhancement
3804 Description: user file locks should fail when not mounting with flock option
3805 Details : Set up an error-returning stub in ll_file_operations.lock field
3806 to prevent incorrect behaviour when client is mounted without
3807 flock option. Also, set up properly f_op->flock field for
3811 Frequency : always on ia64
3813 Description: "lfs df" loops on printing out MDS statfs information
3814 Details : The obd_ioctl_data was not initialized and in some systems
3815 this caused a failure during the ioctl that did not return
3816 an error. Initialize the struct and return an error on failure.
3819 Frequency : SLES 9 only
3821 Description: Error of copying files with lustre special EAs as root
3822 Details : Client side always return success for setxattr call for lustre
3823 special xattr (currently only "trusted.lov").
3826 Frequency : rarely on clusters with both ia64+i386 clients
3828 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3829 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3830 clients are doing concurrent writes to the same file it is
3831 possible that the ia64 clients may LASSERT because the OST
3832 extent locks are not PAGE_SIZE aligned. Ensure that grown
3833 locks are always aligned on the request boundary.
3836 Frequency : specific use, occasional
3838 Description: Overwriting in use executable truncates on-disk binary image
3839 Details : If one node attempts to overwrite an executable in use by
3840 another node, we now correctly return ETXTBSY instead of
3841 truncating the file.
3843 Severity : enhancement
3845 Description: Async OSC create to avoid the blocking unnecessarily.
3846 Details : If a OST has no remain object, system will block on the creating
3847 when need to create a new object on this OST. Now, ways use
3848 pre-created objects when available, instead of blocking on an
3849 empty osc while others are not empty. If we must block, we block
3850 for the shortest possible period of time.
3855 Description: chmod on Lustre root is propagated to other clients
3856 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3857 invalid by the follow_mount time.
3860 Frequency : liblustre clients only
3862 Description: Race in 'instant cancel' lock handling could lead to such locks
3863 never to be granted in case of SMP MDS
3864 Details : Do not destroy not yet granted but cbpending locks in
3868 Frequency : replay/resend of open
3870 Description: non null lock assetion failure in mds_intent_policy
3871 Details : Trying to replay/resend lockless open requests resulted in
3872 mds_open() returning 0 with no lock. Now it sets a flag if
3873 a lock is going to be returned.
3875 Severity : enhancement
3877 Description: Checksum enhancements
3878 Details : New checksum enhancements allow for resending RPCs that failed
3881 Severity : enhancement
3883 Description: Tunables on number of dirty pages in cacche
3884 Details : Allow to set limit on number of dirty pages cached.
3889 Description: client crash on unmount - lock still has references
3890 Details : In some error handling cases it was possible to leak a lock
3891 reference on a client while accessing a file. This was not
3892 harmful to the client during operation, but would cause the
3893 client to crash when the filesystem is unmounted.
3896 Frequency : specific case, rare
3898 Description: ETXTBSY on mds though file not in use by client
3899 Details : ETXTBSY is no longer incorrectly returned when attempting to
3900 chmod or chown a directory that the user previously tried to
3901 execute or a currently-executing binary.
3904 Frequency : extremely rare except on liblustre-based clients
3906 Description: Lustre space not freed when files are deleted
3907 Details : Clean up open-unlinked files after client eviction. Previously
3908 the unlink was skipped and the files remained as orphans.
3913 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3914 Details : In some cases it was possible to send a blocking callback to
3915 a client doing a glimpse, even though that client didn't get
3916 a lock granted. When the glimpse lock is cancelled on the OST
3917 the freed lock is left on the waiting list and corrupted the list.
3920 Frequency : all core dumps
3922 Description: Broke core dumps to lustre
3923 Details : Negative dentry may be unhashed if parent does not have UPDATE
3924 lock, but some callers, e.g. do_coredump, expect dentry to be
3925 hashed after successful create, hash it in ll_create_it.
3927 ------------------------------------------------------------------------------
3929 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3931 * Support for kernels:
3932 2.6.9-42.0.2.EL (RHEL 4)
3933 2.6.5-7.276 (SLES 9)
3934 2.4.21-40.EL (RHEL 3)
3935 2.6.12.6 vanilla (kernel.org)
3939 Frequency : always on RHEL 3
3941 Description: Number of open files grows over time
3942 Details : The number of open files grows over time, whether or not
3943 Lustre is started. This was due to a filp leak introduced
3944 by one of our kernel patches.
3946 ------------------------------------------------------------------------------
3948 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3950 * Support for kernels:
3951 2.6.9-42.EL (RHEL 4)
3952 2.6.5-7.276 (SLES 9)
3953 2.4.21-40.EL (RHEL 3)
3954 2.6.12.6 vanilla (kernel.org)
3959 Bugzilla : 5719, 9635, 9792, 9684
3960 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3961 Details : If a server is under heavy load and cannot reply to new
3962 connection requests before the client resends the (re)connect,
3963 the connection handling code can behave badly if two service
3964 threads are concurrently handing separate (re)connections from
3965 the same client. Add better locking to the connection handling
3966 code, and ensure that only a single connection will be processed
3967 for a given client UUID, even if the lock is dropped.
3969 Severity : enhancement
3971 Description: add TCP zero-copy support to kernel
3972 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3973 sends if the hardware supports scatter-gather and checksumming.
3974 This allows socklnd to do client-write and server-read more
3975 efficiently and reduce CPU utilization from skbuf copying.
3978 Frequency : only if NFS exporting from client
3980 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3981 Details : If setting attributes on a file created under NFS that had
3982 never been opened it would be possible to oops the client
3983 if the file had no objects.
3986 Frequency : always for liblustre
3988 Description: liblustre client does MDS+OSTs setattr RPC for each write
3989 Details : When doing a write from a liblustre client, the client
3990 incorrectly issued an RPC to the MDS and each OST the file was
3991 striped over in order to update the timestamps. When writing
3992 with small chunks and many clients this could overwhelm the MDS
3993 with RPCs. In all cases it would slow down the write because
3994 these RPCs are unnecessary.
3996 Severity : enhancement
3998 Description: allow number of MDS service threads to be changed at module load
3999 Details : It is now possible to change the number of MDS service threads
4000 running. Adding "options mds mds_num_threads={N}" to the MDS's
4001 /etc/modprobe.conf will set the number of threads for the next
4002 time Lustre is restarted (assuming the "mds" module is also
4003 reloaded at that time). The default number of threads will
4004 stay the same, 32 for most systems.
4009 Description: OST crash if filesystem is unformatted or corrupt
4010 Details : If an OST is started on a device that has never been formatted
4011 or if the filesystem is corrupt and cannot even mount then the
4012 error handling cleanup routines would dereference a NULL pointer.
4017 Description: NULL pointer deref in llap_from_page.
4018 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4019 incorrect) mapping if the page was truncated/reclaimed while it was
4020 searched for. Check for this condition and skip such pages when
4021 doing readahead. Introduce extra check to llap_from_page() to
4022 verify page->mapping->host is non-NULL (so page is not anonymous).
4025 Frequency : Sometimes when using sys_sendfile
4027 Description: "page not covered by a lock" warnings from ll_readpage
4028 Details : sendfile called ll_readpage without right page locks present.
4029 Now we introduced ll_file_sendfile that does necessary locking
4030 around call to generic_file_sendfile() much like we do in
4034 Frequency : with certain MDS communication failures at client mount time
4036 Description: NULL pointer deref after failed client mount
4037 Details : a client connection request may delayed by the network layer
4038 and not be sent until after the PTLRPC layer has timed out the
4039 request. If the client fails the mount immediately it will try
4040 to clean up before the network times out the request. Add a
4041 reference from the request import to the obd device and delay
4042 the cleanup until the network drops the request.
4045 Frequency : occasionally during client (re)connect
4047 Description: assertion failure during client (re)connect
4048 Details : processing a client connection request may be delayed by the
4049 client or server longer than the client connect timeout. This
4050 causes the client to resend the connection request. If the
4051 original connection request is replied in this interval, the
4052 client may trip an assertion failure in ptlrpc_connect_interpret()
4053 which thought it would be the only running connect process.
4056 Frequency : only with obd_echo servers and clients that are rebooted
4058 Description: kernel BUG accessing uninitialized data structure
4059 Details : When running an obd_echo server it did not start the ping_evictor
4060 thread, and when a client was evicted an uninitialized data
4061 structure was accessed. Start the ping_evictor in the RPC
4062 service startup instead of the OBD startup.
4064 Severity : enhancement
4065 Bugzilla : 10193 (patchless)
4066 Description: Remove dependency on various unexported kernel interfaces.
4067 Details : No longer need reparent_to_init, exit_mm, exit_files,
4068 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4071 Frequency : rare (only users of deprecated and unsupported LDAP config)
4073 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4074 Details : LDAP apparently contains 'lustreName' attributes instead of
4075 'name'. A simple remapping of the name is sufficient.
4078 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4080 Description: waiting_locks_callback trips kernel BUG if client is evicted
4081 Details : Running with the dump_on_timeout debug flag turned on makes
4082 it possible that the waiting_locks_callback() can try to dump
4083 the Lustre kernel debug logs from an interrupt handler. Defer
4084 this log dumping to the expired_lock_main() thread.
4086 Severity : enhancement
4088 Description: Support NFS exporting on 2.6 kernels.
4089 Details : Implement non-rawops metadata methods for NFS server to use without
4090 changing NFS server code.
4093 Frequency : very rare (synthetic metadata workload only)
4095 Description: two racing renames might cause an MDS thread to deadlock
4096 Details : Running the "racer" program may cause one MDS thread to rename
4097 a file from being the source of a rename to being the target of
4098 a rename at exactly the same time that another thread is doing
4099 so, and the second thread has already enqueued these locks after
4100 doing a lookup of the target and is trying to relock them in
4101 order. Ensure that we don't try to re-lock the same resource.
4104 Frequency : only very large systems with liblustre clients
4106 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4107 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4108 all OSTs in parallel. This allows the eviction of stale liblustre
4109 clients to proceed much faster than if they were done in series,
4110 and also offers similar improvements for other set_info RPCs.
4115 Description: excessive CPU usage during initial read phase on client
4116 Details : During the initial read phase on a client, it would agressively
4117 retry readahead on the file, consuming too much CPU and impacting
4118 performance (since 1.4.5.8). Improve the readahead algorithm
4119 to avoid this, and also improve some other common cases (read
4120 of small files in particular, where "small" is files smaller than
4121 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4126 Description: MDS crash when receiving packet with unknown intent.
4127 Details : Do not LBUG in unknown intent case, just return -EFAULT
4129 Severity : enhancement
4130 Bugzilla : 9293, 9385
4131 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4132 Details : Do not serialize getattr (non-intent version) and statfs.
4135 Frequency : occasional, when OST network is overloaded/intermittent
4137 Description: client evicted by OST after bulk IO timeout
4138 Details : If a client sends a bulk IO request (read or write) the OST
4139 may evict the client if it is unresposive to its data GET/PUT
4140 request. This is incorrect if the network is overloaded (takes
4141 too long to transfer the RPC data) or dropped the OST GET/PUT
4142 request. There is no need to evict the client at all, since
4143 the pinger and/or lock callbacks will handle this, and the
4144 client can restart the bulk request.
4147 Frequency : Always when mmapping file with no objects
4149 Description: client crashes when mmapping file with no objects
4150 Details : Check that we actually have objects in a file before doing any
4151 operations on objects in ll_vm_open, ll_vm_close and
4157 Description: Request leak when working with deleted CWD
4158 Details : Introduce advanced request refcount tracking for requests
4159 referenced from lustre intent.
4161 Severity : Enhancement
4163 Description: Cache open file handles on client.
4164 Details : MDS now will return special lock along with openhandle, if
4165 requested and client is allowed to hold openhandle, even if unused,
4166 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4167 closing files for every read/write openration.
4169 Severity : Enhancement
4171 Description: Cache open negative dentries on client when possible.
4172 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4173 negative dentries on lock revocation.
4178 Description: Remounting a client read-only wasn't possible with a zconf mount
4179 Details : It wasn't possible to remount a client read-only with llmount.
4181 Severity : enhancement
4182 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4183 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4184 MPICH is included to add Lustre-specific ADIO interfaces.
4185 This is based closely on the UFS ADIO layer and only differs
4186 in file creation, in order to allow the OST striping to be set.
4187 This is user-contributed code and not supported by CFS.
4192 Description: extended inode attributes (immutable, append-only) work improperly
4193 when 2.4 and 2.6 kernels are used on client/server or vice versa
4194 Details : Introduce kernel-independent values for these flags.
4196 Severity : enhancement
4199 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4200 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4201 granularity. Now, Fractional MB granularity is be supported,
4202 this is very useful for low memory system.
4204 Severity : enhancement
4206 Description: Getattr by fid
4207 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4208 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4209 have locks and avoids getattr rpc requests for it on every stat.
4212 Frequency : Always, for filesystems larger than 2TB
4214 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4215 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4216 filesystems larger than 2TB to be mounted with mballoc enabled.
4219 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4221 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4222 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4223 filesystems larger than 2TB is mounted then the kernel will
4224 silently corrupt the start of the filesystem. CONFIG_LBD is
4225 enabled for all CFS-supported kernels, but the possibility of
4226 this happening with a modified kernel config exists.
4228 Severity : enhancement
4230 Description: add client O_DIRECT support for 2.6 kernels
4231 Details : It is now possible to do O_DIRECT reads and writes to files
4232 in the Lustre client mountpoint on 2.6 kernel clients.
4234 Severity : enhancement
4236 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4237 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4238 parallel, not waiting for response from every OST before sending
4239 a rpc to the next OST.
4244 Description: setattr vs write race when updating file timestamps
4245 Details : Client processes that update a file timestamp into the past
4246 right after writing to the file (e.g. tar) it is possible that
4247 the updated file modification time can be reset to the current
4248 time due to a race between processing the setattr and write RPC.
4250 Severity : enhancement
4252 Description: Bring 'lfs find' closer in line with regular Linux find.
4253 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4254 -print0 options and obtains all the needed info through the lustre
4257 Severity : enhancement
4259 Description: support up to 1024 configured devices on one node
4260 Details : change obd_dev array from statically allocated to dynamically
4261 allocated structs as they are first used to reduce memory usage
4266 Description: Flush dirty partially truncated pages during truncate
4267 Details : Immediatelly flush partially truncated pages in filter_setattr,
4268 this way we completely avoid having any pages in page cache on OST
4269 and can retire ugly workarounds during writes to flush such pages.
4274 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4275 Details : protect inode from truncation within vfs_unlink() context
4276 just take a reference before calling vfs_unlink() and release it
4277 when parent's i_sem is free.
4282 Description: last_id value checked outside lock on OST caused LASSERT failure
4283 Details : If there were multiple MDS->OST object precreate requests in
4284 flight, it was possible that the OST's last object id was checked
4285 outside a lock and incorrectly tripped an assertion. Move checks
4286 inside locks, and discard old precreate requests.
4289 Frequency : always, if extents are used on OSTs
4291 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4292 Details : OSTs using the extents format would not zero the high 16 bits of
4293 the index physical block number. This is not a problem for any
4294 OST filesystems smaller than 16TB, and no kernels support ext3
4295 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4296 new/modified files) and can be fixed for existing filesystems
4297 with e2fsprogs-1.39-cfs1.
4302 Description: import connection selection may be incorrect if timer wraps
4303 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4304 import connections to be ignored if the 32-bit jiffies counter
4305 wraps. Use a 64-bit jiffies counter.
4308 Frequency : very large clusters immediately after boot
4310 Description: LNET request buffers exhausted under heavy short-term load
4311 Details : If a large number of client requests are generated on a service
4312 that has previously never seen so many requests it is possible
4313 that the request buffer growth cannot keep up with the spike in
4314 demand. Instead of dropping incoming requests, they are held in
4315 the LND until the RPC service can accept more requests.
4318 Frequency : Sometimes during replay
4320 Description: Assertion failure in ll_local_open after replay.
4321 Details : If replay happened on an open request reply before we were able
4322 to set replay handler, reply will become not swabbed tripping the
4323 assertion in ll_local_open. Now we set the handler right after
4324 recognising of open request
4327 Frequency : very rare
4329 Description: kernel reports "badness in vsnprintf"
4330 Details : Reading from the "recovery_status" /proc file in small chunks
4331 may cause a negative length in lprocfs_obd_rd_recovery_status()
4332 call to vsnprintf() (which is otherwise harmless). Exit early
4333 if there is no more space in the output buffer.
4335 Severity : enhancement
4337 Description: clear OBD RPC statistics by writing to them
4338 Details : It is now possible to clear the OBD RPC statistics by writing
4339 to the "stats" file.
4344 Description: Client mtime is not the same on different clients after utimes
4345 Details : In some cases, the client was using the utimes() syscall on
4346 a file cached on another node. The clients now validate the
4347 ctime from the MDS + OSTs to determine which one is right.
4352 Description: Inability to activate failout mode
4353 Details : lconf script incorrectly assumed that in pythong string's numeric
4354 value is used in comparisons.
4357 Frequency : always with multiple stripes per file
4359 Description: Inefficient object allocation for mutli-stripe files
4360 Details : When selecting which OSTs to stripe files over, for files with
4361 a stripe count that divides evenly into the number of OSTs,
4362 the MDS is always picking the same starting OST for each file.
4363 Return the OST selection heuristic to the original design.
4368 Description: mount failures may take full timeout to return an error
4369 Details : Under some heavy load conditions it is possible that a
4370 failed mount can wait for the full obd_timeout interval,
4371 possibly several minutes, before reporting an error.
4372 Instead return an error as soon as the status is known.
4374 Frequency : quota enabled and large files being deleted
4376 Description: releasing more than 4GB of quota at once hangs OST
4377 Details : If a user deletes more than 4GB of files on a single OST it
4378 will cause the OST to spin in an infinite loop. Release
4379 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4384 Description: statfs data retrieved from /proc may be stale or zero
4385 Details : When reading per-device statfs data from /proc, in the
4386 {kbytes,files}_{total,free,avail} files, it may appear
4387 as zero or be out of date.
4390 Frequency : systems with MD RAID1 external journal devices
4392 Description: lconf's call to blkid is confused by RAID1 journal devices
4393 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4394 of returning all block devices that match the journal UUID.
4397 Frequency : always, for aggregate stripe size over 4GB
4399 Description: assertion fails when trying to use 4GB stripe size
4400 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4401 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4404 Frequency : always on ppc64
4406 Description: the first write on an ext3 filesystem with mballoc got stuck
4407 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4408 perform endianness conversion.
4410 ------------------------------------------------------------------------------
4412 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4414 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4415 INTEROPERATE with older versions automatically. Please read the
4416 user documentation before upgrading any part of a live system.
4417 * WARNING: Lustre networking configuration changes are required with
4418 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4421 * Support for kernels:
4422 2.6.9-22.0.2.EL (RHEL 4)
4423 2.6.5-7.244 (SLES 9)
4424 2.6.12.6 vanilla (kernel.org)
4427 Severity : enhancement
4428 Bugzilla : 7981/8208
4429 Description: Introduced Lustre Networking (LNET)
4430 Details : LNET is new networking infrastructure for Lustre, it includes
4431 a reorganized network configuration mode (see the user
4432 documentation for full details) as well as support for routing
4433 between different network fabrics. Lustre Networking Devices
4434 (LNDs) for the supported network fabrics have also been
4435 created for this new infrastructure.
4437 Severity : enhancement
4438 Description: Introduced Access control lists
4439 Details : clients can set ACLs on files and directories in order to have
4440 more fine-grained permissions than the standard Unix UGO+RWX.
4441 The MDS must be started with the "-o acl" mount option.
4443 Severity : enhancement
4444 Description: Introduced filesystem quotas
4445 Details : Administrators may now establish per-user quotas on the
4448 Severity : enhancement
4450 Description: Configuration change for the XT3
4451 The PTLLND is now used to run Lustre over Portals on the XT3
4452 The configure option(s) --with-cray-portals are no longer used.
4453 Rather --with-portals=<path-to-portals-includes> is used to
4454 enable building on the XT3. In addition to enable XT3 specific
4455 features the option --enable-cray-xt3 must be used.
4460 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4461 Details : A race between allocating a new llcd and re-getting the llcd_lock
4462 allowed another thread to grab newly-allocated llcd.
4464 Severity : enhancement
4466 Description: 2.6 OST async journal commit and locking fix to improve performance
4467 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4468 2.6 kernels are now async as they already were in 2.4 kernels so
4469 that they can commit concurrently with the network bulk transfer.
4470 For block-allocated files the filter allocation semaphore is held
4471 to avoid filesystem fragmentation during allocation. BKL lock
4472 removed for 2.6 xattr operations where it is no longer needed.
4477 Description: lconf incorrectly determined whether two IP networks could talk
4478 Details : In some more complicated routing and multiple-network
4479 configurations, lconf will avoid trying to make a network
4480 connection to a disjoint part of the IP space. It was doing the
4481 math incorrectly for one set of cases.
4486 Description: Fix for potential infinite loop processing records in an llog.
4487 Details : If an llog record is corrupted/zeroed, it is possible to loop
4488 forever in llog_process(). Validate the llog record length
4489 and skip the remainder of the block on error.
4492 Frequency : occasional (liblustre only)
4494 Description: liblustre could not open files whose last component is a symlink
4495 Details : sysio_path_walk() would incorrectly pass the open intent to
4496 intermediate path components.
4499 Frequency : rare (liblustre only with non-standard tuning)
4500 Bugzilla : 7201 (7350)
4501 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4502 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4503 after it has been released, only do this for referenced locks
4505 Severity : enhancement
4507 Description: specifying an (invalid) directory default stripe_size of -1
4508 would reset the directory default striping
4509 Details : stripe_size -1 was used internally to signal directory stripe
4510 removal, now use "all default" to signal dir stripe removal
4511 as a directory striping of "all default" is not useful
4514 Frequency : common for large clusters running liblustre clients
4516 Description: doing an ls when liblustre clients are running is slow
4517 Details : sending a glimpse AST to a liblustre client waits for every AST
4518 to time out, as liblustre clients will not respond. Since they
4519 cannot cache data we refresh the OST lock LVB from disk instead.
4521 Severity : enhancement
4523 Description: doing an ls at the same time as file IO can be slow
4524 Details : enqueue and other "small" requests can be blocked behind many
4525 large IO requests. Create a new OST IO portal for non-IO
4526 requests so they can be processed faster.
4529 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4531 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4532 created file with mode 000
4533 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4534 allow an ftruncate() as a non-root user to a file with mode 000.
4535 Lustre now respects this flag to disable mode checks when
4536 truncating a file owned by the user
4539 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4542 Description: Revoking locks from clients that went dead or catatonic might take
4544 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4545 cancellation of such locks instant on servers without waiting for
4546 any reply from clients. Clients drops these locks when cancel
4547 notification from server is received without replying.
4550 Frequency : liblustre-only, when liblustre client dies or becomes busy
4552 Description: Doing ls on Linux clients can take a long time with active
4554 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4555 granting such locks to it in the first place if possible. Locks
4556 are taken by proxy on the OST during the read or write and
4557 dropped immediately afterward. Add connect flags handling, do
4558 not grant locks to liblustre clients for glimpse ASTs.
4560 Severity : enhancement
4562 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4563 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4564 sequential reads and behaves badly with applications doing random
4565 reads. Improve it by reading ahead at least read region, and
4566 avoiding excessive large RPC for small reads.
4568 Severity : enhancement
4570 Description: Creating more than 1000 files for a single job may cause a load
4571 imbalance on the OSTs if there are also a large number of OSTs.
4572 Details : qos_prep_create() uses an OST index reseed value that is an
4573 even multiple of the number of available OSTs so that if the
4574 reseed happens in the middle of the object allocation it will
4575 still utilize the OSTs as uniformly as possible.
4580 Description: OST or MDS may oops in ping_evictor_main()
4581 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4582 but doesn't restart at beginning of obd_exports_timed list
4585 Severity : enhancement
4587 Description: improve by-nid export eviction on the MDS and OST
4588 Details : allow multiple exports with the same NID to be evicted at one
4589 time without re-searching the exports list.
4592 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4594 Description: MDS may oops in groups_free()
4595 Details : in rare race conditions a newly allocated group_info struct is
4596 freed again, and this can be NULL. The 2.4 compatibility code
4597 for groups_free() checked for a NULL pointer, but 2.6 did not.
4600 Frequency : common for liblustre clients doing little filesystem IO
4601 Bugzilla : 9352, 7313
4602 Description: server may evict liblustre clients accessing contended locks
4603 Details : if a client is granted a lock or receives a completion AST
4604 with a blocking AST already set it would not reply to the AST
4605 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4608 Frequency : lfs setstripe, only systems with more than 160 OSTs
4610 Description: unable to set striping with a starting offset beyond OST 160
4611 Details : llapi_create_file() incorrectly limited the starting stripe
4612 index to the maximum single-file stripe count.
4615 Frequency : LDAP users only
4617 Description: lconf did not handle in-kernel recovery with LDAP properly
4618 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4620 Severity : enhancement
4622 Description: bind OST threads to NUMA nodes to improve performance
4623 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4624 node and do their allocations there to localize memory access
4626 Severity : enhancement
4628 Description: llmount can determine client NID directly from Myrinet (GM)
4629 Details : the client NID code from gmnalnid was moved directly into
4630 llmount, removing the need to use this or specifying the
4631 client NID explicitly when mounting GM clients with zeroconf
4634 Frequency : if client is started with down MDS
4636 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4637 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4638 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4639 ptlrpc_import_delayed_req() because we didn't check if the
4640 request was interrupted, and we also didn't break out of the
4641 event loop if there was no timeout
4646 Description: data loss during non-page-aligned writes to a single file from
4647 both multiple nodes and multiple threads on one node at same time
4648 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4649 inconsistency led to false short-reads, that were cached and later
4650 used by ->prepare_write() to fill in partially written page,
4651 leading to data loss.
4654 Frequency : always, if lconf --abort_recovery used
4656 Description: lconf --abort_recovery fails with 'Operation not supported'
4657 Details : lconf was attempting to abort recovery on the MDT device and not
4660 Severity : enhancement
4662 Description: remove cleanup logs
4663 Details : replace lconf-generated cleanup logs with lustre internal
4664 cleanup routines. Eliminates the need for client-cleanup and
4667 Severity : enhancement
4669 Description: add support for EAs (user and system) on lustre filesystems
4670 Details : it is now possible to store extended attributes in the Lustre
4671 client filesystem, and with the user_xattr mount option it
4672 is possible to allow users to store EAs on their files also
4674 Severity : enhancement
4676 Description: Add possibility (config option) to show minimal available OST free
4678 Details : When compiled with --enable-mindf configure option, statfs(2)
4679 (and so, df) will return least minimal free space available from
4680 all OSTs as amount of free space on FS, instead of summary of
4681 free spaces of all OSTs.
4683 Severity : enhancement
4685 Description: do not expand extent locks acquired on OST-side
4686 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4687 by server: they are not cached anyway.
4690 Frequency : when mmap is used/binaries executed from Lustre
4692 Description: Unmmap pages before throwing them away from read cache.
4693 Details : llap_shrink cache now attempts to unmap pages before discarding
4694 them (if unmapping failed - do not discard). SLES9 kernel has
4695 extra checks that trigger if this unmapping is not done first.
4700 Description: lconf didn't resolve symlinks before checking to see whether a
4701 given mountpoint was already in use
4704 Frequency : when migrating failover services
4705 Bugzilla : 6395, 9514
4706 Description: When migrating a subset of services from a node (e.g. failback
4707 from a failover service node) the remaining services would
4708 time out and evict clients.
4709 Details : lconf --force (implied by --failover) sets the global obd_timeout
4710 to 5 seconds in order to quickly disconnect, but this caused
4711 other RPCs to time out too quickly. Do not change the global
4712 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4714 Severity : enhancement
4715 Frequency : if MDS is started with down OST
4716 Bugzilla : 9439,5706
4717 Description: Allow startup/shutdown of an MDS without depending on the
4718 availability of the OSTs.
4719 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4720 Add appropriate locking and lov-osc refcounts for safe
4721 cleaning. Add osc abort_inflight calls in case the
4722 synchronize never started.
4725 Frequency : occasional (Cray XT3 only)
4727 Description: root not authorized to access files in CRAY_PORTALS environment
4728 Details : The client process capabilities were not honoured on the MDS in
4729 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4730 been accessed by an authorized user then root was able to access
4731 the file on the local client also. The root user capabilities
4732 are now allowed on the MDS, as this environment has secure UID.
4735 Frequency : occasional
4737 Description: ldiskfs "too long searching" message happens too often
4738 Details : A debugging message (otherwise harmless) prints too often on
4739 the OST console. This has been reduced to only happen when
4740 there are fragmentation problems on the filesystem.
4745 Description: Division by zero in statfs when all OSCs are inactive
4746 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4747 lov_statfs divides by value returned by lov_get_stripecnt().
4751 Bugzilla : 9489, 3273
4752 Description: First write from each client to each OST was only 4kB in size,
4753 to initialize client writeback cache, which caused sub-optimal
4754 RPCs and poor layout on disk for the first writen file.
4755 Details : Clients now request an initial cache grant at (re)connect time
4756 and so that they can start streaming writes to the cache right
4757 away and always do full-sized RPCs if there is enough data.
4758 If the OST is rebooted the client also re-establishes its grant
4759 so that client cached writes will be honoured under the grant.
4764 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4765 Details : Now I/O RPCs go to different portal number and (presumably) fast
4766 lock requests (and glimses) and other RPCs get their own service
4767 threads pool that should be able to service those RPCs
4770 Severity : enhancement
4772 Description: Ability to exchange lustre version between client and servers and
4773 issue warnings at client side if client is too old. Also for
4774 liblustre clients there is ability to refuse connection of too old
4776 Details : New 'version' field is added to connect data structure that is
4777 filled with version info. That info is later checked by server and
4781 Frequency : rare, liblustre only.
4782 Bugzilla : 9296, 9581
4783 Description: Two simultaneous writes from liblustre at offset within same page
4784 might proceed at the same time overwriting eachother with stale
4786 Details : I/O lock withing llu_file_prwv was released too early, before data
4787 actually was hitting the wire. Extended lock-holding time until
4788 server acknowledges receiving data.
4791 Frequency : extremely rare. Never observed in practice.
4793 Description: avoid generating lustre_handle cookie of 0.
4794 Details : class_handle_hash() generates handle cookies by incrementing
4795 global counter, and can hit 0 occasionaly (this is unlikely, but
4796 not impossible, because initial value of cookie counter is
4797 selected randonly). Value of 0 is used as a sentinel meaning
4798 "unassigned handle" --- avoid it. Also coalesce two critical
4799 sections in this function into one.
4801 Severity : enhancement
4803 Description: allow liblustre clients to delegate truncate locking to OST
4804 Details : To avoid overhead of locking, liblustre client instructs OST to
4805 take extent lock in ost_punch() on client's behalf. New connection
4806 flag is added to handle backward compatibility.
4808 Severity : enhancement
4809 Bugzilla : 4928, 7341, 9758
4810 Description: allow number of OST service threads to be specified
4811 Details : a module parameter allows the number of OST service threads
4812 to be specified via "options ost ost_num_threads={N}" in the
4813 OSS's /etc/modules.conf or /etc/modprobe.conf.
4817 Bugzilla : 6146, 9635, 9895
4818 Description: servers crash with bad pointer in target_handle_connect()
4819 Details : In rare cases when a client is reconnecting it was possible that
4820 the connection request was the last reference for that export.
4821 We would temporarily drop the export reference and get a new
4822 one, but this may have been the last reference and the export
4823 was just destroyed. Get new reference before dropping old one.
4825 Severity : enhancement
4826 Frequency : if client is started with failover MDS
4828 Description: Allow multiple MDS hostnames in the mount command
4829 Details : Try to read the configuration from all specified MDS
4830 hostnames during a client mount in case the "primary"
4833 Severity : enhancement
4835 Description: Stop sending data to evicted clients as soon as possible.
4836 Details : Check if the client we are about to send or are sending data to
4837 was evicted already. (Check is done every second of waiting,
4838 for which l_wait_event interface was extended to allow checking
4839 of exit condition at specified intervals).
4842 Frequency : rare, normally only when NFS exporting is done from client
4844 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4846 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4849 Frequency : always, if srand() is called before liblustre initialization
4851 Description: Liblustre uses system PRNG disturbing its usage by user application
4852 Details : Introduce internal to lustre fast and high-quality PRNG for
4853 lustre usage and make liblustre and some other places in generic
4854 lustre code to use it.
4856 Severity : enhancement
4857 Bugzilla : 9477, 9557, 9870
4858 Description: Verify that the MDS configuration logs are updated when xml is
4859 Details : Check if the .xml configuration logs are newer than the config
4860 logs stored on the MDS and report an error if this is the case.
4861 Request --write-conf, or allow starting with --old_conf.
4863 Severity : enhancement
4865 Description: Handle symlinks in the path when checking if Lustre is mounted.
4866 Details : Resolve intermediate symlinks when checking if a client has
4867 mounted a filesystem to avoid duplicate client mounts.
4872 Description: lconf can hit an error exception but still return success.
4873 Details : The lconf command catches the Command error exception at the top
4874 level script context and will exit with the associated exit
4875 status, but doesn't ensure that this exit status is non-zero.
4880 Description: failure of ptlrpc thread startup can cause oops
4881 Details : Starting a ptlrpc service thread can fail if there are a large
4882 number of threads or the server memory is very fragmented.
4883 Handle this without oopsing.
4886 Frequency : always, only if liblustre and non-default acceptor port was used
4888 Description: liblustre cannot connect to servers with non-default acceptor port
4889 Details : tcpnal_set_default_params() was not called and was therefore
4890 ignoring the environment varaible TCPNAL_PORT, as well as other
4891 TCPNAL_ environment variables
4896 Description: two objects could be created on the same OST for a single file
4897 Details : If an OST is down, in some cases it was possible to create two
4898 objects on a single OST for a single file. No problems other
4899 than potential performance impact and spurious error messages.
4903 Bugzilla : 5681, 9562
4904 Description: Client may oops in ll_unhash_aliases
4905 Details : Client dcache may become inconsistent in race condition.
4906 In some cases "getcwd" can fail if the current directory is
4912 Description: Inode refcounting problems in NFS export code
4913 Details : link_raw functions used to call d_instantiate without obtaining
4914 extra inode reference first.
4918 Bugzilla : 9942, 9903
4919 Description: Referencing freed requests leading to crash, memleaks with NFS.
4920 Details : We used to require that call to ll_revalidate_it was always
4921 followed by ll_lookup_it. Also with revalidate_special() it is
4922 possible to call ll_revalidate_it() twice for the same dentry
4923 even if first occurence returned success. This fix changes semantic
4924 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4925 reference on a request referred from the intent.
4926 ll_intent_release() then releases such a request.
4929 Frequency : rare, normally benchmark loads only
4931 Description: unlinked inodes were kept in memory on the client
4932 Details : If a client is repeatedly creating and unlinking files it
4933 can accumulate a lot of stale inodes in the inode slab cache.
4934 If there is no other client load running this can cause the
4935 client node to run out of memory. Instead flush old inodes
4936 from client cache that have the same inode number as a new inode.
4939 Frequency : SLES9 2.6.5 kernel and long filenames only
4940 Bugzilla : 9969, 10379
4941 Description: utime reports stale NFS file handle
4942 Details : SLES9 uses out-of-dentry names in some cases, which confused
4943 the lustre dentry revalidation. Change it to always use the
4947 Frequency : rare, unless heavy write-truncate concurrency is continuous
4948 Bugzilla : 4180, 6984, 7171, 9963, 9331
4949 Description: OST becomes very slow and/or deadlocked during object unlink
4950 Details : filter_destroy() was holding onto the parent directory lock
4951 while truncating+unlinking objects. For very large objects this
4952 may block other threads for a long time and slow overall OST
4953 responsiveness. It may also be possible to get a lock ordering
4954 deadlock in this case, or run out of journal credits because of
4955 the combined truncate+unlink. Solution is to do object truncate
4956 first in one transaction without parent lock, and then do the
4957 final unlink in a new transaction with the parent lock. This
4958 reduces the lock hold time dramatically.
4961 Frequency : rare, 2.4 kernels only
4963 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4964 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4965 before unmounting their filesystem in order to avoid deadlock.
4966 The kernel_locked() function in 2.4 kernels only checks whether
4967 the kernel lock is held, not whether it is this process that is
4968 holding it as 2.6 kernels do.
4973 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4974 Details : The client ptlrpc code may be trying to reconnect to a down
4975 server before a previous connection attempt has timed out.
4976 Increase the reconnect interval to be longer than the connection
4977 timeout interval to avoid sending duplicate connections to
4981 Frequency : echo_client brw_test command
4983 Description: fix echo_client to work with OST preallocated code
4984 Details : OST preallocation code (5137) didn't take echo_client IO path
4985 into account: echo_client calls filter methods outside of any
4986 OST thread and, hence, there is no per-thread preallocated
4987 pages and buffers to use. Solution: hijack pga pages for IO. As
4988 a byproduct, this avoids unnecessary data copying.
4992 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4993 Description: Client can oops in mdc_commit_close() after open replay
4994 Details : It was possible for the MDS to return an open request with no
4995 transaction number in mds_finish_transno() if the client was
4996 evicted, but without actually returning an error. Clients
4997 would later try to replay that open and may trip an assertion
4998 Simplify the client close codepath, and always return an error
4999 from the MDS in case the open is not successful.
5002 Frequency : rare, 2.6 OSTs only
5004 Description: OST may deadlock under high load on fragmented files
5005 Details : If there was a heavy load and highly-fragmented OST filesystems
5006 it was possible to have all the OST threads deadlock waiting on
5007 allocation of biovecs, because the biovecs were not released
5008 until the entire RPC IO was completed. Instead, release biovecs
5009 as soon as they are complete to ensure forward IO progress.
5011 Severity : enhancement
5013 Description: Support for specifying external journal device at mount
5014 Details : If an OST or MDS device is formatted with an external journal
5015 device, this device major/minor is stored in the ext3 superblock
5016 and may not be valid for failover. Allow detecting and
5017 specifying the external journal at mount time.
5022 Description: Mounting an MDS with pending unlinked files may cause oops
5023 Details : target_finish_recovery() calls mds_postrecov() which returned
5024 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5025 considers this an error and immediately begins cleaning up the
5026 lov, just after starting the mds_lov process
5028 Severity : enhancement
5030 Description: Implement 'lfs df' to report actual free space on per-OST basis
5031 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5032 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5033 usage of objects; '-h' to report in human readable format.
5035 ------------------------------------------------------------------------------
5037 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5044 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5045 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5046 at startup during orphan unlinking. Instead, skip buddy update
5047 before setup, as it will be regenerated after recovery is complete.
5052 Description: If an OST is inactive, its locks might reference stale inodes.
5053 Details : lov_change_cbdata() must iterate over all namespaces, even if
5054 they are inactive to clear inode references from the lock.
5056 Severity : enhancement
5057 Frequency : occasional, if non-standard max_dirty_mb used
5059 Description: Client will block write RPCs if not enough grant
5060 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5061 then the client will block writes while waiting for another RPC
5062 to complete instead of consuming its dirty limit. With change
5063 we get improved performance when max_dirty_mb is small.
5065 Severity : enhancement
5066 Bugzilla : 3389, 6253
5067 Description: Add support for supplementary groups on the MDS.
5068 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5069 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5070 lookups for user supplementary groups into a cache.
5074 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5075 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5076 for liblustre O_EXCL is enforced by MDS.
5080 Description: readdir never returns NULL in liblustre.
5081 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5082 dentry in struct dirent.
5086 Frequency : liblustre only, depends on application IO pattern
5087 Description: liblustre clients evicted if not contacting servers
5088 Details : Don't put liblustre clients into the ping_evictor list, so
5089 they will not be evicted by the pinger ever.
5091 Severity : enhancement
5093 Description: Add ability to evict clients by NID from MDS.
5094 Details : By echoing "nid:$NID" string into
5095 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5096 $NID would be instantly evicted from this MDS and from all active
5097 OSTs connected to it.
5101 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5102 Details : lookup_it_finish() used to query file size from OSTs that was not
5107 Description: service threads change working directory to that of init
5108 Details : Starting lustre service threads may pin the working directory
5109 of the parent thread, making that filesystem busy. Threads
5110 now change to the working directory of init to avoid this.
5114 Frequency : during shutdown only
5115 Description: shutdown with a failed MDS or OST can cause unmount to hang
5116 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5121 Frequency : 2.6 only
5122 Description: chmod/chown may include an extra supplementary group
5123 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5124 supplementary group and if none is specified this is used.
5127 Bugzilla : 5479 (6816)
5129 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5130 Details : If lookup is in progress on a file that is unlinked we might try
5131 to revalidate the inode and fail in revalidate after lookup is
5132 complete and ll_file_open() enqueues the open again but
5133 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5136 Frequency : always, if lconf --abort_recovery used
5138 Description: lconf --abort_recovery fails with 'Operation not supported'
5139 Details : lconf was attempting to abort recovery on the MDT device and not
5142 ------------------------------------------------------------------------------
5144 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5149 Frequency : rare (only unsupported configurations with a node running as an
5151 Bugzilla : 6514, 5137
5152 Description: Mounting a Lustre file system on a node running as an OST could
5154 Details : OSTs now preallocates memory needed to write out data at
5155 startup, instead of when needed, to avoid having to
5156 allocate memory in possibly low memory situations.
5157 Specifically, if the file system is mounted on on OST,
5158 memory pressure could force it to try to write out data,
5159 which it needed to allocate memory to do. Due to the low
5160 memory, it would be unable to do so and the node would
5161 become unresponsive.
5163 Severity : enhancement
5165 Description: Addition of lconf --service command line option
5166 Details : lconf now accepts a '--service <arg>' option, which is
5167 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5169 Severity : enhancement
5171 Description: Failover mode is now the default for OSTs.
5172 Details : By default, OSTs will now run in failover mode. To return to
5173 the old behaviour, add '--failout' to the lmc line for OSTs.
5175 Severity : enhancement
5177 Description: Health checks are now provided for MDS and OSTs
5178 Details : Additional detailed health check information on MSD and OSTs
5179 is now provided through the procfs health_check value.
5182 Frequency : occasional, depends on IO load
5184 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5185 after numerous create/delete cycles
5186 Details : The ext3 inode allocation policy would not allocate new inodes
5187 very well on the OSTs because there are no new directories
5188 being created. Instead we look for groups with free space if
5189 the parent directories are nearly full.
5194 Description: Network or server problems during mount may cause partially
5195 mounted clients instead of returning an error.
5196 Details : The config llog parsing code may overwrite the error return
5197 code during mount error handling, returning success instead
5203 Description: MDS can fail to allocate large reply buffers
5204 Details : After long uptimes the MDS can fail to allocate large reply
5205 buffers (e.g. zconf client mount config records) due to memory
5206 fragmentation or consumption by the buffer cache. Preallocate
5207 some large reply buffers so that these replies can be sent even
5208 under memory pressure.
5212 Frequency : rare (liblustre)
5213 Description: fsx running with liblustre complained that using truncate() to
5214 extend the file doesn't work. This patch corrects that issue.
5215 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5216 ATTR_SIZE and lsm use in llu_setattr_raw.
5220 Frequency : rare, only 2.6 kernels
5221 Description: Unusual file access patterns on the MDS may result in inode
5222 data being lost in very rare circumstances.
5223 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5224 code in the 2.6 kernel caused the inode and/or EA data not to
5225 be read from disk, causing single-file corruption.
5229 Frequency : rare, only 2.6 filesystems using extents
5230 Description: Heavy concurrent write and delete load may cause data corruption.
5231 Details : It was possible under high-load situations to have an extent
5232 metadata block in the block device cache from a just-unlinked
5233 file overwrite a newly-allocated data block. We now unmap any
5234 metadata buffers that alias just-allocated data blocks.
5238 Frequency : filesystems with default stripe_count larger than 77
5239 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5240 Details : lconf specifies an inode size of 4096 bytes when the default
5241 stripe_count is larger than 77. This conflicts with the default
5242 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5243 this case to avoid pinning too much memory for large EAs.
5245 ------------------------------------------------------------------------------
5247 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5252 Frequency : rare (extremely heavy IO load with hundreds of clients)
5254 Description: Client is evicted, gets IO error writing to file
5255 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5256 caused clients to be evicted for AST timeouts. The fixes in
5257 bug 5192 mean we no longer need to have such short AST timeouts
5258 so ldlm_timeout has been increased.
5261 Frequency : occasional during --force or --failover shutdown under load
5262 Bugzilla : 5949, 4834
5263 Description: Server oops/LBUG if stopped with --force or --failover under load
5264 Details : a collection of import/export refcount and cleanup ordering
5265 issues fixed for safer force cleanup
5268 Frequency : only filesystems larger than 120 OSTs
5269 Bugzilla : 5990, 6223
5270 Description: lfs getstripe would oops on a very large filesystem
5271 Details : lov_getconfig used kfree on vmalloc'd memory
5274 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5275 Bugzilla : 6242, 6243
5276 Description: reading from files that had been truncated to a non-zero size
5277 but never opened returned no data
5278 Details : ll_file_read() reads zeros from no-object files to EOF
5283 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5284 Details : There's little harm in aborting MDS/OSS recovery and letting it
5285 try again, so I removed the LASSERT and return an error instead.
5287 Severity : enhancement
5289 Description: New debugging infrastructure for tracking down data corruption
5290 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5291 (b) cover more of the client-side code path, and (c) try to narrow
5292 down where problems occurred
5296 Bugzilla : 3819, 4364, 4397, 6313
5297 Description: Racing close and eviction MDS could cause assertion in mds_close
5298 Details : It was possible to get multiple mfd references during close and
5299 client eviction, leading to one thread referencing a freed mfd.
5301 Severity: : enhancement
5302 Bugzilla : 3262, 6359
5303 Description: Attempts to reconnect to servers are now more aggressive.
5304 Details : This builds on the enhanced upcall-less recovery that was added
5305 in 1.4.2. When trying to reconnect to servers, clients will
5306 now try each server in the failover group every 10 seconds. By
5307 default, clients would previously try one server every 25 seconds.
5312 Description: After recovery, certain operations trigger a failed
5313 assertion on a client.
5314 Details : Failing over an mds, using lconf -d --failover, while a
5315 client was doing a readdir() call would cause the client to
5316 LBUG after recovery completed and the readdir() was resent.
5318 Severity : enhancement
5320 Description: Default groups are now added by lconf
5321 Details : You can now run lconf --group <servicename> without having to
5322 manually add groups with lmc.
5325 Frequency : occasional
5327 Description: Nodes with an elan id of 0 trigger a failed assertion
5330 Frequency : always when accessing e.g. tty/console device nodes
5332 Description: tty and some other devices nodes cannot be used on lustre
5333 Details : file's private_data field is used by device data and lustre
5334 values in there got lost. New field was added to struct file to
5335 store fs-specific private data.
5338 Frequency : when exporting Lustre via NFS
5340 Description: NFSD failed occasionally when looking up a path component
5341 Details : NFSD is looking up ".." which was broken in ext3 directories
5342 that had grown large enough to become hashed.
5345 Frequency : Clusters with multiple interfaces not on the same subnet
5347 Description: Nodes will repeatedly try to reconnect to an interface which it
5348 cannot reach and report an error to the log.
5349 Details : Extra peer list entries will be created by lconf with some peers
5350 unreachable. lconf now validates the peer before adding it.
5353 Frequency : Only if a default stripe is set on the filesystem root.
5355 Description: Setting a default stripe on the filesystem root prevented the
5356 filesystem from being remounted.
5357 Details : The client was sending extra request flags in the root getattr
5358 request and did not allocate a reply buffer for the dir EA.
5361 Frequency : occasional, higher if lots of files are accessed by one client
5362 Bugzilla : 6159, 6097
5363 Description: Client trips assertion regarding lsm mismatch/magic
5364 Details : While revalidating inodes the VFS looks up inodes with ifind()
5365 and in rare cases can find an inode that is being freed.
5366 The ll_test_inode() code will free the lsm during ifind()
5367 when it finds an existing inode and then the VFS later attaches
5368 this free lsm to a new inode.
5372 Bugzilla : 6422, 7030
5373 Description: MDS deadlock between mkdir and client eviction
5374 Details : Creating a new file via mkdir or mknod (starting a transaction
5375 and getting the ns lock) can deadlock with client eviction
5376 (gets ns lock and trying to finish a synchronous transaction).
5379 Frequency : occasional
5380 Description: While starting a server, the fsfilt_ext3 module could not be
5382 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5383 kernels. Previously, lconf would still use the ext3 name
5384 when trying to load modules. Now, it will correctly use
5385 ext3 on 2.4 and ldiskfs on 2.6.
5387 Severity : enhancement
5388 Description: The default stripe count has been changed to 1
5389 Details : The interpretation of the default stripe count (0, to lfs
5390 or lmc) has been changed to mean striping across a single
5391 OST, rather than all available. For general usage we have
5392 found a stripe count of 1 or 2 works best.
5394 Severity : enhancement
5395 Description: Add support for compiling against Cray portals.
5396 Details : Conditional compiling for some areas that are different
5400 Frequency : occasional
5401 Bugzilla : 6409, 6834
5402 Description: Creating files with an explicit stripe count may lead to
5403 a failed assertion on the MDS
5404 Details : If some OSTs are full or unavailable, creating files may
5405 trigger a failed assertion on the MDS. Now, Lustre will
5406 try to use other servers or return an error to the
5410 Frequency : occasional
5412 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5413 caused lock timeout during readahead (since 1.4.2).
5414 Details : Processes doing readahead might match a lock that hasn't been
5415 granted yet if there are overlapping and conflicting lock
5416 requests. The readahead process waits on ungranted lock
5417 (original lock is CBPENDING), while OST waits for that process
5418 to cancel CBPENDING read lock and eventually evicts client.
5420 Severity : enhancement
5422 Description: Initial enabling of flock support for clients
5423 Details : Implements fcntl advisory locking and file status functions.
5424 This feature is provided as an optional mount flag (default
5425 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5426 locking are implemented yet, and those that are are not guaranteed
5427 to be completely correct in production environments.
5428 mount -t lustre -o [flock|noflock] ...
5431 Frequency : occasional
5433 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5434 assertion in the ext3 JBD (journaling) layer.
5435 Details : The b_committed_data struct is protected by the big kernel lock
5436 in 2.4 kernels, serializing journal_commit_transaction() and
5437 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5438 access to this struct. In 2.6 kernels there is finer grained
5439 locking to improve SMP performance of the JBD layer.
5443 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5448 Description: merge and cleanup kernel patches.
5451 -----------------------------------------------------------------------------
5453 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5455 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5456 versions of Lustre. Please update all servers and clients to
5457 version 1.4.2 or later at the same time. You must also run
5458 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5459 to update the configuration logs.
5461 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5462 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5463 - swab llog records of type '0' so we get proper header size/idx (5861)
5464 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5465 - fix rename of one directory over another leaking an inode (5953)
5466 - avoid SetPageDirty on 2.6 (5981)
5467 - don't re-add just-being-destroyed locks to the waiting list (5653)
5468 - when creating new directories, inherit the parent's custom
5469 striping settings if present parent (3048)
5470 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5471 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5472 - don't hold client locks on temporary worklist from l_lru (5666)
5473 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5474 - automatically evict dead clients (5921)
5475 - Update file size properly in create+truncate+fstat case (6196)
5476 - Do not unhash mountpoint dentries, do not allow removal of
5478 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5479 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5480 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5481 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5482 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5483 - Correctly handle reads of files with no objects (6243)
5484 - lctl recover will also mark a device active if deactivate used (5933)
5486 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5487 - allow --write-conf on an MDS with different nettype than client (5619)
5488 - don't write config llogs to MDS for mounts not from that MDS (5617)
5489 - lconf should create multiple TCP connections from a client (5201)
5490 - init scripts are now turned off by default; run chkconfig --on
5491 lustre and chkconfig --on lustrefs to use them
5492 - upcalls are no longer needed for clients to recover to failover
5494 - add --abort-recovery option to lconf to abort recovery on device
5496 - add support for an arbitrary number of OSTs (3026)
5497 - Quota support protocol changes.
5498 - forward compatibility changes to wire structs (6007)
5499 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5500 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5501 - improved "lustre status" script
5502 - initialize blocksize for non-regular files (6062)
5503 - added --disable-server and --disable-client configure options (5782)
5504 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5505 - Vanilla 2.4.29 support
5506 - increase maximum number of obd devices to 520 (6242)
5507 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5508 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5509 SLES 9 2.6.5 kernels
5510 - limit stripes per file to 160 (the maximum EA size) (6093)
5512 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5515 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5516 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5517 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5518 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5519 - small ext3 extents cleanups and fixes (5733)
5520 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5521 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5522 - avoid needless client->OST connect, fix handle mismatch (5317)
5523 - fix DLM error path that led to out-of-sync client, long delays (5779)
5524 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5525 - fix several locking issues related to i_size (5492,5624,5654,5672)
5526 - don't move pending lock onto export if it is already evicted (5683)
5527 - fix kernel oops when creating .foo in unlinked directory (5548)
5528 - fix deadlock in obdfilter statistics vs. object create (5811)
5529 - use time_{before,after} to avoid timer jiffies wrap (5882)
5530 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5531 - Do not leak request if server was not able to process it (5154)
5532 - If mds_open unable to find parent dir, make that negative lookup(5154)
5533 - don't create new directories with extent-mapping (5909, 5936)
5535 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5536 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5537 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5538 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5539 - support for RHEL4 kernel on i686 (5773)
5540 - provide error messages when incompatible logs are encountered (5898)
5542 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5543 * version 1.4.0.10 (1.4.1 release candidate 1)
5545 - don't keep a lock reference when lock is not granted (4238)
5546 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5547 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5548 - drop import inflight refcount on signal_completed_replay error (5255)
5549 - unlock page after async write error during send (3677)
5550 - handle missing objects in filter_preprw_read properly (5265)
5551 - no transno return for symlink open, don't save no-trasno open (3440)
5552 - don't try to complete elan receive that already failed (4012)
5553 - free RPC server reply state on error (5406)
5554 - clean up thread from ptlrpc_start_thread() on error (5160)
5555 - readahead could read extra page into cache that wasn't ejected (5388)
5556 - prevent races in class_attach/setup/cleanup/detach (5260)
5557 - don't dereference de->d_inode after l_dput of de (5458)
5558 - use "int" for stripe value returned from lock_to_stripe (5544)
5559 - mballoc allocation and error-checking fixes in 2.6 (5504)
5560 - block device patches to fix I/O request sizes in 2.6 (5482)
5561 - look up hostnames for IB nals (5602)
5562 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5563 - don't start multiple acceptors for the same port (5277)
5564 - fix incorrect LASSERT in mds_getattr_name (5635)
5565 - export a proc file for general "ping" checking (5628)
5566 - fix "lfs check" to not block when the MDS is down (5628)
5568 - service request history (4965)
5569 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5570 - create an "evict_client" /proc entry on OSTs, like the MDS has
5571 - fix mount usage message, return errors per mount(8) (5168)
5572 - change grep [] to grep "[]" in tests so they work in more UMLs
5573 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5574 - remove ancient LOV_MAGIC_V0 EA support (5047)
5575 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5576 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5577 - allow readahead allocations to fail when low on memory (5383)
5578 - mmap locking landed again, after considerable improvement (2828)
5579 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5581 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5584 - send OST transaction number in read/write reply to free req (4966)
5585 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5586 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5587 - the watchdog thread now runs as interruptible (5246)
5588 - flock/lockf fixes (but it's still disabled, pending 5135)
5589 - don't use EXT3 constants in llite code (5094)
5590 - memory shortage at startup could cause assertion (5176)
5592 - reorganization of lov code
5593 - single portals codebase
5595 - add extents/mballoc support (5025)
5596 - direct I/O reads in the obdfilter (4048)
5597 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5599 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5602 - send OST transaction number in read/write reply to free req (4966)
5603 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5604 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5605 - the watchdog thread now runs as interruptible (5246)
5606 - handle missing objects in filter_preprw_read properly (5265)
5607 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5608 - drop import inflight refcount on signal_completed_replay error (5255)
5609 - unlock page after async write error during send (3677)
5610 - return original error code on reconstructed replies (3761)
5611 - no transno return for symlink open, don't save no-trasno open (3440)
5613 - add pid to ldlm debugging output (4922)
5614 - bump the watchdog timeouts -- we can't handle 30sec yet
5615 - extra debugging for orphan dentry/inode bug (5259)
5617 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5620 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5621 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5622 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5623 - fix NULL dereference in /proc/sys/portals/routes (4827)
5624 - allow failed mdc_close() operations to be interrupted (4561)
5625 - stop precreate on OST before MDS would time out on it (4778)
5626 - don't send partial-page writes before EOF from client (4410)
5627 - discard client grant for sub-page writes on large-page clients (4520)
5628 - don't free dentries not owned by NFS code, check generation (4806)
5629 - fix lsm leak if mds_create_objects() fails (4801)
5630 - limit debug_daemon file size, always print CERROR messages (4789)
5631 - use transno after validating reply (3892)
5632 - process timed out requests if import state changes (3754)
5633 - update mtime on OST during writes, return in glimpse (4829)
5634 - add mkfsoptions to LDAP (4679)
5635 - use ->max_readahead method instead of zapping global ra (5039)
5636 - don't interrupt __l_wait_event() during strace
5638 - add software watchdogs to catch hung threads quickly (4941)
5639 - make lustrefs init script start after nfs is mounted
5640 - fix CWARN/ERROR duplication (4930)
5641 - return async write errors to application if possible (2248)
5642 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5643 - print NAL number in %x format (4645)
5644 - update barely-supported suse-2.4.21-171 series (4842)
5645 - support for sles 9 %post scripts
5646 - support for building 2.6 kernel-source packages
5647 - support for sles km_* packages
5649 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5652 - ignore -ENOENT errors in osc_destroy (3639)
5653 - notify osc create thread that OSC is being cleaned up (4600)
5654 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5655 - reconstruct ost_handle() like mds_handle() (4657)
5656 - create a new thread to do import eviction to avoid deadlock (3969)
5657 - let lconf resolve symlinked-to devices (4629)
5658 - don't unlink "objects" from directory with default EA (4554)
5659 - hold socknal file ref over connect in case target is down (4394)
5660 - allow more than 32000 subdirectories in a single directory (3244)
5661 - fix blocks count for O_DIRECT writes (3751)
5662 - OST returns ENOSPC from object create when no space left (4539)
5663 - don't send truncate RPC if file size isn't changing (4410)
5664 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5665 - bind to privileged port in socknal and tcpnal (3689)
5667 - rate limit CERROR/CWARN console message to avoid overload (4519)
5668 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5669 - basic mmap support (3918)
5670 - kernel patch series update from b1_4 (4711)
5672 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5675 - avoid crash during MDS cleanup with OST shut down (2775)
5676 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5677 - don't use bad inodes on the MDS (3744)
5678 - dynamic object preallocation to improve recovery speed (4236)
5679 - don't hold spinlock over lock dumping or change debug flags (4401)
5680 - don't zero obd_dev when it is force cleaned (3651)
5681 - print grants to console if they go negative (4431)
5682 - "lctl deactivate" will stop automatic recovery attempts (3406)
5683 - look for existing locks in ldlm_handle_enqueue() (3764)
5684 - don't resolve lock handle twice in recovery avoiding race (4401)
5685 - revalidate should check working dir is a directory (4134)
5687 - don't always mark "slow" obdfilter messages as errors (4418)
5689 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5692 - don't close LustreDB during write_conf until it is done (3860)
5693 - fix typo in lconf for_each_profile (3821)
5694 - allow dumping logs from multiple threads at one time (3820)
5695 - don't allow multiple threads in OSC recovery (3812)
5696 - fix debug_size parameters (3864)
5697 - fix mds_postrecov to initialize import for llog ctxt (3121)
5698 - replace config semaphore with spinlock (3306)
5699 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5700 - don't allow enqueue to complete on a destroyed export (3822)
5701 - down write_lock before checking llog header bitmap (3825)
5702 - recover from lock replay timeout (3764)
5703 - up llog sem before sending rpc (3652)
5704 - reduce ns lock hold times when setting kms (3267)
5705 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5706 - fix NULL deref and obd_dev leak on setup error (3312)
5707 - replace some LBUG about llog ops with error handling (3841)
5708 - don't match INVALID dentries from d_lookup and spin (3784)
5709 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5710 - fix invalid assertion in ptlrpc_set_wait (3880)
5712 - add libwrap support for the TCP acceptor (3996)
5713 - add /proc/sys/portals/routes for non-root route listing (3994)
5714 - allow setting MDS UUID in .xml (2580)
5715 - print the stack of a process that LBUGs (4228)
5717 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5720 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5721 - make sure to unset replay flag from failed open requests (3440)
5722 - if default stripe count is 0, use OST count for inode size (3636)
5723 - update parent mtime/ctime on client for create/unlink (2611)
5724 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5725 - free recovery state on server during a forced cleanup (3571)
5726 - unregister_reply for resent reqs (3063)
5727 - loop back devices mounting and status check on 2.6 (3563)
5728 - fix resource-creation race that can provoke i_size == 0 (3513)
5729 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5730 - more debugging for page-accounting assertion (3746)
5731 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5732 - avoid deadlock after precreation failure (3758)
5733 - fix race and lock order deadlock in orphan handling (3450, 3750)
5734 - add validity checks when grabbing inodes from l_ast_data (3599)
5736 - add /proc/.../recovery_status to obdfilter (3428)
5737 - lightweight CDEBUG infrastructure, debug daemon (3668)
5738 - change default OSC RPC parameters to be better on small clusters
5739 - turn off OST read cache for files smaller than 32MB
5740 - install man pages and include them in rpms (3100)
5741 - add new init script for (un)mounting lustre filesystems (2593)
5742 - run chkconfig in %post for init scripts (3701)
5743 - drop scimac NAL (unmaintained)
5745 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5748 - clean kiobufs before and after use (3485)
5749 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5750 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5751 - change init script to fail more clearly if not run as root (1528)
5752 - allow clients to reconnect during replay (1742)
5753 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5754 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5755 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5756 - fix cond_resched (3554)
5757 - only evict unfinished clients after recovery (3515)
5758 - allow bulk resend, prevent data loss (3570)
5759 - dynamic ptlrpc request buffer allocation (2102)
5760 - don't allow unlinking open directory if it isn't empty (2904)
5761 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5762 - remove extraneous obd dereference causing LASSERT failure (3334)
5763 - don't use get_cycles() when creating temp. files on the mds (3156)
5764 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5765 - handle EEXIST for set-stripe, set proper directory name (3336)
5767 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5768 - fix ksocknal_fmb_callback() error messages (2918)
5770 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5773 - don't copy lvb into (possibly NULL) reply on error (2983)
5774 - don't deref dentry after dput, don't free lvb on error (2922)
5775 - use the kms to determine writeback rpc length (2947)
5776 - increment oti_logcookies when osc is inactive (2948)
5777 - update client's i_blocks count via lvb messages (2543)
5778 - handle intent open/close of special files properly (1557)
5779 - mount MDS with errors=remount-ro, like obdfilter (2009)
5780 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5781 - don't use cancelling-locks' kms values (2947)
5782 - use highest lock extent for kms, not last one (2925)
5783 - don't dereference ERR_PTR() dentry in error handling path (3107)
5784 - fix thread race in portals_debug_dumplog() (3122)
5785 - create lprocfs device entries at setup instead of at attach (1519)
5786 - common AST error handler, don't evict client on completion race (3145)
5787 - zero nameidata in detach_mnt in 2.6 (3118)
5788 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5789 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5790 - initialize RPC timeout timer earlier for 2.6 (3219)
5791 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5792 - print nal/nid for unknown nid (3258)
5793 - additional checks for oscc recovery before doing precreate (3284)
5794 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5795 - don't crash in mdc_close for bad permissions on open (3285)
5796 - zero i_rdev for non-device files (3147)
5797 - clear page->private before handing to FS, better assertion (3119)
5798 - tune the read pipeline (3236)
5799 - fix incorrect decref of invalidated dentry (2350)
5800 - provide read-ahead stats and refine rpc in flight stats (3328)
5801 - don't hold journal transaction open across create RPC (3313)
5802 - update atime on MDS at close time (3265)
5803 - close LDAP connection when recovering to avoid server load (3315)
5804 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5805 - don't leak open file on MDS after open resend (3325)
5806 - serialize filter_precreate and filter_destroy_precreated (3329)
5807 - loop device shouldn't call sync_dev() for nul device (3092)
5808 - clear page cache after eviction (2766)
5809 - resynchronize MDS->OST in background (2824)
5810 - refuse to mount the same filesystem twice on same mountpoint (3394)
5811 - allow llmount to create routes for mounting behind routers (3320)
5812 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5813 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5814 - fix rare problem with rename on htree directories (3417)
5816 - allow default OST striping configuration per directory (1414)
5817 - fix compilation for qswnal for 2.6 kernels (3125)
5818 - increase maximum number of MDS request buffers for large systems
5819 - change liblustreapi to be useful for external progs like lfsck (3098)
5820 - increase local configuration timeout for slow disks (3353)
5821 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5823 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5826 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5827 - don't overwrite extent policy data in reply if lock was blocked (2901)
5828 - drop filter export grants atomically with removal from device (2663)
5829 - del obd_self_export from work_list in class_disconnect_exports (2908)
5830 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5831 - swab reply message in mdc_close, other PPC fixes (2464)
5832 - fix destroying of named logs (2325)
5833 - overwrite old logs when running lconf --write_conf (2264)
5834 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5835 - fix race in target_handle_connect (2898)
5836 - mds_reint_create() should take same inode create lock (2926)
5837 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5838 - don't close files for self_export to avoid uninitialized obd (2936)
5839 - allow MDS with the same name as client node (2939)
5840 - hold dentry reference for closed log files for unlink (2325)
5841 - reserve space for all logs during transactions (2059)
5842 - don't evict page beyond end of stripe extent (2925)
5843 - don't oops on a deleted current working directory (2399)
5844 - handle hard links to targets without a parent properly (2517)
5845 - don't dereference NULL lock when racing during eviction (2867)
5846 - don't grow lock extents when lots of conflicting locks (2919)
5848 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5851 - account for cache space usage on clients to avoid data loss (974)
5852 - lfsck support in lustre kernel code (2349)
5853 - reduce journal credits needed for BRW writes (2370)
5854 - orphan handling to avoid losing space on client/server crashes
5855 - ptlrpcd can be blocked, stopping ALL progress (2477)
5856 - use lock value blocks to assist in proper KMS, faster stat (1021)
5857 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5858 - recovery for initial connections (2355)
5859 - fixes for mds_cleanup_orphans (1934)
5860 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5861 - block all file creations until orphan recovery completes (1901)
5862 - client remove rq_connection from request struct (2423)
5863 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5864 - recovery timer race (2670)
5865 - mdc_close recovey bug (2532)
5866 - ptlrpc cleanup bug (2710)
5867 - mds timeout on local locks (2588)
5868 - namespace lock held during RPCs (2431)
5869 - handle interrupted sync write properly (2503)
5870 - don't try to handle a message that hasn't been replied to (2699)
5871 - client assert failure during cleanup after abort recovery (2701)
5872 - leak mdc device after failed mount (2712)
5873 - ptlrpc_check_set allows timedout requests to complete (2714)
5874 - wait for inflight reqs when ptlrpcd finishes (2710)
5875 - make sure unregistered services are removed from the srv_list
5876 - reset bulk XID's when resending them (caught by 1138 test)
5877 - unregister_bulk after timeout
5878 - fix lconf error (2694)
5879 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5880 - readahead locks pages, leaves pending causing memory pressure (2673)
5881 - increase OST request buffers to 4096 on large machines (2729)
5882 - fix up permission of existing directories in simple_mkdir (2661)
5883 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5884 - don't assign transno to errored transactions (2742)
5885 - don't delete objects on OST if given a bogus objid from MDS (2751)
5886 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5887 - if rq_no_resend, then timeout request after recovery (2432)
5888 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5889 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5890 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5891 - MDS recovery shouldn't delete objects that it has given out (2730)
5892 - if enqueue arrives after completion, don't clobber LVB (2819)
5893 - don't unlock pages twice when trigger_group_io returns error (2814)
5894 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5895 - don't write pages to disk if there was an error (1450)
5896 - don't ping imports that have recovery disabled (2676)
5897 - take buffered bytes into account when balancing socknal conn (2817)
5898 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5899 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5900 - remove little-endian swabbing of llog records (1987)
5901 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5902 - retry reposting request buffers if they fail (1191)
5903 - grow extent at grant time to avoid granting a revoked lock (2809)
5904 - lock revoke doesn't evict page if covered by a second lock (2765)
5905 - disable VM readahead to avoid reading outside lock extents (2805)
5907 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5908 - updated kernel patches for hp-2.4.20 kernel (2681)
5910 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5913 - fix truncated write corruption (2366)
5914 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5916 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5917 - ensure len(uuid) < 37 in lmc (1171)
5918 - fix ia64 OOPS in llog_test (2255)
5919 - zero end of page at obdfilter for partial page writes (2648)
5920 - don't leave stale dentries around after renames (bug 2428)
5921 - fix timeouts when evicting a client with a single lock held (2642)
5922 - set deadline for the initial HELLO message to drain (2634)
5923 - print out dotted-quad IP addresses in the socknal (2302)
5925 - additional debugging for MDS client eviction problem (2443)
5926 - fix mkfsoptions support for osts (2603, 2604)
5928 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5931 - add series for the vanilla 2.6.0 kernel
5932 - add series for the vanilla 2.4.24 kernel
5933 - add series for a cray x86/64 UL kernel drop
5934 - fix xattr patches for the vanilla 2.4.19 series
5936 - generate true UUIDs in lmc (1171)
5937 - have portals stack dumping break in UML (2466)
5938 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5939 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5940 - wake callback waiting threads on client eviction (2460)
5941 - Add --ptldebug and --subsystem to lmc (1719)
5942 - update assertion to allow safe interrupt allocation
5943 - set rq_no_resend for cancel requests (2432)
5944 - recalculate ptlrpcd timeout after resend (2494)
5945 - call vfs_rmdir when removing pending directories (2368)
5946 - fix renaming a file to itself (2429)
5947 - lmc creates a default one-stripe lov (2454)
5948 - expand procfs space to handle large clusters (2326)
5949 - increase UML stack to avoid overflow
5950 - update lconf's list of debug and subsystem masks
5951 - fix lfs find --obd (2510)
5952 - /proc tunable for disabling filter read caching (2591)
5953 - stop rpm packages from altering slapd.conf (2301)
5954 - disable nagle in the socknal under 0conf (2578)
5955 - choose mds inode size based on stripe count (2572)
5956 - fix kernel-source rpm problems (2516)
5958 - add --disable-doc to avoid pdf generation (2421)
5959 - update documentation, tests, type-os, comments
5960 - avoid format warnings on ia64
5961 - remove the TOE NAL
5962 - tiny code cleanups by removing unused fields
5964 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5967 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5968 - remove the most blatant lies from BUILDING file (2371)
5969 - change default debug level to reasonable production setting
5970 - reduce client side cache size to reduce cache flush time
5971 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5972 - make TCP zerocopy and pinger support enabled by default (2476)
5973 - sync writes completed after process exits caused crashes (2319)
5974 - maintain correct mount count on the MDS (2356)
5975 - backout 1557, because 2316 wasn't really fixed
5976 - better file I/O statistics gathering in /proc
5977 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5978 - another kernel patch to fix zero-copy TCP function export
5979 - don't take duplicate lock when processing re-sent getattr (2420)
5980 - lctl uses obd_self_export instead of creating new conn (2353)
5981 - MDS/OST recovery case which requires object creation asserted (2425)
5982 - move lfs from /usr/sbin to /usr/bin in packages
5983 - fix race between mds_client_add and mds_client_free (2417)
5984 - use kmalloc instead of slabs in portals (2430)
5985 - don't create duplicate records when a failover MDS is present (2442)
5986 - remove unnecessary mount age check (2332)
5987 - don't remove directory inodes from locks prematurely (2451)
5988 - don't break if MDS service name is the same as hostname (2103)
5989 - fix races in client write RPC generation when cache full (2482)
5991 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5994 - remove now-unused request->rq_obd (278)
5995 - if an allocation fails, print out how much memory we've used (1933)
5996 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5997 - add the "configurable stack size" patch to most series files (1256)
5998 - ability to write large log records, for 100+ OST configs (2306)
5999 - fix NULL deref when filter_prep fails (2314)
6000 - fix operator precedence error in filter_sync
6001 - dynamic allocation of socknal TX descriptors (2315)
6002 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6003 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6004 - ensure that CWARN messages in Portals always get to the syslog
6005 - __init/__exit are not for prototype decls (ldlm_init/exit)
6006 - x86-64 compile warning fixes
6007 - fix gateway LMC keyword conflict (2318)
6008 - fix MDS lock inversions in getattr/reint paths (1844)
6009 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6010 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6011 - DLM race condition prevented some lock evictions (2328)
6012 - ENOMEM detection and retry on socknal sends (2230)
6013 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6014 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6015 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6016 - better error messages when a client is rejected during recovery (1505)
6017 - avoid cancelling locks which were never granted, after failure (2330)
6018 - fix i_sem/journal inversion in mds_client_add (2333)
6019 - fix truncate/getattr lock cycle deadlock (2334)
6020 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6021 - fix two rare exit paths which could leak an l_lock() ref (2321)
6022 - fencepost error in MDS/OST orphan recovery (2226)
6023 - make log record alignment 8 bytes (1988)
6024 - lstripe now fails when requested offset > ost_count (2237)
6025 - ensure that all kernel series have a complete list.h (1607)
6026 - fix crashes in special-file operations (2316)
6027 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6029 - allow configurable automake binary, for testing new versions
6030 - small update to the lfs documentation
6032 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6034 * fix negative export reference count in fsfilt_sync (2312)
6036 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6037 * release candidate 0.9.1
6039 - orphans are moved into the PENDING directory for possible recovery
6040 - replayed opens now open by fid for orphan/rename safety (1042)
6041 - last close of an orphan inode generates a transno (683)
6042 - chdir() and mount() now pin the directory entry (1020)
6043 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6044 - discard very old requests without processing them (1502)
6045 - remove some common, well-understood CERRORs (1505)
6046 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6047 - clear "grant" flags in OST replies until OST grant code lands (1644)
6048 - fix read performance by not clobbering i_blksize on client (1598)
6049 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6050 - make LRU size a /proc tunable, clears locks when reduced (707)
6051 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6052 - server locks take references on exports now (1558)
6053 - build fixes for 2.4.20-rh trees (1663)
6054 - return an error from lov_create if all OSCs are inactive (1751)
6055 - fix import levels when a reconnect happens without a timeout (1597)
6056 - exit early from mds_open if we get a lookup error (1749)
6057 - partial page read at EOF wouldn't wait for disk before sending (1642)
6058 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6059 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6060 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6061 - write/truncate lock inversion (1639)
6062 - Don't auto-load obdclass, portals modules during cleanup (1495)
6063 - fix timestamps from jumping to "now" (1763)
6064 - extra journal assertions (1648)
6065 - add an extra multiunlink test (1771)
6066 - fix read_record/write_record API (1776)
6067 - fix leak of offset_extent, possible incorrect i_size later (1772)
6068 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6069 - Debugging for the kqswnal_get_idle_tx problems (1820)
6070 - Allow recovery to be attempted multiple times (1536)
6071 - Write out MDS last_rcvd file after it is first created (1600)
6072 - Fix tx_descriptor leak in failed transmit situations (1827)
6073 - ext3 journaling fixes for assertion failure after IO error (1871)
6074 - class_export_put() on freed export after completion AST error (1896)
6075 - Fix revalidate looping in VFS (1322)
6076 - Don't access a freed export during MDS_REINT timeout (1521)
6077 - Add open-unlink recovery support on the MDS (1673,1764)
6078 - Return an error if no MDS data was read from last_rcvd (1946)
6079 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6080 - Don't setup a disk device that doesn't match exported UUID (317)
6081 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6082 - avoid committing NULL handle in force close
6083 - local.sh is now a one-stripe LOV configuration
6084 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6085 - fix ext3 htree duplicate directory entry corruption (1516)
6086 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6087 - update to new LOV EA format (2097)
6088 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6090 2003-06-15 Phil Schwan <phil@clusterfs.com>
6093 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6094 - per-import/export recovery handling (958, 931, 959)
6095 - multiple last-rcvd slots, for serving multiple FSes (949)
6096 - connections are again shared between multiple imp/exports (963, 964)
6097 - "umount -f" would hang if any requests needed to be sent (393, 978)
6098 - avoid pinning large req buffer by copying for queued messages (989)
6099 - add "uuid" to "lctl device" command to help upcalls (991)
6100 - "open" RPCs with transnos would confuse recovery counters (1037)
6101 - do proper endian conversion of all wire messages (288, 340, 891)
6102 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6103 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6104 - fix ptlrpc_connection leak in target_handle_connect (1174)
6105 - fix import refcounting bug in OST and MDS cleanup (1134)
6106 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6107 - fix very unlikely obd_types race condition (501)
6108 - remove osc_open hack for echo_client (1187)
6109 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6110 - a failure in read_inode2 leads to deadlock (1139)
6111 - cancel ack-locks as soon as transaction is committed (1072)
6112 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6113 - make sure to commitrw after any preprw to avoid deadlock (1162)
6114 - failing to execute a file in a lustre FS would lock inode (1203)
6115 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6116 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6117 - obd_uuid2tgt would walk off the end of the list (1255)
6118 - on IA64 the osc would give portals incorrect bulk size (1258)
6119 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6120 - fix lock inversion caused by new llite matching code (1282)
6121 - limit the number of dirty pages on a client to 10MB (1286)
6122 - timed out locks were not being corrected cancelled (1289)
6123 - fix O_DIRECT above 4GB on IA-32 (1292)
6124 * major user-visible changes
6125 - fail out/fail over policy now controlled by the upcall (993)
6127 - add OBD_PING to check server availability and failure (954)
6128 - lustre messages are now sent in sending host order (288, 340, 891)
6129 - add eadatalen to MDS getattr reply (340)
6130 - OST read replies may contain second buffer, with per-page status (593)
6132 2003-03-11 Phil Schwan <phil@clusterfs.com>
6135 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6136 - failed open()s could cause deadlock; fixed (867, 869)
6137 - stop cancelling OST locks when files are closed (481)
6138 - overlapping XID spaces caused network corruption (851, 853)
6139 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6140 - setattr_raw added, to do single-RPC, server-side setattrs
6141 - lmc/lconf syntax change for OST UUIDs
6142 - fix crashy race condition between ptlrpc_free_req and osc_close
6143 - don't use request in mdc_enqueue if we hit a timeout (889)
6144 - don't set the inode i_size for regular files from the MDS (896)
6145 - handle out of order completion AST (842)
6146 - don't LBUG if a lock request times out after receiving AST (913)
6147 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6148 - if a bad lock AST arrives, send an error instead of dropping entirely
6149 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6150 - fix leak in bulk IO when only partially completed (899, 900, 926)
6151 - fix O_DIRECT for ia64 (55)
6152 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6153 - C-z after timeout could hang a process forever; fixed (977)
6155 - client-side I/O cache (678, 924, 929, 941, 970)
6157 - READPAGE and SETATTRs which don't take server-side locks get
6160 2003-02-11 Phil Schwan <phil@clusterfs.com>
6163 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6164 - Fix fsfilt_extN_readpage() to read a full page of directory
6165 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6166 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6167 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6168 - Most intent operations were removed, in favour of a new RPC mode
6169 that does a single RPC to the server and bypasses most of the VFS
6170 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6171 - Aggressively cancel local locks on DLM servers
6172 - mds_reint_unlink sends EA to the client if it's the last nlink.
6173 client uses that EA to unlink OST objects.
6174 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6175 - recursive symlinks were fixed (439)
6176 - fixed NULL deref in DEBUG_REQ
6177 - filter_update_lastobjid no longer calls sync, which annoyed extN
6178 - fixed multi-client small-writes to a single file problem (445)
6179 - fixed mtime updates during file writes (607)
6180 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6181 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6182 - fixed Linux OST size reporting problem (444, 656)
6183 - OST now updates object mtime with writes or setattr (607, 619)
6184 - client verifies file size before zeroing page past EOF (445)
6185 - OST now writes last allocated objid to disk with allocation (108)
6186 - LOV on echo now works (409)
6188 - mds_reint_unlink sends a new buffer, with the EA included. this
6189 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6190 set if a regular file was being unlinked, and it was the last link
6191 - use PtlGet from the target for bulk writes (315)
6192 - OST now updates object mtime with writes or setattr (607, 619)
6193 - LDLM now has a grant-time callback to revalidate locked items, if
6195 - Many MDS operations were reorganized to combat race conditions
6197 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6198 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6200 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6203 - Fully reactivate OST imports after reconnection (512, others)
6204 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6205 - More graceful error handling for truncating on dead OST (515)
6206 - Don't error out unless we're actually accessing dead stripes (474)
6207 - Fix garbage sizes when stripes are missing (410)
6208 - LRU counters were broken, causing constant lock purge (433, 432)
6209 - garbage on read from stripes with failed OSTs (441)
6210 - mark OSCs as active before reconnecting during recovery (438)
6211 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6212 - lfind did not preserve OST order in output (443)
6213 - symlinks cause hung clients, incorrect data (439)
6214 - stop dereferencing request after dropping refcount (457)
6215 - don't LASSERT(spin_is_locked) on non-SMP (455)
6216 - fixes for many rename() bugs
6217 - fstat didn't correctly synchronize attributes (399)
6218 - server must handle lock cancellation during blocking AST prep (487)
6219 - bulk descriptors were free()d too soon (511)
6220 - fix paths in lconf, which would load incorrect modules (451, 507)
6221 - fix confusing lconf 'host not found' error message (386)
6222 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6223 - fix race condition in mdc_blocking_ast() for inode access (526)
6224 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6225 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6226 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6227 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6228 - fix lstripe in several ways (kernel side) (527)
6229 - fix request leak in ldlm_cli_enqueue (262)
6230 - incorrect OSC was marked inactive after OST failure
6231 - call mds_fs_cleanup before unmounting filesystem (524)
6232 - fix races between taking ns_lock and ldlm_lock_change_resource
6233 - fix races updating LOV export open file list
6234 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6235 - fix recovery NULL deref in ldlm_cli_cancel_unused
6236 - fix some DLM races by using new hash table for lock handles (419)
6237 - permit the client to specify desired inodes, at replay
6238 - duplicate requests when we queue them for replay reintegration
6239 - fix last_rcvd offset calculation
6240 - sync after each recovered transaction, so we always make progress
6241 - never, not always, ERESTART requests without transnos
6242 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6244 - skip replay if the MDS says that the client is already connected
6245 - don't check for a recovery-enabled export to match lctl's UUID
6246 - don't INC_USE_COUNT for phantom exports
6247 - don't crash when cleaning up phantom exports (567)
6248 - don't double-finish or set replay data for errored mdc_open requests
6249 - abort requests when they time out, so we don't get old replies
6250 - send/receive replies for AST messages again
6251 - if the client says that it doesn't have the lock, cancel it on the
6253 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6254 mark it as destroyed, it will all work out when decref is called
6255 - fix module use counts (22, 581)
6257 - ASTs now expect a reply (server cancels lock on error reply)
6259 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6262 - fix many simultaneous client startup (392)
6263 - fix dentry->d_it clobbering
6264 - credentials weren't being shipped for readdir/getattr operations
6265 - remove invalid assertions triggered during some concurrent MD
6267 - proper Lustre versions added (336, 389)
6268 - fix memory leak for create error case (398)
6269 - fix LOV locking bug that would get cli/srv out of sync
6270 - fix echo client over LOV (409)
6271 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6272 - fix double-O_EXCL intent crash (424)
6273 - avoid sending multiple lock CANCELs (352)
6275 - MDS can do multi-client recovery (modulo bugs in new code)
6277 - many updates, edits, cleanups
6279 2002-11-18 Phil Schwan <phil@clusterfs.com>
6282 - fix null d_it dereference (346)
6283 - fix full OST/dbench hang (333)
6284 - fix permission problem with file removal (286)
6285 - fix removal of OSCs from LOV when they fail
6286 - fix NULL deref during bulk timeout (214)
6287 - fix problems related to multiple filesystems on one MDS (241)
6288 - fixed serious subtle metadata locking bugs
6289 - free locks on clients when inodes are removed due to memory
6291 - fix inode pointer in lock data (285)
6292 - partial support for multiple MDS on a single host (241)
6293 - data locks weren't cancelled at clear_inode time (290, 311)
6294 - intent locks could lead to unbounded lock growth (205)
6295 - added a maximum lock count, an LRU list, and a flusher
6296 - fix multiple rename (365)
6297 - properly abstracted the echo client
6298 - OSC locked 1 byte too many; fixed
6299 - rewrote brw callback code:
6300 - fixed recovery bugs related to LOVs (306)
6301 - fixed too-many-pages-in-one-write crash (191)
6302 - fixed (again) crash in sync_io_timeout (214)
6303 - probably fixed callback-related race (385)
6305 - Add capability to MDS protocol
6306 - LDLM cancellations and callbacks on different portals
6308 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6311 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6312 - defer open object destruction to close time (601981, 138)
6313 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6314 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6315 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6316 - don't call obd_setattr in ll_file_release for destroyed objects
6317 * protocol change to lustre_msg: move |version| and add |flags|
6318 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6319 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6320 * added replay of create, unlink, link and rename operations during
6321 MDS failover; recovery should be much more robust now
6322 * remove failed OSCs from LOVs (only lov_create uses this so far)
6323 * the lustre-HOWTO was brought (more) up to date (582544)
6325 2002-10-23 Phil Schwan <phil@clusterfs.com>
6328 - in-use dentries weren't being reused properly (617851)
6329 - prevent multiple LDLM setup (599178)
6330 - fix LOV size calculations for truncate (617853)
6331 - fix client handling of MDS intent errors (POSIX)
6332 - fix permission bug in lovstripe.c test (624321)
6333 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6334 - truncate past end of file could corrupt data
6335 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6336 - a race in recovery could return ETIMEDOUT to apps (623947)
6337 - building outside the source directory was fixed
6338 * the lustre-HOWTO was brought (more) up to date (582544)
6339 * major progress was made on recovery functionality
6341 2002-10-10 Phil Schwan <phil@clusterfs.com>
6344 - recovery deadlock fix
6345 - rm -rf causes LBUG fix (617817)
6346 - file open by multiple tasks fix (618962)
6347 - directory permissions bugs (602707 and 620007)
6348 - journal_stop fixed with locking (611313)
6349 - O_APPEND failures resolved (618273, perhaps 614459)
6350 - lconf PATH fix (619770)
6351 - IA64 build fix (621450)
6352 - RPC buffer sizes scale with amount of memory
6354 2002-10-01 Phil Schwan <phil@clusterfs.com>
6357 - locks would be cancelled without throwing away data pages,
6358 resulting in inconsistent data (605627)
6359 - inode attributes were not always being refreshed (605627, 612449)
6360 - lconf now continues to cleanup after lctl reports an error
6361 - MDS now enforces user permissions (602707)
6362 - lprocfs cleanup fixed, but not yet enabled (614157)
6363 - fixed infinite server hang, should a client not respond to an AST
6364 - avoid going into recovery if user calls readlink() with a buffer
6365 that's too small (613941)
6366 - AST RPCs no longer require replies (614867) -- this may be changed
6367 - don't crash server if client sends an IOV that's too big (611336)
6368 - fixed lock conversion deadlock (611892)
6369 - fixed the following of symlinks (614622)
6370 * recovery: the server can remove locks from a client that dies, other
6371 clients can make progress
6372 * more extN patch fixes
6373 * compile-time configurable ptlrpc buffer allocations
6375 - collaborative read cache document
6376 - Lustre Lite Performance CDR document-in-progress
6378 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6381 - fix typo in patch-2.4.18
6383 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6386 - clear ptlrpc request each time in handle_incoming_request()
6387 - unlink of files now destroys the object on the OST
6389 2002-09-19 Peter Braam <braam@clusterfs.com>
6391 * add hard link support
6392 * change obdfile creation method
6393 * kernel patch changed
6395 2002-09-19 Peter Braam <braam@clusterfs.com>
6398 - stack overflow bug in extN fixed
6400 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6402 * documentation updates
6403 - add man pages for config tools
6404 - update tests/README to describe testing with new config tools
6405 - finish metadata API descriptions
6406 * bug fixes and cleanups
6407 - statfs workaround for 16TB limit
6408 - LOV stripe allocation improved, can stripe on subset of OSTs
6409 - LOV file size/IO offset was wrong for files > 4GB in size
6410 - object EA data was being dropped, caused files to be unreadable
6411 - memory overflow with non-LOV OST caused memory corruption
6412 - fixed regression tests to work with new config tools, obdfilter
6413 - fixed bug when directory size became larger than 1 block
6414 - fixed bug (for single client case) when PWD was deleted
6415 - invalidate local directory pages when doing intent-based ops
6416 - avoid LDLM oops when lock callback contained bad data
6418 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6420 * documentation updates
6421 * bug fixes and cleanups
6422 - configuration tools
6425 - 64-bit compile warnings
6426 - 64-bit internal statfs data
6428 * test_brw on persistent OST devices
6430 * lprocfs (disabled)
6432 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6434 * documentation updates
6435 * bug fixes and cleanups
6436 * configuration tools
6438 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6441 * many small fixes to 0_5_4
6442 * io/network handling
6443 * thinkos in MDS operations
6445 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6448 * crucial basic fixes to 0.5.3
6449 * IOR, Iozone work over Elan
6452 2002-08-07 Phil Schwan <phil@clusterfs.com>
6453 * version 0_5_3, our first alpha
6454 * we use the new Portals iovs
6455 * documentation updates
6456 * bug fixes and cleanups
6457 * small changes in the DLM wire protocol
6459 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6460 * version 0_5_1 with some initial stability,
6461 * locking on MD and file I/O.
6462 * documentation updates
6463 * several bug fixes since 0.5.0
6464 * small changes in wire protocol
6466 2002-07-18 Phil Schwan <phil@clusterfs.com>
6468 * delivered as Lustre Light Alpha
6469 * fixed a crash after handling invalid MDS requests
6470 * fixed directory pages for architectures with non-4k pages sizes
6472 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6473 * release version v0_4_4
6474 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6476 * Fixes a number of interruption problems with OST operations.
6477 * Update documentation for portals header changes
6478 * Move all wire protocol structs/defines to lustre_idl.h
6479 * Fixes symlink length bug.
6480 * Add tcpdump to repository.
6482 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6483 * release version v0_4_3
6484 * Fixes statfs for inodes on extN.
6485 * Fixes bug in runtests which would delete /etc/hosts.
6486 * Use 64-bit object IDs wherever possible (not into VFS though)
6487 Remove ost_get_info, which is unused by lustre, and out of date.
6489 2002-07-03 Peter Braam <braam@clusterfs.com>
6490 * release version v0_4_2 Fixes a lookup error (type not passed)
6491 * move forward to head of Portals
6492 * move forward to latest Lustre kernel
6494 2002-06-25 Peter Braam <braam@clusterfs.com>
6495 * release version v0_4_1. Hopefully stable on single node use.