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 2.6.32-504.30.3.el6 (RHEL6.6)
8 2.6.32-573.3.1.el6 (RHEL6.7)
9 3.10.0-229.7.2.el7 (RHEL7.1)
10 3.0.101-0.47.52 (SLES11 SP3)
11 * Client known to build on unpatched kernels:
12 2.6.32-431.29.2.el6 (RHEL6.5)
13 2.6.32-504.30.3.el6 (RHEL6.6)
14 2.6.32-573.3.1.el6 (RHEL6.7)
15 3.10.0-229.7.2.el7 (RHEL7.1)
16 3.0.101-0.47.52 (SLES11 SP3)
18 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
19 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
20 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
21 stack overflow. For more information, please refer to bugzilla 17630.
22 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
23 the re-exporting nodes to be mounted with "32bitapi" mount option
25 --------------------------------------------------------------------------------
27 03-10-2015 Intel Corporation
29 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
30 for currently supported client and server kernel versions.
31 * Server known to build on patched kernels:
32 2.6.32-431.29.2.el6 (RHEL6.5)
33 2.6.32-504.8.1.el6 (RHEL6.6)
34 3.0.101-0.46 (SLES11 SP3)
35 * Client known to build on unpatched kernels:
36 2.6.32-431.29.2.el6 (RHEL6.5)
37 2.6.32-504.8.1.el6 (RHEL6.6)
38 3.10.0-123.20.1.el7 (RHEL7)
39 3.0.101-0.46 (SLES11 SP3)
40 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
41 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
42 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
43 stack overflow. For more information, please refer to bugzilla 17630.
44 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
45 the re-exporting nodes to be mounted with "32bitapi" mount option
47 Severity : enhancement
49 Description: control OST-index in IDIF via ROCOMPAT flag.
50 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
51 last_rcvd file. For new formatted OST device, it will be auto set;
52 for the case of upgrading from old OST device, you can enable it
53 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
54 enabled, for new created OST-object, its IDIF-in-LMA will contain
55 the OST-index; for the existing OST-object, the OSD will convert
56 old format IDIF as new format IDIF with OST-index stored in the
57 LMA EA when accessing such OST-object or via OI scrub. Once such
58 flag is enabled, it cannot be reverted back, so the system cannot
59 be downgraded to the orignal incompatible version.
61 --------------------------------------------------------------------------------
63 07-30-2014 Intel Corporation
65 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
66 for currently supported client and server kernel versions.
67 * Server known to build on patched kernels:
68 2.6.32-431.20.3.el6 (RHEL6.5)
69 3.0.101-0.31 (SLES11 SP3)
70 * Client known to build on unpatched kernels:
71 2.6.32-431.20.3.el6 (RHEL6.5)
72 3.0.101-0.31 (SLES11 SP3)
75 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
76 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
77 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
78 stack overflow. For more information, please refer to bugzilla 17630.
79 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
80 the re-exporting nodes to be mounted with "32bitapi" mount option
82 --------------------------------------------------------------------------------
84 10-31-2013 Intel Corporation
86 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
87 * Server builds for kernels:
88 2.6.32-358.18.1.el6 (RHEL6.4)
89 * Client builds for unpatched kernels:
90 2.6.32-358.18.1.el6 (RHEL6.4)
91 3.0.92-0.8 (SLES11 SP3)
93 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
94 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
95 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
96 stack overflow. For more information, please refer to bugzilla 17630.
97 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
98 the re-exporting nodes to be mounted with "32bitapi" mount option
100 --------------------------------------------------------------------------------
102 05-30-2013 Intel Corporation
104 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
105 * Server builds for kernels:
106 2.6.32-358.6.2.el6 (RHEL6.4)
107 * Client builds for unpatched kernels:
108 2.6.32-358.6.2.el6 (RHEL6.4)
109 3.0.74-0.6.6 (SLES11 SP2)
111 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
112 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
113 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
114 stack overflow. For more information, please refer to bugzilla 17630.
115 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
116 the re-exporting nodes to be mounted with "32bitapi" mount option
118 --------------------------------------------------------------------------------
120 9-30-2012 Whamcloud, Inc.
122 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
123 * Server support for kernels:
124 2.6.32-279.5.1.el6 (RHEL6)
125 * Client support for unpatched kernels:
126 2.6.32-279.5.1.el6 (RHEL6)
128 2.6.18-238.19.1.el5 (RHEL5)
129 2.6.16.60-0.69.1 (SLES10)
130 * Recommended e2fsprogs version: 1.41.90.wc4
131 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
132 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
133 stack overflow. For more information, please refer to bugzilla 17630.
134 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
135 the re-exporting nodes to be mounted with "32bitapi" mount option
137 Severity : enhancement
139 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
141 Severity : enhancement
143 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
145 Severity : enhancement
147 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
149 Severity : enhancement
151 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
153 Severity : enhancement
155 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
157 --------------------------------------------------------------------------------
159 04-30-2012 Whamcloud, Inc.
161 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
162 * Server support for kernels:
163 2.6.32-220.4.2.el6 (RHEL6)
164 * Client support for unpatched kernels:
165 2.6.32-220.4.2.el6 (RHEL6)
167 2.6.18-238.19.1.el5 (RHEL5)
168 2.6.16.60-0.69.1 (SLES10)
169 * Recommended e2fsprogs version: 1.41.90.wc4
170 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
171 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
172 stack overflow. For more information, please refer to bugzilla 17630.
173 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
174 the re-exporting nodes to be mounted with "32bitapi" mount option
176 Severity : enhancement
178 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
180 Severity : enhancement
182 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
184 Severity : enhancement
186 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
188 Severity : enhancement
190 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
192 --------------------------------------------------------------------------------
194 09-30-2011 Whamcloud, Inc.
196 * Server support for kernels:
197 2.6.18-238.19.1.el5 (RHEL5)
198 2.6.32-131.6.1.el6 (RHEL6)
199 * Client support for unpatched kernels:
200 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
201 2.6.18-238.19.1.el5 (RHEL5)
202 2.6.32-131.6.1.el6 (RHEL6)
203 2.6.29.4-167.fc11 (FC11)
204 2.6.16.60-0.69.1 (SLES10)
206 * Recommended e2fsprogs version: 1.41.90.wc3
207 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
208 Lustre filesystem with 4K stack may cause a stack overflow. For more
209 information, please refer to bugzilla 17630.
210 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
211 mounted with "32bitapi" mount option
213 Severity : enhancement
214 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
216 Severity : enhancement
217 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
219 Severity : enhancement
220 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
223 Frequency : only with SLES10
225 Description: Use OFED "KMP" provided by Novell
226 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
227 Lustre is now built against this package. That means you need to
228 install the ofed-kmp package from Novell for the patchless client.
229 Note that the ofed-kmp that Novell ships may not exactly match the
230 kernel version but should still be compatible.
232 Severity : enhancement
234 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
235 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
237 --------------------------------------------------------------------------------
239 2010-07-15 Oracle, Inc.
241 * Server support for kernels:
242 2.6.18-164.11.1.el5 (RHEL 5)
243 2.6.18-164.11.1.0.1.el5 (OEL 5)
244 * Client support for unpatched kernels:
245 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
246 2.6.18-164.11.1.el5 (RHEL 5),
247 2.6.18-164.11.1.0.1.el5 (OEL 5)
248 2.6.16.60-0.42.8 (SLES 10),
250 2.6.29.4-167.fc11 (FC11)
251 * Recommended e2fsprogs version: 1.41.10-sun2
253 Severity : enhancement
255 Description: Distribute lfs_migrate script with lustre
256 Details : The lfs_migrate mentioned in the manual is now included in the
257 lustre tree in lustre/scripts and will distributed in lustre RPMs.
259 Severity : enhancement
261 Description: Failover nid list added to import proc file.
263 Severity : enhancement
265 Description: "som_preview" mount option required to use SOM
266 Details : The clients will need to use som_preview mount option
270 Description: Lustre does not do 1MB IOs to HW RAID
271 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
273 Severity : enhancement
275 Description: lfs quota output is not very convenient for awk/sed-parsing
276 Details : Some positions in lfs quota output table could be empty or
277 non-empty which made it hard to parse it with scripts, now
278 a dash is put instead of space where there is not supposed
283 Description: Update to OFED 1.5.1
284 Details : OFED 1.5.1 will be used in environments that don't use
285 vendor supplied OFED.
289 Description: Remove lnet. prefix from lctl params display
290 Details : Also renamed lnet.memused to lnet_memused to
295 Description: Use conf_param -d to remove permanant settings.
299 Description: Add list_param -R
300 Details : lctl list_param now supports recursive mode.
303 Bugzilla : 21829, 22312
304 Description: Fix llobdstat.
305 Details : 1. Now updating the write_bytes counter
306 2. Port script fixes from 1.8.
310 Description: decrease the usage of memory on clients.
311 Details : 1. On clients, recycle dentries and inodes unused.
312 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
317 Description: llobdstat fix and enhancement
318 Details : add a counter to set a limit to how many samples will be returned
319 fix a wildcard in the path to limit to obdfilter stats only
323 Description: lfs check only allowed for root
324 Details : Code cleanups and added a regresion test of non-root use of
329 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
330 Details : Fix hr->hr_index overload.
334 Description: Fix FIEMAP functionality.
338 Description: FC11 patchless client support added
342 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
343 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
347 Description: New acc-sm script and various test-framework enhancments.
351 Description: Previously landed patch was reverted.
355 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
359 Description: Add ldiskfs maxdirsize mount option.
363 Description: Fix mkfs.lustre support for 16TB LUNs.
367 Description: Print mmp_check_interval
368 Details : print mmp_check_interval and make it possible to abort
369 mount operation in case it takes too long.
373 Description: Add cascading_rw to lustre mpi tests.
377 Description: Store SOM data in EA.
378 Details : Also several SOM related cleanups.
382 Description: Add refcount to md_open_data
386 Description: Use refcount on lov_request_set to prevent memory corruption.
390 Description: Statahead: use parent dir's i_mutex for synchronization.
394 Description: loadgen improvements.
398 Description: Enable peer health detection in ksocklnd and ko2iblnd.
402 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
406 Description: Fix race on obd_devcie:md_stats
410 Description: Fix SLES 11 client builds.
411 Details : Resolved uninitalized pointer gcc warnings.
415 Description: Allow non-root access for "lfs check".
416 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
420 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
422 Severity : enhancement
424 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
425 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
429 Description: Sanity test_27p fails.
430 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
431 Sanity tests 27 were cleaned up and improved.
435 Description: Unify req format on client/servers
436 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
437 buffers and swabbers approach..
442 Description: drop join file stuff.
444 Severity : enhancement
446 Description: Add 2.6.27 and SLES11 (patchless client) support
450 Description: Fix in ptlrpc_expire_one_request() to print the signed time
453 Severity : enhancement
455 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
458 Severity : enhancement
460 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
462 Severity : enhancement
464 Description: correctly handle big reply message.
465 Details : send LNet event if reply is bigger than buffer and adjust this buffer
470 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
472 Severity : enhancement
474 Description: Add support for OEL5.
478 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
480 Severity : enhancement
482 Description: "weak-modules" support
483 Details : Implement "weak-modules" support which enables kernel modules
484 to be used with any kernel that implements the same kABI. In
485 order to achieve this modules are now installed in
486 /lib/modules/$(uname -r)/updates/kernel on all distributions.
490 Description: client could not reconnect to OST because of an active request.
491 Details : The client now retries to reconnect to the same server, if a
492 connect request failed with EBUSY or -EAGAIN. Test case 35b
493 conf-sanity.sh has been created.
497 Description: Don't allow make backward step on assignin osc next id.
498 Details : race between allocation next id and ll_sync thread can be cause
499 of set wrong osc next id and can be kill valid ost objects.
503 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
507 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
511 Description: Error handling in osc_statfs_interpret() has been improved.
512 Details : Check in osc_statfs_interpret() for EBADR.
516 Description: Do not send statfs() requests to OSTs disabled by administrator.
517 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
519 Severity : enhancement
521 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
523 Severity : enhancement
525 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
527 Severity : enhancement
529 Description: Enhancement for lfs(1) command to use numeric uid/gid.
533 Description: Improved cleanup procedure for conf-sanity test_46a.
537 Description: Do not start bulk transfer, if client tried to reconnect.
538 Details : Do not start bulk transfer, if client tried to reconnect,
539 but failed due to active rpcs. Check for exp_abort_active_req
540 has been implemented.
544 Description: Prevent inconsistences between linux and lustre mount structures.
545 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
546 Make the sleep interruptible.
548 Severity : enhancement
550 Description: provide server to client comms path
551 Details : server to client communications path via new LDLM_SET_INFO rpc
555 Description: truncate starts GFP_FS allocation under transaction causing deadlock
556 Details : ldiskfs_truncate calls grab_cache_page which may start page
557 allocation under an open transaction. This may lead to
558 calling prune_icache with consequent lustre reentrance.
562 Description: send correctly lsm on open replay
563 Details : MDS is trust to LSM size on replay open, but client can set wrong size
568 Description: enable client interface failover
569 Details : When a child reconnects from another NID, properly update export
570 nid hash position and ldlm reverse import.
572 Severity : enhancement
574 Description: Communicate OST degraded/readonly state via statfs to MDS
575 Details : Flags in the statfs returned from OSTs indicate whether the
576 OST is in a degraded RAID state, or if the filesystem has
577 turned read-only after a filesystem error is detected.
582 Description: don't panic if EPROTO was hit when reading symlink
583 Details : correctly handling request reference in error cases.
585 Severity : enhancement
587 Description: Add LustreNetLink, a kernel-userspace communcation path.
589 Severity : enhancement
591 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
596 Description: access to llog context before init.
597 Details : move handling CATALOGS file at osc layer and forbid access to llog
601 Frequency : rare, only when client is writing to OST on local node
603 Description: Avoid deadlock for local client writes
604 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
605 memory freeing context. This allows OST threads to set the
606 PF_MEMALLOC flag on task structures in order to allocate memory
607 from reserved pools and complete IO.
608 Use GFP_HIGHUSER for OST allocations for non-local client writes,
609 so that the OST threads generate memory pressure and allow
610 inactive pages to be reclaimed.
612 Severity : enhancement
614 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
619 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
620 Details : move ldlm namespace creation in setup phase to avoid grab
621 _lprocfs_lock with cli_sem held.
625 Description: Temporarily disable grant shrink.
626 Details : Disable the feature for debugging.
629 Frequency : only when formatting test filesystems
631 Description: Unable to run several mkfs.lustre on loop devices at the same time
632 Details : mkfs.lustre returns error 256 on the concurrent loop devices
633 formatting. The solution is to proper handle the error.
635 Severity : enhancement
637 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
639 Severity : enhancement
641 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
646 Description: resolve race between obd_disconnect and class_disconnect_exports
647 Details : if obd_disconnect will be called to already disconnected export he
648 forget release one reference and osc module can't unloaded.
650 Severity : enhancement
652 Description: Allow tuning service thread via /proc
653 Details : For each service a new
654 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
655 created that can be used to set min/max thread counts, and get the
656 current number of running threads.
661 Descriptoin: don't return error if have partially created objects for file.
662 Details : lov_update_create_set uses set->set_success as index for created
663 objects, so if some requests will be failed, they will have hole
664 at end of array and qos_shrink_lsm for allocate correct lsm.
666 Severity : enhancement
668 Description: Update OFED support to 1.4.1
673 Description: Reduce small size read RPC
674 Details : Set read-ahead limit for every file and only do read-ahead when
675 available read-ahead pages are bigger than 1M to avoid small size
678 Severity : enhancement
680 Description: move AT tunable parameters for more consistent usage
681 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
683 Severity : enhancement
685 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
686 Details : allow skip disconnected ost for send statfs request and hide error
692 Description: fix racy locking of mballoc block bitmaps causing BUG
693 Details : The locking of the mballoc buddy bitmap and the in-memory
694 block bitmap was using two different spin locks in some
695 cases. This made it possible to incorrectly access the
696 mballoc bitmap while another process was modifying it,
697 causing a sanity assertion to fail. While no on-disk corruption
698 was reported, there was some risk of this happening.
701 Frequency : rare, on llog test 6
703 Descriptoin: don't allow connect to already connected import
704 Details : allowing connect to already connected import is hide connecting problem.
709 Descriptoin: allow kill process which wait statahead result
710 Details : for some reasons 'ls' can stick in waiting result from statahead,
711 in this case need way for kill this process.
713 Severity : enhancement
715 Description: Add state history info file, enhance import info file
716 Details : Track import connection state changes in a new osc/mdc proc file;
717 add overview-type data to the osc/mdc import proc file.
719 Severity : enhancement
721 Description: MDS create should not wait for statfs RPC while holding DLM lock.
723 Severity : enhancement
725 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
730 Descriptoin: OST grant too much space to client even there are not enough space.
731 Details : Client will shrink its grant cache to OST if there are no write
732 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
733 this grant cache if there are already not enough avaible space
734 (left_space < total_clients * 32M).
737 Frequency : start MDS on uncleanly shutdowned MDS device
739 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
740 Details : stay in waiting mds<>ost recovery finished produce random bugs
741 due race between two ll_sync thread for one lov target. send
742 ACTIVATE event only if connect realy finished and import have
746 Frequency : rare, connect and disconnect target at same time
748 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
749 Details : don't call obd_disconnect under lov_lock. this long time
750 operation and can block ptlrpcd which answer to connect request.
755 Descriptoin: don't lose wakeup for imp_recovery_waitq
756 Details : recover_import_no_retry or invalidate_import and import_close can
757 both sleep on imp_recovery_waitq, but we was send only one wakeup
761 Frequency : always with long access acl
763 Descriptoin: mds can't pack reply with long acl.
764 Details : mds don't control size of acl but they limited by reint/getattr
767 Severity : enhancement
769 Description: Update to SLES10 kernel-2.6.16.60-0.33.
771 Severity : enhancement
773 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
776 Frequency : start MDS on uncleanly shutdowned MDS device
778 Descriptoin: aborting recovery hang on MDS
779 Details : don't throttle destroy RPCs for the MDT.
782 Frequency : on remount
784 Description: external journal device not working after the remount
785 Details : clear dev_rdonly flag for external journal devices in
791 Description: shutdown vs evict race
792 Details : client_disconnect_export vs connect request race.
793 if client will evicted at this time - we start invalidate
794 thread without referece to import and import can be freed
798 Frequency : rare, need acl's on inode.
800 Description: client can't handle ost additional correctly
801 Details : if ost was added after client connected to mds client can have
802 hit lnet_try_match_md ... to big messages to wide striped files.
803 in this case need teach client to handle config events about add
804 lov target and update client max ea size at that event.
806 Severity : enhancement
808 Description: Changelogs
809 Details : Changelogs are a lightweight mechanism to track filesystem
810 metadata and namespace changes. The changelog is recorded
811 permanently on the MDTs, and is periodically "consumed" / purged
812 when records are no longer needed.
814 Severity : enhancement
816 Description: compact fld format with extents
817 Details : Store range of seq rather than every seq in FLD. Seq
818 controller update FLD rather than clients. In Case of CMD, mdt0
819 has FLD, all other metadata server act as non persistent proxy
820 for FLD queries and cache fld entries in fld cache.
825 Description: don't skip ost target if they assigned to file
826 Details : Drop slow OSCs if we can, but not for requested start idx.
827 This means "if OSC is slow and it is not the requested
828 start OST, then it can be skipped, otherwise skip it only
829 if it is inactive/recovering/out-of-space.
833 Description: more cleanup in mds_lov
834 Details : not send LOV EA under replay, we can't know about they size at this
835 time. Don't allow client connect to mds before any ost connected,
836 for avoid problems with LOV EA size and returning EIO to client.
838 Severity : enhancement
840 Description: Interoperability at server side (Disk interoperability)
842 Severity : enhancement
844 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
846 Severity : enhancement
848 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
850 Severity : enhancement
852 Description: New client IO stack (CLIO).
854 Severity : enhancement
856 Description: Commit on sharing. Eliminate inter-client dependencies between
857 uncommitted transactions by doing transaction commits.
858 Thereby clients may recovery independently.
861 Frequency : Create a symlink file with a very long name
863 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
864 Details : If there is no extra space in the request for early cancels,
865 ldlm_req_handles_avail() returns 0 instead of a negative value.
867 Severity : enhancement
869 Description: Add /proc entry for import status
870 Details : The mdc, osc, and mgc import directories now have
871 an import directory that contains useful import data for debugging
874 Severity : enhancement
876 Description: Re-disable certain /proc logging
877 Details : Enable and disable client's offset_stats, extents_stats and
878 extents_stats_per_process stats logging on the fly.
881 Frequency : Only on FC kernels 2.6.22+
883 Description: oops in statahead
884 Details : Do not drop reference count for the dentry from VFS when lookup,
885 VFS will do that by itself.
887 Severity : enhancement
889 Description: Generic /proc file permissions
890 Details : Set /Proc file permissions in a more generic way to enable non-
891 root users operate on some /proc files.
895 Description: Hitting mdc_commit_close() ASSERTION
896 Details : Properly handle request reference release in
897 ll_release_openhandle().
901 Description: quota recovery deadlock during mds failover
902 Details : This patch includes att18982, att18236, att18237 in bz14840.
904 1. fix osts hang when mds does failover with quotaon
905 2. prevent watchdog storm when osts threads wait for the
910 Frequency : only patchless client
911 Description: add workaround for race between add/remove dentry from hash
913 Severity : enhancement
915 Description: Allow OST glimpses to return PW locks
919 Description: LBUG when llog conf file is full
920 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
924 Description: Prevent import from entering FULL state when server in recovery
928 Description: service mount cannot take device name with ":"
929 Details : Only when device name contains ":/" will mount treat it as
935 Description: replace ptlrpcd with the statahead thread to interpret the async
936 statahead RPC callback
940 Frequency : on recovery
941 Description: I/O failures after umount during fail back
942 Details : if client reconnected to restarted server we need join to recovery
943 instead of find server handler is changed and process self
944 eviction with cancel all locks.
946 Severity : enhancement
948 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
950 Severity : enhancement
952 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
954 Severity : enhancement
956 Description: Upcall on Lustre log has been dumped
957 Details : Allow for a user mode script to be called once a Lustre log has
958 been dumped. It passes the filename of the dumped log to the
959 script, the location of the script can be specified via
960 /proc/sys/lnet/debug_log_upcall.
965 Description: avoid idr_remove called for id which is not allocated.
966 Details : Move assigment s_dev for clustered nfs to end of initialization,
967 to avoid problem with error handling.
972 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
973 Details : When connection is reused this not moved from CONN_UNUSED_HASH
974 into CONN_USED_HASH and this prodice warning when put connection
975 again in unused hash.
978 Severity : enhancement
980 Description: File striping can now be set to use an arbitrary pool of OSTs.
982 Severity : enhancement
984 Description: Export bytes_read/bytes_write count on OSC/OST.
988 Description: Early reply size mismatch, MGC loses connection
989 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
990 the connect flags are properly negotiated.
995 Description: quotas are not honored with O_DIRECT
996 Details : all writes with the flag O_DIRECT will use grants which leads to
997 this problem. Now using OBD_BRW_SYNC to guard this.
1001 Description: add quota statistics
1002 Details : 1. sort out quota proc entries and proc code.
1003 2. add quota statistics
1005 Severity : enhancement
1007 Description: enable quota support for HEAD.
1011 Description: Properly propagate oinfo flags from lov to osc for statfs
1012 Details : restore missing copy oi_flags to lov requests.
1014 Severity : enhancement
1016 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1017 : l_getgroups(8), lst(8), routerstat(8)
1018 Details : included man pages for llobdstat(8), llstat(8),
1019 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1021 Severity : enhancement
1023 Description: configure's --enable-quota should check the
1024 : kernel .config for CONFIG_QUOTA
1025 Details : configure is terminated if --enable-quota is passed but
1026 : no quota support is in kernel
1030 Description: Remove portals compatibility
1031 Details : Remove portals compatibility, not interoperable with releases
1036 Description: Resolve device initialization race
1037 Details : Prevent proc handler from accessing devices added to the
1038 obd_devs array but yet be intialized.
1040 Severity : enhancement
1042 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1044 Severity : enhancement
1046 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1051 Description: Using wrong pointer in osc_brw_prep_request
1052 Details : Access to array[-1] can produce panic if kernel compiled with
1053 CONFIG_PAGE_ALLOC enabled
1057 Description: Client runs out of low memory
1058 Details : Consider only lowmem when counting initial number of llap pages
1062 Description: *optional* service tags registration
1063 Details : if the "service tags" package is installed on a Lustre node
1064 When the filesystem is mounted, a local-node service tag will
1065 be created. See http://inventory.sun.com/ for more information
1066 about the Service Tags asset management system.
1070 Description: Kernel BUG tries to release flock
1071 Details : Lustre does not destroy flock lock before last reference goes
1072 away. So always drop flock locks when client is evicted and
1073 perform unlock regardless of successfulness of speaking to MDS.
1077 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1081 Description: sanity test 65a fails if stripecount of -1 is set
1082 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1087 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1088 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1091 Severity : enhancement
1093 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1097 Description: do not process already freed flock
1098 Details : flock can possibly be freed by another thread before it reaches
1099 to ldlm_flock_completion_ast.
1103 Description: LBUG during stress test
1104 Details : Need properly lock accesses the flock deadlock detection list.
1108 Description: oops in page fault handler
1109 Details : kernel page fault handler can return two special 'pages' in error
1110 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1114 Description: timeout with invalidate import.
1115 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1116 should be handled by ptlrpcd. This produce long age waiting and
1117 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1119 Severity : enhancement
1121 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1125 Description: enable MGS and MDT services start separately
1126 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1127 a co-located MGS without starting the MGS, which is a complement
1128 to 'nosvc' mount option.
1132 Frequency : after recovery
1133 Description: precreate to many object's after del orphan.
1134 Details : del orphan st in oscc last_id == next_id and this triger growing
1135 count of precreated objects. Set flag LOW to skip increase count
1136 of precreated objects.
1140 Frequency : rare, on clear nid stats
1141 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1142 Details : when clean nid stats sometimes try destroy live entry,
1143 and this produce panic in free.
1147 Description: Stack overflow during MDS log replay
1148 ease stack pressure by using a thread dealing llog_process.
1152 Description: wait until IO finished before start new when do lock cancel.
1153 Details : VM protocol want old IO finished before start new, in this case
1154 need wait until PG_writeback is cleared until check dirty flag and
1155 call writepages in lock cancel callback.
1157 Severity : enhancement
1159 Description: using special macro for print time and cleanup in includes.
1163 Description: mds_mfd_close() ASSERTION(rc == 0)
1164 Details : In mds_mfd_close(), we need protect inode's writecount change
1165 within its orphan write semaphore to prevent possible races.
1169 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1173 Frequency : rare, on shutdown ost
1174 Description: don't hit live lock with umount ost.
1175 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1176 use shrink_dcache_sb instead.
1180 Description: don't panic with use echo client
1181 Details : echo client pass NULL as client nid pointer and this produce null
1182 pointer dereference.
1186 Description: fix build on ppc32
1187 Details : compile code with -m64 flag produce wrong object file for ppc32.
1191 Description: add message levels for liblustreapi
1195 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1196 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1197 will result in an -EAGAIN and in the exit call path we call
1198 alloc_rr() which will with increasing aggressiveness attempt to
1199 aquire precreated objects on the minimum number of required OSCs.
1203 Description: Use old size assignment to avoid deadlock
1204 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1205 the scheduling while holding a spinlock. We do not need locking
1206 for size in ll_update_inode() because size is only updated from
1207 the MDS for directories or files without objects, so there is no
1208 other place to do the update, and concurrent access to such inodes
1209 are protected by the inode lock.
1213 Description: resolve "_IOWR redefined" build error on SLES10
1217 Description: dump the memory debugging after all modules are unloaded to
1218 suppress false negative in conf_sanity test 39
1220 Severity : enhancement
1222 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1227 Description: fix SLES kernel versioning
1228 Details : the kernel version for our SLES 10 kernel did not include a "-"
1229 before the "smp" at the end. while this was not a problem in
1230 general, it did mean that software trying to use the kernel
1231 version to try to detect a vendor specific kernel would fail.
1232 this was most evident by the OFED build scripts.
1236 Description: Don't update lov_desc members until making sure they are valid
1237 Details : When updating lov_desc members via proc fs, need fix their
1238 validities before doing the real update.
1242 Description: don't put request into delay list while invalidate in flight.
1243 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1244 import in flight. this produce timeout for invalidate and sometimes
1245 can cause stale data.
1249 Frequency : on ppc only
1250 Description: not convert ost objects for directory because it's not exist.
1251 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1255 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1257 Severity : enhancement
1259 Description: Update to support 2.6.22.14 vanilla kernel.
1263 Frequency : rare, on recovery
1264 Description: read procfs can produce deadlock in some situation
1265 Details : Holding lprocfs lock which send rpc can produce block for destroy
1266 obd objects and this also block reconnect with -EALREADY. This
1267 isn't fix all lprocfs bugs - but make it rare.
1269 Severity : enhancement
1271 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1274 Frequency : frequent on X2 node
1276 Description: mdc_set_open_replay_data LBUG
1277 Details : Set replay data for requests that are eligible for replay.
1281 Description: lustre_mgs: operation 101 on unconnected MGS
1282 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1283 MGC, and late on MGC cannot successfully connect to MGS and a lot
1284 of the error messages complaining that MGS is not connected.
1288 Frequency : on network error
1289 Description: panic with double free request if network error
1290 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1291 but it's true only for synchronus enqueue, for async enqueue
1292 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1295 Severity : enhancement
1297 Description: client-side metadata stat-ahead during readdir(directory readahead)
1298 Details : perform client-side metadata stat-ahead when the client detects
1299 readdir and sequential stat of dir entries therein
1302 Frequency : on start mds
1304 Description: Implement get_info(last_id) in obdfilter.
1307 Frequency : occasional
1309 Description: Correctly check stale fid, not start epoch if ost not support SOM
1310 Details : open with flag O_CREATE need set old fid in op_fid3 because
1311 op_fid2 was overwritten with new generated fid, but mds can answer
1312 with one of these two fids and both is not stale. Setattr
1313 incorrectly started an epoch and assume will be called
1314 done_writing, but without SOM done_writing ever being called.
1317 Frequency : rare, depends on device drivers and load
1319 Description: MDS or OSS nodes crash due to stack overflow
1320 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1321 In some cases, in conjunction with device drivers that use a lot
1322 of stack the MDS (or possibly OSS) service threads could overflow
1323 the stack. One change which was identified to consume additional
1324 stack has been reworked to avoid the extra stack usage.
1327 Frequency : occasional
1329 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1330 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1331 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1332 oscc_has_objects and see OSC has no objests and call
1333 oscc_internal_create to resend create request.
1335 Severity : enhancement
1337 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1339 Severity : enhancement
1341 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1344 Frequency : very rare
1346 Description: Fix replay if there is an un-replied request and open
1347 Details : In some cases, older replay request will revert the
1348 mcd->mcd_last_xid on MDS which is used to record the client's
1349 latest sent request.
1351 Severity : enhancement
1353 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1355 Severity : enhancement
1357 Description: Add rhel5 support to HEAD.
1359 Serverity : enhancement
1361 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1366 Description: Don't allow skipping OSTs if index has been specified.
1367 Details : Don't allow skipping OSTs if index has been specified, make
1368 locking in internal create lots better.
1372 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1373 Details : access to bitfield in structure is always rounded to long
1374 and this produce problem with not atomic change any bit.
1378 Description: Lustre make rpms failed.
1379 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1380 builds Lustre rpms from tarball.
1383 Frequency : rare on shutdown ost
1385 Description: If llog cancel was not send before clean_exports phase, this can
1386 produce deadlock in llog code.
1387 Details : If llog thread has last reference to obd and call class_import_put
1388 this produce deadlock because llog_cleanup_commit_master wait when
1389 last llog_commit_thread exited, but this never success because was
1390 called from llog_commit_thread.
1394 Description: allow userland application know is lost one of stripes.
1395 Details : fill lvb_blocks with error code on ost and return it to
1396 application if error flag found.
1400 Description: NULL lov_tgts causing MDS oops
1401 Details : more safe checks for NULL lov_tgts for avoid oops.
1403 Severity : enhancement
1405 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1409 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1410 Details : If an OST is down the MDS will hang indefinitely in
1411 obd_statfs_rqset() waiting for the statfs data. While for
1412 MDS QOS usage of statfs, it should not stuck in waiting.
1414 Severity : enhancement
1416 Description: remote_acl support
1417 Details : Support ACL-based permission check for remote user.
1418 Support setfacl/getfacl for remote user with the utils
1419 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1420 the system "{s,g}etfacl" utils.
1422 Severity : enhancement
1424 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1426 Severity : enhancement
1428 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1432 Description: make mgs_setparam() handle fsname containing dash
1433 Details : fsname containing a dash does not work with lctl conf_param
1435 Severity : enhancement
1437 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1439 Severity : enhancement
1441 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1445 Frequency : rare, at shutdown
1446 Description: access already free / zero obd_namespace.
1447 Details : if client_disconnect_export was called without force flag set,
1448 and exist connect request in flight, this can produce access to
1449 NULL pointer (or already free pointer) when connect_interpret
1450 store ocd flags in obd_namespace.
1454 Frequency : only at startup
1455 Description: not alloc memory with spinlock held.
1456 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1457 if any spinlock held.
1459 Severity : enhancement
1461 Description: make lustre randomly fail allocating memory
1462 Details : Make lustre randomly failed allocating memory for testing purpose.
1464 Severity : enhancement
1466 Description: lost problems with lov objid file
1467 Details : Fixes some scability and access to not inited memory problems
1468 in work with lov objdid file.
1473 Description: lfs find does not continue on file error
1474 Details : Continue other files processing when a file/dir is absent.
1478 Description: Inconsistent usage of lustre_pack_reply()
1479 Details : Standardize the usage of lustre_pack_reply() such that it
1480 always generate a CERROR on failure.
1483 Frequency : occasional
1485 Description: MDS hang or stay in waiting lock
1486 Details : If client receive lock with CBPENDING flag ldlm need send lock
1487 cancel as separate rpc, to avoid situation when cancel request
1488 can't processed due all i/o threads stay in wait lock.
1492 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1496 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1500 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1504 Description: Update to RHEL4 latest kernel.
1506 Severity : enhancement
1508 Description: Build SLES10 patchless client fails
1509 Details : The configure was broken by run ./configure with
1510 --with-linux-obj=.... argument for patchless client. When the
1511 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1512 can't search header adequately. Use absolute path such as
1513 -I($LINUX)/include instead.
1517 Description: interrupt oig_wait produce painc on resend.
1518 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1519 private set, and this produce situation when rq_ptlrpcd_data not
1520 copyed to new allocated request and triggered LBUG on assert
1521 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1522 ptlrpcd set if request is changed and can be safety changed to use
1527 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1528 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1529 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1530 these macros and replace them with their difinitions instead.
1534 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1535 Details : Update which_patch & target file for SLES10 latest kernel.
1538 Description: Remove client patches from SLES 10 kernel.
1539 Details : This causes SLES 10 clients to behave as patchless clients
1540 even on a Lustre-patched (server) kernel.
1542 Severity : enhancement
1544 Description: self-adjustable client's lru lists
1545 Details : use adaptive algorithm for managing client cached locks lru
1546 lists according to current server load, other client's work
1547 pattern, memory activities, etc. Both, server and client
1548 side namespaces provide number of proc tunables for controlling
1551 Severity : enhancement
1553 Description: light-weight GSS support
1554 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1555 and reduce performance overhead.
1557 Severity : enhancement
1559 Description: Linux keyring support
1560 Details : Support using service of Linux keyring for Lustre GSS internal
1561 context refresh/cache mechanism.
1565 Description: Fix errors in lfs documentation
1566 Details : Fixes man pages
1570 Description: don't use GFP_* in generic Lustre code.
1571 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1575 Description: obdclass is limited by single OBD_ALLOC(idarray)
1576 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1580 Description: warnings with build patchless client with vanila 2.6.19 and up
1581 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1582 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1586 Frequency : Only for SLES
1587 Description: sanity_quota fail test_1
1588 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1589 file, which makes TSTID[2] inunique.
1593 Frequency : Only for SLES9
1594 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1595 Details : Update target/ChangeLog/which_patch .
1599 Frequency : Only for test_10 in sanity-quota.sh
1600 Description: a bug in quota test code
1601 Details : checking a test flag in a wrong place
1605 Frequency : Only for test_13 in sanity-quota.sh
1606 Description: a bug in quota test script
1607 Details : shouldn't check overfull allocated quota
1611 Description: jbd statistics
1612 Details : Port older jbd statistics patch for sles10
1614 Severity : enhancement
1616 Description: Add printing inode info into message about error in writepage.
1621 Description: reply_lock_interpret crash due to race with it and lock cancel.
1622 Details : Do not replay locks that are being cancelled. Do not reference
1623 locks by their address during replay, just by their handle.
1627 Frequency : When flocks are used.
1628 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1629 Details : Flock locks might destroy just granted lock if it could be merged
1630 with another existing flock, this is done in completion handler,
1631 so teach ldlm_cli_enquque_fini that this is a valid case for
1637 Description: Oops in read and write path when failing to allocate lock.
1638 Details : Check if lock allocation failed and return error back.
1642 Description: lstripe command fails for valid OST index
1643 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1644 instead of lov->desc.ld_active_tgt_count.
1648 Description: sanity.sh failed test 103
1649 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1654 Description: df doesn't work properly if diskfs blocksize != 4K
1655 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1659 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1660 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1662 Severity : enhancement
1664 Description: Adaptive timeouts
1665 Details : RPC timeouts adapt to changing server load and network
1666 conditions to reduce resend attempts and improve recovery time.
1670 Description: llapi_file_create() does not allow some changes
1671 Details : add llapi_file_open() that allows specifying the mode and
1672 open flags, and also returns an open file handle.
1676 Description: merge and cleanup kernel patches.
1677 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1681 Description: Add journal checksum support.(Kernel part)
1682 Details : The journal checksum feature adds two new flags i.e
1683 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1684 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1685 indicates that the commit block contains the checksum for
1686 the blocks described by the descriptor blocks. Now commit
1687 record can be sent to disk without waiting for descriptor
1688 blocks to be written to disk. This behavior is controlled
1689 using JBD2_FEATURE_ASYNC_COMMIT flag.
1693 Description: OSS needs mutliple precreate threads
1694 Details : Add ability to start more than one create thread per OSS.
1698 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1699 Details : execute lfs setstripe on client
1703 Description: mds_obd_create error creating tmp object
1704 Details : When the user sets quota on root, llog will be affected and can't
1705 create files and write files.
1708 Frequency : Always on ia64 patchless client, and possibly others.
1710 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1711 Details : This allows the patchless client to be loaded on architectures
1712 without this export.
1716 Description: RedHat Update kernel for RHEL5
1717 Details : Add the kernel config file for RHEL5.
1721 Description: RedHat Update kernel for RHEL5
1722 Details : Modify the kernel config file more closer RHEL5.
1726 Description: Build failure against Centos5 (RHEL5)
1727 Details : Define PAGE_SIZE when it isn't present.
1731 Description: test_7 on sanity-quota.sh can't be used on separate machines
1732 Details : add facet to handle it
1736 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1737 Details : with NFS, the anon dentry's parent was set to itself in
1738 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1739 corresponding dentry other than use rec->ur_name.
1741 Severity : enhancement
1743 Description: lfs setstripe enhancement
1744 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1748 Description: enable data checksumming by default
1749 Details : enable checksum by default, allow --disable-checksum
1750 configure option and "-o nochecksum" mount option. Checksums
1751 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1755 Description: lustre support for RHEL5
1756 Details : Add support for RHEL5.
1760 Description: Client eviction due to failover config
1761 Details : after a connection loss, the lustre client should attempt to
1762 reconnect to the last active server first before trying the
1763 other potential connections.
1767 Description: when mds and osts use different quota unit(32bit and 64bit),
1768 quota will be released repeatly.
1769 Details : void sending multiple quota reqs to mds, which will keep the
1770 status between the reqs.
1774 Description: osts not allocated evenly to files
1775 Details : change the condition to increase offset_idx
1778 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1780 Description: Data corruption for OSTs that are formatted larger than 2TB
1782 Details : When generating the bio request for lustre file writes the
1783 sector number would overflow a temporary variable before being
1784 used for the IO. The data reads correctly from Lustre (which
1785 will overflow in a similar manner) but other file data or
1786 filesystem metadata may be corrupted in some cases.
1790 Description: Tune the kernel for good SCSI performance.
1791 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1792 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1797 Description: rewrite ext2-derived code in obdclass/uuid.c
1798 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1799 from scratch preserving functionality.
1803 Description: "lfs find -obd UUID" prints directories
1804 Details : "lfs find -obd UUID" will return all directory names instead
1805 of just file names. It is incorrect because the directories
1806 do not reside on the OSTs.
1810 Description: use i_size_read and i_size_write in 2.6 port
1811 Details : replace inode->i_size access with i_size_read/write()
1814 Frequency : when removing large files
1816 Description: scheduling issue during removal of large Lustre files
1817 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1818 It causes scheduling issues when removing large files (17TB in the
1822 Frequency : only with liblustre clients on XT3
1824 Description: evictions taking too long
1825 Details : allow llrd to evict clients directly on OSTs
1828 Frequency : only on ppc
1830 Description: /proc/fs/lustre/devices broken on ppc
1831 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1832 We should make sure the type of 'index' is loff_t and then cast
1833 explicitly as needed below. Do not assign an explicitly cast
1837 Frequency : only for rhel5
1839 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1840 Details : Modify the target file & which_kernel.
1842 Severity : enhancement
1844 Description: omit set fsid for export NFS
1845 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1846 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1850 Frequency : Always, for kernels after 2.6.16
1851 Description: Fix warning idr_remove called for id=.. which is not allocated.
1852 Details : Last kernels save old s_dev before kill super and not allow
1853 to restore from callback - restore it before call kill_anon_super.
1857 Description: buffer overruns could theoretically occur
1858 Details : llapi_semantic_traverse() modifies the "path" argument by
1859 appending values to the end of the origin string, and a
1860 overrun may occur. Adding buffer overrun check in liblustreapi.
1864 Description: Fix error on 'ls .' at the top of the Lustre mount.
1865 Details : Don't revalidate dentry if it is a root dentry.
1869 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1870 Details : Modify vm-tunables-rhel4.patch.
1874 Description: Kernel config for 2.6.18-vanilla.
1875 Details : Modify targets/2.6-vanilla.target.in.
1876 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1877 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1878 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1879 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1883 Description: improve handling recoverable errors
1884 Details : if request processig with error which can be recoverable on server
1885 request should be resend, otherwise page released from cache and
1890 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1891 Details : Modify target file & which_patch.
1892 A flaw was found in the IA32 system call emulation provided
1893 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1894 value could be stored in the %RAX register, which could trigger an
1895 out-of-bounds system call table access. An untrusted local user
1896 could exploit this flaw to run code in the kernel
1897 (ie a root privilege escalation). (CVE-2007-4573).
1901 Description: O_DIRECT bypasses client statistics.
1902 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1903 read_ahead_stats were updated. copied the stats section from
1904 osc_send_oap_rpc() into async_internal().
1908 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1912 Description: change order of libsysio includes
1913 Details : '#include sysio.h' should always come before '#include xtio.h'
1917 Description: handle "serious error: objid * already exists" more gracefully
1918 Details : If LAST_ID value on disk is smaller than the objects existing in
1919 the O/0/d* directories, it indicates disk corruption and causes an
1920 LBUG(). If the object is 0-length, then we should use the existing
1921 object. This will help to avoid a full fsck in most cases.
1923 Severity : enhancement
1925 Description: adapt the lustre_config script to support the upgrade case
1926 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1927 server targets to 1.6 in parallel.
1931 Description: To avoid grant space > avaible space when the disk is almost
1932 full. Without this patch you might see the error "grant XXXX >
1933 available" or some LBUG about grant, when the disk is almost
1935 Details : In filter_check_grant, for non_grant cache write, we should
1936 check the left space by if (*left > ungranted + bytes), instead
1937 of (*left > ungranted), because only we are sure the left space
1938 is enough for another "bytes", then the ungrant space should be
1939 increase. In client, we should update cl_avail_grant only there
1940 is OBD_MD_FLGRANT in the reply.
1944 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1945 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1946 described in CVE-2007-4573. This problem could allow untrusted
1947 local users to gain root access.
1950 Frequency : when using O_DIRECT and quotas
1952 Description: Incorrect file ownership on O_DIRECT output files
1953 Details : block usage reported by 'lfs quota' does not take into account
1954 files that have been written with O_DIRECT.
1959 Description: touch file failed when fs is not full
1960 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1961 otherwise we can get ENOSP while fs is not full.
1965 Description: parallel lock callbacks
1966 Details : Instead of sending blocking and completion callbacks as separated
1967 requests, adding them to a set and sending in parallel.
1970 Frequency : only for Cray XT3
1971 Bugzilla : 12829/13455
1972 Description: Changing primary group doesn't change the group lustre assigns to
1974 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1975 overridden with the primary group provided by the group upcall,
1976 whereas the supplied fsgid can be trusted if it is in the list of
1977 supplementary groups returned by the group upcall.
1979 Severity : enhancement
1981 Description: Allow masking D_WARNING, D_ERROR messages from console
1982 Details : Console messages can now be disabled via lnet.printk.
1986 Description: User code with malformed file open parameter crashes client node
1987 Details : Before packing join_file req, all the related reference should be
1988 checked carefully in case some malformed flags cause fake
1989 join_file req on client.
1993 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1994 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1995 mode is equal to the granted mode. However, it should also check
1996 that LDLM_FL_CP_REQD is not set.
2000 Description: Heavy nfs access might result in deadlocks
2001 Details : After ELC code landed, it is now improper to enqueue any mds
2002 locks under och_sem, because enqueue might want to decide to
2003 cancel open locks for same inode we are holding och_sem for.
2007 Description: Client eviction while running blogbench
2008 Details : A lot of unlink operations with concurrent I/O can lead to a
2009 deadlock causing evictions. To address the problem, the number of
2010 oustanding OST_DESTROY requests is now throttled to
2011 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2012 callbacks are processed in priority.
2016 Description: enable ACLs on MDS by default
2017 Details : ACLs must be enabled on MDS by default.
2020 Frequency : PPC/PPC64 only
2022 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2023 Details : fix duplicated definitions between asm-ppc64/types.h and
2024 lustre_types.h on PPC.
2027 Frequency : PPC/PPC64 only
2029 Description: asm-ppc/segment.h does not exist
2030 Details : fix compile issue on PPC.
2034 Description: better handle error messages in extents code
2037 Frequency : RHEL4 only
2039 Description: mkfs is very slow on IA64/RHEL4
2040 Details : A performance regression has been discovered in the MPT Fusion
2041 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2042 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2043 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2045 Severity : enhancement
2047 Description: SNMP support enhancement
2048 Details : Adding total number of sampled request for an MDS node in snmp
2051 Severity : enhancement
2053 Description: Optimize ldlm waiting list processing for PR extent locks
2054 Details : When processing waiting list for read extent lock and meeting read
2055 lock that is same or wider to it that is not contended, skip
2056 processing rest of the list and immediatelly return current
2057 status of conflictness, since we are guaranteed there are no
2058 conflicting locks in the rest of the list.
2062 Description: Time out and refuse to reconnect
2063 Details : When the failover node is the primary node, it is possible
2064 to have two identical connections in imp_conn_list. We must
2065 compare not conn's pointers but NIDs, otherwise we can defeat
2066 connection throttling.
2070 Description: port llog fixes from b1_6 into HEAD
2071 Details : Port llog reference couting and some llog cleanups from b1_6
2072 (bug 10800) into HEAD, for protect from panic and access to already
2073 free llog structures.
2077 Description: Detect stride IO mode in read-ahead
2078 Details : When a client does stride read, read-ahead should detect that and
2079 read-ahead pages according to the detected stride pattern.
2083 Description: data checksumming impacts single node performance
2084 Details : add support for several checksum algorithm. Currently, only CRC32
2085 and Adler-32 are supported. The checksum type can be changed on
2086 the fly via /proc/fs/lustre/osc/*/checksum_type.
2090 Description: use adler32 for page checksums
2091 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2096 Description: build for x2 fails
2097 Details : fix compile issue on Cray systems.
2101 Description: Properly match for duplicate locks
2102 Details : Due to different lock order from skiplists code, we need to
2103 traverse entire list for now
2106 Frequency : only on PPC/SLES10
2108 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2109 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2110 be defined. Add a hack in mkfs_lustre.c to work around this compile
2115 Description: LASSERT on MDS when client holding flock lock dies
2116 Details : ldlm pool logic depends on number of granted locks equal to
2117 number of released locks which is not true for flock locks, so
2118 just exclude such locks from consideration.
2122 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2123 Details : Use fsfilt_sync() for both the whole filesystem sync and
2124 individual file sync to eliminate dangerous inode locking
2125 with I_LOCK that can lead to a deadlock.
2129 Description: performance in 1.6.3
2130 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2131 sure that 1MB requests are not fragmented by the block layer.
2133 Severity : enhancement
2135 Description: organize the server-side client stats on per-nid basis
2136 Details : Change the structure of stats under obdfilter and mds to
2144 The "uuid"s file would list the uuids of _active_ exports.
2145 And the clear entry is to clear all stats and stale nids.
2147 Severity : enhancement
2149 Description: eliminate client locks in face of contention
2150 Details : file contention detection and lockless i/o implementation
2151 for contended files.
2155 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2158 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2160 Description: MDS may lose file striping (and hence file data) in some cases
2161 Details : If there are additional extended attributes stored on the MDS,
2162 in particular ACLs, SELinux, or user attributes (if user_xattr
2163 is specified for the client mount options) then there is a risk
2164 of attribute loss. Additionally, the Lustre file striping
2165 needs to be larger than default (e.g. striped over all OSTs),
2166 and an additional attribute must be stored initially in the
2167 inode and then increase in size enough to be moved to the
2168 external attribute block (e.g. ACL growing in size) for file
2173 Description: skiplist implementation simplification
2174 Details : skiplists are used to group compatible locks on granted list
2175 that was implemented as tracking first and last lock of each
2176 lock group the patch changes that to using doubly linked lists
2180 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2181 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2182 because LASSERT on both the data supplied by a client, and the
2183 data on disk is dangerous and incorrect.
2185 Severity : enhancement
2187 Description: Slow truncate/writes to huge files at high offsets.
2188 Details : Directly associate cached pages to lock that protect those pages,
2189 this allows us to quickly find what pages to write and remove
2190 once lock callback is received.
2194 Description: more ldlm soft lockups
2195 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2196 starve other threads from the resource lock for a long time in
2197 case of long waiting queue, so change the debug level from
2198 D_OTHER to the less frequently used D_INFO.
2200 Severity : enhancement
2202 Description: add -gid, -group, -uid, -user options to lfs find
2206 Description: Hung threads in invalidate_inode_pages2_range
2207 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2208 once one is completed. As a result, some RPCs are stuck in the
2209 queue and are never sent.
2213 Description: filter threads hungs on waiting journal commit
2214 Details : Cleanup filter group llog code, then only filter group llog will
2215 be only created in the MDS/OST syncing process.
2219 Description: Procfs and llog threads access destoryed import sometimes.
2220 Details : Sync the import destoryed process with procfs and llog threads by
2221 the import refcount and semaphore.
2223 Severity : enhancement
2225 Description: openlock cache of b1_6 port to HEAD
2230 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2231 Details : If insertion of an extent fails, then discard the inode
2232 preallocation and free data blocks else it can lead to duplicate
2237 Description: don't always update ctime in ext3_xattr_set_handle()
2238 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2239 In some cases the ctime should not be updated, for example for
2240 2.0->1.8 compatibility it is necessary to delete an xattr and it
2241 should not update the ctime.
2245 Bugzilla : 15713/16362
2246 Description: Assertion in iopen_connect_dentry in 1.6.3
2247 Details : looking up an inode via iopen with the wrong generation number can
2248 populate the dcache with a disconneced dentry while the inode
2249 number is in the process of being reallocated. This causes an
2250 assertion failure in iopen since the inode's dentry list contains
2251 both a connected and disconnected dentry.
2255 Description: assertion failure in ldlm_handle2lock()
2256 Details : fix a race between class_handle_unhash() and class_handle2object()
2257 introduced in lustre 1.6.5 by bug 13622.
2262 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2263 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2268 Description: oops in sd_iostats_seq_show()
2269 Details : unloading/reloading the scsi low level driver triggers a kernel
2270 bug when trying to access the sd iostat file.
2275 Description: Kernel panics during QLogic driver reload
2276 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2277 patch, causing memory corruption.
2282 Description: journal_dev option does not work in b1_6
2283 Details : pass mount option during pre-mount.
2285 Severity : enhancement
2287 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2288 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2289 extent information of a file. It can be used to map logical blocks
2290 in a file to physical blocks in the block device.
2294 Description: LDLM soft lockups - improvement
2295 Details : It is be possible to send the lock handle along with each read
2296 or write request because the client is already doing a lock match
2297 itself so there isn't any reason the OST should have to re-do that
2301 Frequency : only on Cray X2
2303 Description: X2 build failures
2304 Details : fix build failures on Cray X2.
2308 Description: xid & resent requests
2309 Details : Initialize RPC XID from clock at startup (randomly if clock is
2312 Severity : enhancement
2314 Description: Add lustre_start utility to start or stop multiple Lustre servers
2319 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2320 Details : In case of memory pressure, list_del() can be called twice on
2321 req->rq_history_list, causing a kernel oops.
2325 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2326 Details : fix stack overflow in the distributed lock manager by defering
2327 export eviction after a failed AST to the elt thread instead of
2328 handling it in the dlm interpret routine.
2332 Description: Convert some comments to new format.
2333 Details : Update documenting comments to match doxygen conventions.
2337 Description: Grammar fixes.
2338 Details : A couple of trivial spelling fixes.
2342 Description: OSD_COUNTERS-mandatory
2343 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2344 to simplify the code.
2348 Description: simplify cmm_device freeing logic.
2349 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2353 Description: Add lockdep support to dt_object_operations locking interface.
2354 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2355 parameter indicating lock ordering. Update mdd code to use new
2360 Description: Introduce failloc constants for lockless IO tests.
2361 Details : Add two new failloc constants to test lockless IO. Only one of
2362 them in implemented---another is checked in yet to be landed
2367 Description: Add lockdep support for inode mutex.
2368 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2369 in the situations where Lustre has to lock more than one inode
2374 Description: Add optional invariants checking support.
2375 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2376 configure switch. This macro is to be used for consistency and
2377 sanity checks that are too expensive to be left in `production'
2382 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2383 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2388 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2389 Details : Use spin_lock_nested() in (the only) situation where more than
2390 one ldlm_lock is locked simultaneously. Also, fix possible
2391 dead-lock in ldlm_lock_change_resource() by enforcing particular
2396 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2397 Details : Instead of specifying each ldlm_lock call-back through separate
2398 parameter, wrap them into struct ldlm_callback_suite.
2402 Description: Kill join_lru obd method and its callers.
2403 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2404 in memory---a requirement implemented through ->o_join_lru() obd
2405 method in HEAD. Remove this method and its users.
2409 Description: Add asynchronous ldlm ENQUEUE completion handler.
2410 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2411 case that stock ldlm_completion_ast() cannot handle as it waits
2412 until lock is granted. Introduce new ldlm_completion_ast_async()
2413 for this. Also comment ldlm_completion_ast().
2417 Description: ldlm_error <-> errno conversion.
2418 Details : Add functions to map (rather arbitrary) between LDLM error codes
2419 and standard errno values. CLIO needs this to prevent LDLM specific
2420 constants from escaping ldlm and osc.
2424 Description: Kill unused ldlm_handle2lock_ns() function.
2425 Details : Kill unused ldlm_handle2lock_ns() function.
2429 Description: Add lu_ref support to ldlm_lock
2430 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2431 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2432 and ldlm_resource. LDLM interface has to be changed, because code that
2433 releases a reference on a lock, has to "know" what reference this is.
2434 In the most frequent case
2436 lock = ldlm_handle2lock(handle);
2438 LDLM_LOCK_PUT(lock);
2440 no changes are required. When any other reference (received _not_ from
2441 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2442 instead of LDLM_LOCK_PUT().
2444 Arguably, changes are pervasive, and interface requires some discipline
2445 for proper use. On the other hand, it was very instrumental in finding
2446 a few leaked lock references.
2450 Description: Add ldlm_lock_addref_try().
2451 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2452 attempts to addref a lock that might be being canceled
2457 Description: Add ldlm_weigh_callback().
2458 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2459 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2460 hard-coded `number of pages' logic.
2464 Description: Add lockdep annotations to llog code.
2465 Details : Use appropriately tagged _nested() locking calls in the places
2466 where llog takes more than one ->lgh_lock lock.
2470 Description: Add loi_kms_set().
2471 Details : Wrap kms updates into a helper function.
2475 Description: Constify instances of struct lsm_operations.
2476 Details : Constify instances of struct lsm_operations.
2480 Description: lu_conf support.
2481 Details : On a server, a file system object is uniquely identified
2482 by a fid, which is sufficient to locate and load all object
2483 state (inode). On a client, on the other hand, more data are
2484 necessary instantiate an object. Change lu_object_find() and
2485 friends to take additional `lu_conf' argument describing object.
2486 Typically this includes layout information.
2490 Description: lu_context fixes.
2491 Details : Introduce new lu_context functions that are needed on the client
2492 side, where some system threads (ptlrpcd) are shared by multiple
2493 modules, and so cannot be stopped during module shutdown.
2497 Description: Add start and stop methods to lu_device_type_operations.
2498 Details : Introduce two new methods in lu_device_type_operations, that are
2499 invoked when first instance of a given type is created and last one
2500 is destroyed respectively. This is need by CLIO.
2504 Description: Add lu_ref support to struct lu_device.
2505 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2506 track leaked references.
2510 Description: Introduce lu_kmem_descr.
2511 Details : lu_kmem_descr and its companion interface allow to create
2512 and destroy a number of kmem caches at once.
2516 Description: Fix lu_object finalization race.
2517 Details : Fix a race between lu_object_find() finding an object and its
2518 concurrent finalization. This race is (most likely) not possible
2519 on the server, but might happen on the client.
2523 Description: Introduce lu_ref interface.
2524 Details : lu_ref is a debugging module allowing to track references to
2525 a given object. It is quite cpu expensive, and has to be
2526 explicitly enabled with --enable-lu_ref. See usage description
2531 Description: Factor lu_site procfs stats into a separate function.
2532 Details : Separate lu_site stats printing code into a separate function
2533 that can be reused on a client.
2537 Description: Constify instances of struct {lu,dt,md}_device_operations.
2538 Details : Constify instances of struct {lu,dt,md}_device_operations.
2542 Description: Introduce struct md_site and move meta-data specific parts of
2543 struct lu_site here.
2544 Details : Move md-specific fields out of struct lu_site into special struct
2545 md_site, so that lu_site can be used on a client.
2549 Description: Kill mdd_lov_destroy().
2550 Details : Remove unused mdd code.
2554 Description: Add st_block checking to multistat.c.
2555 Details : Add st_block checking to multistat.c.
2559 Description: Add lu_ref support to struct obd_device.
2560 Details : Add lu_ref tracking to obd_device.
2564 Description: Kill obd_set_fail_loc().
2565 Details : Remove unused code.
2569 Description: Add special type for ptlrpc_request interpret functions.
2570 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2571 there. Actual usage will be in CLIO.
2575 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2576 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2577 doesn't work with lockdep.
2581 Description: Add rwv.c test program.
2582 Details : New testing program exercising readv(2) and writev(2) (Qian).
2586 Description: Add sendfile.c test program.
2587 Details : New testing program exercising sendfile(2) (Jay).
2591 Description: Ratelimit a message that can be very frequent.
2592 Details : Ratelimit a memory allocation failure message that can
2597 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2598 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2602 Description: Liblustre build fixes.
2603 Details : Liblustre build fixes.
2607 Description: libcfs: add cfs_{need,cond}_resched() interface.
2608 Details : libcfs: add cfs_{need,cond}_resched() definition and
2609 implementations for Linux, NT, and liblustre.
2611 Severity : enhancement
2613 Description: More exported tunables for mballoc
2614 Details : Add support for tunable preallocation window and new tunables for
2615 large/small requests
2619 Description: Detect corruption of block bitmap and checking for preallocations
2620 Details : Checks validity of on-disk block bitmap. Also it does better
2621 checking of number of applied preallocations. When corruption is
2622 found, it turns filesystem readonly to prevent further corruptions.
2626 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2627 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2628 window to 0. If it does hit cache miss, and read-pattern is still
2629 stride-io mode, does not reset the stride window, but also does not
2630 increase the stride window length in this case.
2634 Frequency : only for big-endian servers
2635 Description: Check if system is big-endian while mounting fs with extents feature
2636 Details : Mounting a filesystem with extents feature will fail on big-endian
2637 systems since ext3-based ldiskfs is not supported on big-endian
2638 systems. This can be over-riden with "bigendian_extents" mount option.
2640 Severity : enhancement
2642 Description: The root squash functionality
2643 Details : A security feature, which is to prevent users from being able
2644 to mount lustre on their desktop, run as root, and delete
2645 all of the files in the filesystem. The goal is accomplished by
2646 remapping user id (UID) and group id (GID) of the root user to
2647 a UID and GID specified by the system administartor via Lustre
2648 configuration management server (MGS). The functionality also
2649 allows to specify sets of clients for which the remapping does
2654 Description: Excessive recovery window
2655 Details : With AT enabled, the recovery window can be excessively long (6000+
2656 seconds). To address this problem, we no longer use
2657 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2658 timeout no longer depends on the service time, it is set to
2659 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2660 time via pb_service_time.
2664 Description: Watchdog triggered on MDS failover
2665 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2666 the OSTs know that the MDS "UUID" can be reused for the same export
2667 from a different NID, so we do not need to wait for the export to be
2671 Frequency : rare, only if using MMP with Linux RAID
2673 Description: MMP doesn't work with Linux RAID
2674 Details : While using HA for Lustre servers with Linux RAID, it is possible
2675 that MMP will not detect multiple mounts. To make this work we
2676 need to unplug the device queue in RAID when the MMP block is being
2677 written. Also while reading the MMP block, we should read it from
2678 disk and not the cached one.
2680 Severity : enhancement
2682 Description: open file using fid
2683 Details : A file can be opened using just its fid, like
2684 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2687 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2690 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2691 Details : add proper locking for creation of jbd_4k slab cache
2695 Description: MMP check in ext3_remount() fails without displaying any error
2696 Details : When multiple mount protection fails during remount, proper error
2699 Severity : enhancement
2701 Description: Allow stripe size to be up to 4G-64k
2702 Details : Fix math logic to allow large stripe sizes.
2706 Description: add check for >8TB ldiskfs filesystems
2707 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2708 Don't allow >8TB ldiskfs filesystems to be mounted without
2709 force_over_8tb mount option
2714 Descriptoin: OSS complains "too many blocks" and crashes
2715 Details : In filter_commitrw_write():
2716 1. lquota_chkquota();
2717 2. obdo_from_inode();
2718 3. lquota_pending_commit();
2719 "2." has changed uid and gid in this case, then caused
2720 lquota_chkquota() and lquota_pending_commit() to use
2721 different uid and gid. That is the root of the bug.
2725 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2726 Details : Detect and print out the 'stripe count = -1' case properly.
2730 Description: Out or order replies might be lost on replay
2731 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2732 request with tid smaller than one currently being added, add it
2733 to the start, not end of the list.
2735 --------------------------------------------------------------------------------
2737 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2739 * Support for kernels:
2740 2.6.5-7.283 (SLES 9),
2741 2.6.9-55.EL (RHEL 4),
2742 2.6.16.46-0.14 (SLES 10),
2743 2.6.18.8 vanilla (kernel.org)
2744 * Client support for unpatched kernels:
2745 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2746 2.6.16 - 2.6.22 vanilla (kernel.org)
2747 * Due to recently discovered recovery problems, we do not recommend
2748 using patchless RHEL 4 clients with this or any earlier release.
2749 * Recommended e2fsprogs version: 1.39.cfs8
2750 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2751 * Starting with this release, the ldiskfs backing filesystem required
2752 by Lustre is now in its own package, lustre-ldiskfs. This package
2753 should be installed. It is versioned separately from Lustre and
2754 may be released separately in future.
2758 Description: block reactivating mgc import until all deactivates complete
2759 Details : Fix race when failing back MDT/MGS to itself (testing)
2762 Frequency : at statup only
2764 Description: mds_lov_synchronize race leads to various problems
2765 Details : simultaneous MDT->OST connections at startup can cause the
2766 sync to abort, leaving the OSC in a bad state.
2768 Severity : enhancement
2770 Description: add optional extra BUILD_VERSION info
2771 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2772 to override the lustre version.
2775 Frequency : 2.6.18 servers only
2777 Description: ll_kern_mount() doesn't release the module reference
2778 Details : The ldiskfs module reference count never drops down to 0
2779 because ll_kern_mount() doesn't release the module reference.
2784 Description: server LBUG when using old ost_num_threads parameter
2785 Details : Accept the old ost_num_threads parameter but warn that it
2786 is deprecated, and fix an off-by-one error that caused an LBUG.
2791 Description: Transient SCSI error results in persistent IO issue
2792 Details : iobuf->dr_error is not reinitialized to 0 between two
2796 Frequency : sometimes when underlying device returns I/O errors
2798 Description: OSTs not going read-only during write failures
2799 Details : OSTs are not remounted read-only when the journal commit threads
2800 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2801 instead of the ext3 wrappers.
2805 Description: poor connect scaling with increasing client count
2806 Details : Don't run filter_grant_sanity_check for more than 100 exports
2807 to improve scaling for large numbers of clients.
2810 Frequency : SLES10 only
2812 Description: sanity-quota.sh quotacheck failed: rc = -22
2813 Details : Quotas cannot be enabled on SLES10.
2816 Frequency : liblustre clients only
2818 Description: getdirentries does not give error when run on compute nodes
2819 Details : getdirentries does not fail when the size specified as an argument
2820 is too small to contain at least one entry
2822 Severity : enhancement
2824 Description: Add LNET router traceability for debug purposes
2825 Details : If a checksum failure occurs with a router as part of the
2826 IO path, the NID of the last router that forwarded the bulk data
2827 is printed so it can be identified.
2832 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2833 Details : Due to a race condition, liblustre clients were occasionally
2834 evicted incorrectly.
2836 Severity : enhancement
2838 Description: lfs setstripe use optional parameters instead of postional
2841 Severity : enhancement
2843 Description: Nanosecond timestamp support for ldiskfs
2844 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2845 resolution timestamps. There is not yet support for this at
2846 the Lustre filesystem level.
2849 Frequency : during server recovery
2851 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2852 Details : request with rq_no_resend flag not awake l_wait_event if they get
2856 Frequency : nfs export on patchless client
2858 Description: connectathon hang when test nfs export over patchless client
2859 Details : Disconnected dentry cannot be found with lookup, so we do not need
2860 to unhash it or make it invalid
2863 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2864 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2865 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2870 Description: open req refcounting wrong on reconnect
2871 Details : If reconnect happened between getting open reply from server and
2872 call to mdc_set_replay_data in ll_file_open, we will schedule
2873 replay for unreferenced request that we are about to free.
2874 Subsequent close will crash in variety of ways.
2875 Check that request is still eligible for replay in
2876 mdc_set_replay_data().
2881 Description: disable writes to filesystem when reading health_check file
2882 Details : the default for reading the health_check proc file has changed
2883 to NOT do a journal transaction and write to disk, because this
2884 can cause reads of the /proc file to hang and block HA state
2885 checking on a healthy but otherwise heavily loaded system. It
2886 is possible to return to the previous behaviour during configure
2887 with --enable-health-write.
2889 Severity : enhancement
2891 Description: 64-bit inode version
2892 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2893 to the inode. This will be required for version-based recovery.
2898 Description: MDS fails to start if a duplicate client export is detected
2899 Details : in some rare cases it was possible for a client to connect to
2900 an MDS multiple times. Upon recovery the MDS would detect this
2901 and fail during startup. Handle this more gracefully.
2903 Severity : enhancement
2905 Description: Add -o localflock option to simulate old noflock
2907 Details : This will achieve local-only flock/fcntl locks
2913 Description: log_commit_thread vs filter_destroy race leads to crash
2914 Details : Take import reference before releasing llog record semaphore
2919 Description: Wrong request locking in request set processing
2920 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2921 imp_delayed_list, in this place should be used imp_lock.
2924 Frequency : when reconnection
2926 Description: Grant Leak when osc reconnect to OST
2927 Details : When osc reconnect ost, OST(filter) should check whether it
2928 should grant more space to client by comparing fed_grant and
2929 cl_avail_grant, and return the granted space to client instead
2930 of "new granted" space, because client will call osc_init_grant
2931 to update the client grant space info.
2934 Frequency : when client reconnect to OST
2936 Description: Grant Leak when osc do resend and replay bulk write
2937 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2938 bulk write request, because the grant info while be sync between
2939 OSC and OST when reconnect, and we should ignore the grant info
2940 these of resend/replay write req.
2945 Description: Grant space more than available left space sometimes.
2946 Details : When then OST is about to be full, if two bulk writing from
2947 different clients came to OST. Accord the avaliable space of the
2948 OST, the first req should be permitted, and the second one
2949 should be denied by ENOSPC. But if the seconde arrived before
2950 the first one is commited. The OST might wrongly permit second
2951 writing, which will cause grant space > available space.
2954 Frequency : when client is evicted
2956 Description: Grant might be wrongly erased when osc is evicted by OST
2957 Details : when the import is evicted by server, it will fork another
2958 thread ptlrpc_invalidate_import_thread to invalidate the
2959 import, where the grant will be set to 0. While the original
2960 thread will update the grant it got when connecting. So if
2961 the former happened latter, the grant will be wrongly errased
2962 because of this race.
2967 Description: Checking Stale with correct fid
2968 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2969 to check whether it is stale, because sometimes, we want the
2970 enqueue happened anyway, and op_data.fid2 will not be initialized.
2972 Severity : enhancement
2974 Description: update patchless client
2975 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2978 Frequency : only with 2.4 kernel
2980 Description: random memory corruption
2981 Details : size of struct ll_inode_info is to big for union inode.u and this
2982 can be cause of random memory corruption.
2987 Description: Memory leak in recovery
2988 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2989 It should also check obd_fail before fsfilt_start, otherwise if
2990 fsfilt_start return -EROFS,(failover mds during mds recovery).
2991 then the req will return with repmsg->transno = 0 and rc = EROFS.
2992 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2993 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2994 be freed no matter whether fsfilt_commit success or not.
2997 Frequency : only with huge count clients
2999 Description: Prevents from taking the superblock lock in llap_from_page for
3001 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3002 for llap_from_page call in ll_removepage prevents from taking the
3003 superblock lock for a soon died page.
3008 Description: Not check open intent error before release open handle
3009 Details : in some rare cases, the open intent error is not checked before
3010 release open handle, which may cause
3011 ASSERTION(open_req->rq_transno != 0), because it tries to release
3012 the failed open handle.
3017 Description: Set cat log bitmap only after create log success.
3018 Details : in some rare cases, the cat log bitmap is set too early. and it
3019 should be set only after create log success.
3023 Description: Accessing a block bevice can re-enable I/O when Lustre is
3024 tearing down a device.
3025 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3030 Description: service threads may hog cpus when there are a lot of requests
3031 Details : Insert cond_resched to give other threads a chance to use some CPU
3036 Description: the cat log was not initialized in recovery
3037 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3038 unlink log on mds will not be initialized until mds post
3039 recovery. And also in mds post recovery, the unlink log will
3040 initialization will be done asynchronausly, so there will be race
3041 between add unlink log and unlink log initialization.
3045 Description: brw_stats were being printed incorrectly
3046 Details : brw_stats were being printed as log2 but all of them were not
3047 recorded as log2. Also remove some code duplication arising from
3048 filter_tally_{read,write}.
3052 Frequency : rare, only in recovery.
3053 Description: ASSERTION(req->rq_type != LI_POISON) failed
3054 Details : imp_lock should be held while iterating over imp_sending_list for
3055 prevent destroy request after get timeout in ptlrpc_queue_wait.
3059 Description: replay-single.sh test 52 fails
3060 Details : A lock's skiplist need to be cleanup when it being unlinked
3061 from its resource list.
3065 Description: Short directio read returns full requested size rather than
3067 Details : Direct I/O operations should return actual amount of bytes
3068 transferred rather than requested size.
3070 Severity : enhancement
3072 Description: metadata RPC reduction (e.g. for rm performance)
3073 Details : decrease the amount of synchronous RPC between clients and servers
3074 by canceling conflicing lock before the operation on the client
3075 and packing thier handles into the main operation RPC to server.
3077 Severity : enhancement
3079 Description: Async OSC create to avoid the blocking unnecessarily.
3080 Details : If an OST has no remaining object, system will block on the
3081 creation when it needs to create a new object on this OST. Now,
3082 ways use pre-created objects when available, instead of blocking on
3083 an empty osc while others are not empty. If we must block, we
3084 block for the shortest possible period of time.
3088 Description: improve handling recoverable errors
3089 Details : if request processig with error which can be recoverable on server
3090 request should be resend, otherwise page released from cache and
3093 Severity : enhancement
3095 Description: refine locking for avoid write wrong info into lov_objid file
3096 Details : fix possible races with add new target and write/update data in
3099 --------------------------------------------------------------------------------
3101 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3106 Frequency : on some architectures
3108 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3109 Details : Uninitialized flags sometimes cause configuration commands to
3113 Frequency : patchless clients only
3115 Description: missing __iget() symbol export
3116 Details : The __iget() symbol export is missing. To avoid the need for
3117 this on patchless clients the deathrow inode reaper is turned
3118 off, and we depend on the VM to clean up old inodes. This
3119 dependency was during via the fix for bug 12181.
3121 --------------------------------------------------------------------------------
3123 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3125 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3126 INTEROPERATE with older versions automatically. In many cases a
3127 special upgrade step is needed. Please read the
3128 user documentation before upgrading any part of a live system.
3129 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3130 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3131 * WARNING: Lustre configuration and startup changes are required with
3132 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3136 Severity : enhancement
3138 Description: MountConf
3139 Details : Lustre configuration is now managed via mkfs and mount
3140 commands instead of lmc and lconf. New obd types (MGS, MGC)
3141 are added for dynamic configuration management. See
3142 https://mail.clusterfs.com/wikis/lustre/MountConf for
3145 Severity : enhancement
3147 Description: dynamic OST addition
3148 Details : OSTs can now be added to a live filesystem
3150 Severity : enhancement
3152 Description: startup order invariance
3153 Details : MDTs and OSTs can be started in any order. Clients only
3154 require the MDT to complete startup.
3156 Severity : enhancement
3158 Description: parallel, asynchronous orphan cleanup
3159 Details : orphan cleanup is now performed in separate threads for each
3160 OST, allowing parallel non-blocking operation.
3162 Severity : enhancement
3164 Description: optimized stripe assignment
3165 Details : stripe assignments are now made based on ost space available,
3166 ost previous usage, and OSS previous usage, in order to try
3167 to optimize storage space and networking resources.
3169 Severity : enhancement
3171 Description: Permanently set tunables
3172 Details : All writable /proc/fs/lustre tunables can now be permanently
3173 set on a per-server basis, at mkfs time or on a live system.
3175 Severity : enhancement
3177 Description: Lustre message v2
3178 Details : Add lustre message format v2.
3180 Severity : enhancement
3182 Description: client OST exclusion list
3183 Details : Clients can be started with a list of OSTs that should be
3184 declared "inactive" for known non-responsive OSTs.
3188 Description: SPEC SFS validation failure on NFS v2 over lustre.
3189 Details : Changes the blocksize for regular files to be 2x RPC size,
3190 and not depend on stripe size.
3192 Severity : enhancement
3194 Description: Multiple MD RPCs in flight.
3195 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3196 To avoid overly-overloading MDS, introduce a limit on number of
3197 MDS RPCs in flight for a single client and add /proc controls
3198 to adjust this limit.
3200 Severity : enhancement
3202 Description: client read/write statistics
3203 Details : Add client read/write call usage stats for performance
3204 analysis of user processes.
3205 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3206 file access. extents_stats shows chunk size distribution.
3207 extents_stats_per_process show chunk size distribution per
3210 Severity : enhancement
3212 Description: mds statistics
3213 Details : Add detailed mds operations statistics in
3214 /proc/fs/lustre/mds/*/stats.
3218 Description: Failure of copying files with lustre special EAs.
3219 Details : Client side always return success for setxattr call for lustre
3220 special xattr (currently only "trusted.lov").
3225 Description: Refcount LNET uuids
3226 Details : The global LNET uuid list grew linearly with every startup;
3227 refcount repeated list entries instead of always adding to
3230 Severity : enhancement
3232 Description: Dynamic service threads
3233 Details : Within a small range, start extra service threads
3234 automatically when the request queue builds up.
3237 Frequency : mixed-endian client/server environments
3239 Description: mixed-endian crashes
3240 Details : The new msg_v2 system had some failures in mixed-endian
3243 Severity : enhancement
3245 Description: Easy OST removal
3246 Details : OSTs can be permanently deactivated with e.g. 'lctl
3247 conf_param lustre-OST0001.osc.active=0'
3249 Severity : enhancement
3251 Description: MGS proc entries
3252 Details : Added basic proc entries for the MGS showing what filesystems
3255 Severity : enhancement
3257 Description: provide MGS failover
3258 Details : Added config lock reacquisition after MGS server failover.
3260 Severity : enhancement
3262 Description: add Linux 2.4 support
3263 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3267 Description: a large application tries to do I/O to the same resource and dies
3268 in the middle of it.
3269 Details : Check the req->rq_arrival time after the call to
3270 ost_brw_lock_get(), but before we do anything about
3271 processing it & sending the BULK transfer request. This
3272 should help move old stale pending locks off the queue as
3273 quickly as obd_timeout.
3276 Frequency : when an incorrect nid is specified during startup
3278 Description: ptlrpc connect to non-existant node causes kernel crash
3279 Details : LNET can't be re-entered from an event callback, which
3280 happened when we expire a message after the export has been
3281 cleaned up. Instead, hand the zombie cleanup off to another
3284 Severity : enhancement
3286 Description: plain/inodebits lock performance improvement
3287 Details : Grouping plain/inodebits in granted list by their request modes
3288 and bits policy, thus improving the performance of search through
3292 Frequency : only if OST filesystem is corrupted
3294 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3295 Details : for a short time RPCs with bulk IO are in the replay list,
3296 but replay of bulk IOs is unimplemented. If the OST filesystem
3297 is corrupted due to disk cache incoherency and then replay is
3298 started it is possible to trip an assertion. Avoid putting
3299 committed RPCs into the replay list at all to avoid this issue.
3302 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3304 Description: System hang on startup
3305 Details : This bug allowed the liblustre (e.g. catamount) client to
3306 return to the app before handling all startup RPCs. This
3307 could leave the node unresponsive to lustre network traffic
3308 and manifested as a server ptllnd timeout.
3310 Severity : enhancement
3312 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3313 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3314 boolean to control whether to print peer debug info when a
3315 client's RPC times out.
3318 Frequency : only for kernels with patches from Lustre below 1.4.3
3320 Description: Remove old rdonly API
3321 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3324 Frequency : only for devices with external journals
3326 Description: Set external device read-only also
3327 Details : During a commanded failover stop, we set the disk device
3328 read-only while the server shuts down. We now also set any
3329 external journal device read-only at the same time.
3332 Frequency : when upgrading from 1.4 while trying to change parameters
3334 Description: The wrong (new) MDC name was used when setting parameters for
3335 upgraded MDT's. Also allows changing of OSC (and MDC)
3336 parameters if --writeconf is specified at tunefs upgrade time.
3339 Frequency : when setting specific ost indicies
3341 Description: QOS code breaks on skipped indicies
3342 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3343 created with --index need not be sequential.
3345 Severity : enhancement
3347 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3348 Details : The uninit_groups feature works in conjunction with the kernel
3349 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3350 up the pass1 processing of e2fsck. This is a read-only feature
3351 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3352 mount filesystems that have had this feature enabled.
3354 Severity : enhancement
3356 Description: Improve multi-block allocation algorithm to avoid fragmentation
3357 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3358 allocation locality and avoid filesystem fragmentation.
3360 ------------------------------------------------------------------------------
3362 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3364 * Support for kernels:
3365 2.4.21-47.0.1.EL (RHEL 3)
3366 2.6.5-7.283 (SLES 9)
3367 2.6.9-42.0.10.EL (RHEL 4)
3368 2.6.12.6 vanilla (kernel.org)
3369 2.6.16.27-0.9 (SLES 10)
3370 * Recommended e2fsprogs version: 1.39.cfs5
3372 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3376 Frequency : occasional, depends on client load and configuration
3377 Bugzilla : 12181, 12203
3378 Description: data loss for recently-modified files
3380 Details : In some cases it is possible that recently written or created
3381 files may not be written to disk in a timely manner (this should
3382 normally be within 30s unless client IO load is very high).
3383 The problem appears as zero-length files or files that are a
3384 multiple of 1MB in size after a client crash or client eviction
3385 that are missing data at the end of the file.
3387 This problem is more likely to be hit on clients where files are
3388 repeatedly created and unlinked in the same directory, clients
3389 have a large amount of RAM, have many CPUs, the filesystem has
3390 many OSTs, the clients are rebooted frequently, and/or the files
3391 are not accessed by other nodes after being written.
3393 The presence of the problem can be detected by looking at
3394 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3395 smaller than the second (nr_unused) then dirty files will not
3396 be flushed automatically to disk. "sync; sleep 10" should be
3397 run several times on the node before unmounting it to update
3398 Lustre (this is also safe to run on nodes without this problem).
3400 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3401 that can cause this same problem, so customers using that kernel
3402 also need to update the kernel in addition to Lustre. In order
3403 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3405 It is normal that files written just before a client crash (less
3406 than 30s) may not yet have been flushed to disk, even for local
3410 Frequency : frequent on thin XT3 nodes
3412 Description: UUID collision on thin XT3 Linux nodes
3413 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3414 randomly, since we relied on an insufficiently-seeded PRNG.
3419 Description: OSS hangs after "All ost request buffers busy"
3420 Details : A deadlock between quota and journal operations caused OSS
3421 hangs after printing "All ost request buffers busy."
3424 Frequency : always on liblustre builds
3426 Description: Cleanup compiler warnings on liblustre
3429 Frequency : always on liblustre builds on XT3
3431 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3435 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3436 > 32000 subdirectories
3437 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3438 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3439 subdirectory count crosses 32000. This will aid e2fsck to
3440 correctly handle more than 32000 subdirectories.
3443 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3445 Description: System hang on startup
3446 Details : This bug allowed the liblustre (e.g. catamount) client to
3447 return to the app before handling all startup RPCs. This
3448 could leave the node unresponsive to lustre network traffic
3449 and manifested as a server ptllnd timeout.
3451 Severity : enhancement
3453 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3454 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3455 boolean to control whether to print peer debug info when a
3456 client's RPC times out.
3461 Description: make O_SYNC working on 2.6 kernels
3462 Details : 2.6 kernels use different method for mark pages for write,
3463 so need add a code to lustre for O_SYNC work.
3468 Description: Failure to close file and release space on NFS
3469 Details : Put inode details into lock acquired in ll_intent_file_open.
3470 Use mdc_intent_lock in ll_intent_open to properly
3471 detect all kind of errors unhandled by mdc_enqueue.
3476 Description: proc file read during shutdown sometimes raced obd removal,
3478 Details : Add lock to prevent obd access after proc file removal.
3481 Frequency : Only for files larger than 4GB on 32-bit clients.
3483 Description: improperly doing page alignment of locks
3484 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3485 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3490 Description: RPCs being resent when they shouldn't be
3491 Details : Some RPCs that should not be resent are being resent. This
3492 can cause inconsistencies in the RPC state machine. Do not
3493 resend such requests.
3496 Frequency : rare, only with NFS export
3498 Description: Crash on NFS re-export node
3499 Details : under very unusual load conditions an assertion is hit in
3500 ll_intent_file_open()
3503 Frequency : only if OST filesystem is corrupted
3505 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3506 Details : for a short time RPCs with bulk IO are in the replay list,
3507 but replay of bulk IOs is unimplemented. If the OST filesystem
3508 is corrupted due to disk cache incoherency and then replay is
3509 started it is possible to trip an assertion. Avoid putting
3510 committed RPCs into the replay list at all to avoid this issue.
3515 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3516 Details : Large single O_DIRECT read and write calls can fail to allocate
3517 a sufficiently large buffer to process the request. In case of
3518 allocation failure the allocation is retried with a smaller
3519 buffer and broken into smaller requests.
3521 Severity : enhancement
3523 Description: Add -o localflock option to simulate old noflock behaviour.
3524 Details : This will achieve local-only flock/fcntl locks coherentness.
3529 Description: versioning check is incomplete
3530 Details : Checking the version difference of client vs. server, report
3531 error if the gap is too big.
3536 Description: add support PG_writeback bit
3537 Details : add support for PG_writeback bit for Lustre, for more carefull
3538 work with page cache in 2.6 kernel. This also fix some deadlocks
3539 and remove hack for work O_SYNC with 2.6 kernel.
3541 Severity : enhancement
3543 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3544 Details : The uninit_groups feature works in conjunction with the kernel
3545 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3546 up the pass1 processing of e2fsck. This is a read-only feature
3547 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3548 mount filesystems that have had this feature enabled.
3550 Severity : enhancement
3552 Description: Improve multi-block allocation algorithm to avoid fragmentation
3553 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3554 allocation locality and avoid filesystem fragmentation.
3556 ------------------------------------------------------------------------------
3558 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3560 * Support for kernels:
3561 2.6.9-42.0.3.EL (RHEL 4)
3562 2.6.5-7.276 (SLES 9)
3563 2.4.21-47.0.1.EL (RHEL 3)
3564 2.6.12.6 vanilla (kernel.org)
3565 2.6.16.21-0.8 (SLES10)
3566 * Recommended e2fsprogs version: 1.39.cfs2-0
3568 * The backwards-compatible /proc/sys/portals symlink has been removed
3569 in this release. Before upgrading, please ensure that you change
3570 any configuration scripts or /etc/sysctl.conf files that access
3571 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3572 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3573 in advance of the upgrade on any system running Lustre 1.4.6 or
3574 newer, since /proc/sys/lnet was added in that version.
3575 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3579 Frequency : only when quota is used
3581 Description: avoid scanning export list for quota master
3582 Details : Change the algorithms to avoid scanning export list in order
3583 to improve the efficiency.
3586 Frequency : MDS failover only, very rarely
3588 Description: "went back in time" messages on mds failover
3589 Details : The greatest transno may be lost when the current operation
3590 finishes with an error (transno==0) and the client's last_rcvd
3591 record is over-written. Save the greatest transno in the
3592 mds_last_transno for this case.
3595 Frequency : always for specific kernels and striping counts
3597 Description: client may get "Matching packet too big" without ACL support
3598 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3599 when trying to access files in certain configurations. The
3600 clients should in fact be denied when mounting because they do
3601 not understand ACLs.
3604 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3606 Description: many clients connecting with IO in progress causes connect timeouts
3607 Details : Avoid synchronous journal commits to avoid delays caused by many
3608 clients connecting/disconnecting when bulk IO is in progress.
3609 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3610 OST_IO_PORTAL to avoid delays behind potentially many pending
3614 Frequency : occasionally with multiple writers to a single file
3616 Description: shared writes to file may result in wrong size reported by stat()
3617 Details : Allow growing of kms when extent lock is cancelled
3620 Frequency : always with random mmap IO to multi-striped file
3622 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3623 Details : Only if the hole is at the end of OST object so that kms is too
3624 small. Fix is to increase kms accordingly in ll_nopage.
3627 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3629 Description: writes to a missing object would leak memory on the OST
3630 Details : If there is an inconsistency between the MDS and OST filesystems,
3631 such that the MDS references an object that doesn't exist, writes
3632 to that object will leak memory due to incorrect cleanup in the
3633 error handling path, eventually running out of memory on the OST.
3638 Description: Creating too long symlink causes lustre errors
3639 Details : Check symlink and name lengths before sending requests to MDS.
3642 Frequency : only if flock is enabled (not on by default)
3644 Description: posix locks not released on fd closure on 2.6.9+
3645 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3646 kernels, this caused such locks not to be released on fd close and
3647 then assertions on fs unmount about still used locks.
3650 Frequency : MDS failover only, very rarely
3652 Description: clients may get ASSERTION(granted_lock != NULL)
3653 Details : When request was taking a long time, and a client was resending
3654 a getattr by name lock request. The were multiple lock requests
3655 with the same client lock handle and
3656 mds_getattr_name->fixup_handle_for_resent_request found one of the
3657 lock handles but later failed with ASSERTION(granted_lock != NULL).
3662 Description: handle->h_buffer_credits > 0, assertion failure
3663 Details : h_buffer_credits is zero after truncate, causing assertion
3664 failure. This patch extends the transaction or creates a new
3668 Frequency : NFS re-export or patchless client
3669 Bugzilla : 11179, 10796
3670 Description: Crash on NFS re-export node (__d_move)
3671 Details : We do not want to hash the dentry if we don't have a lock.
3672 But if this dentry is later used in d_move, we'd hit uninitialised
3673 list head d_hash, so we just do this to init d_hash field but
3674 leave dentry unhashed.
3677 Frequency : NFS re-export or patchless client
3679 Description: NFS exports has problem with symbolic link
3680 Details : lustre client didn't properly install dentry when re-exported
3681 to NFS or running patchless client.
3684 Frequency : NFS re-export or patchless client
3686 Description: Various nfs/patchless fixes.
3687 Details : fixes reuse disconected alias for lookup process - this fixes
3688 warning "find_exported_dentry: npd != pd",
3689 fix permission error with open files at nfs.
3690 fix apply umask when do revalidate.
3693 Frequency : occasional
3695 Description: Crash on NFS re-export node
3696 Details : calling clear_page() on the wrong pointer triggered oops in
3697 generic_mapping_read().
3700 Frequency : rarely, using O_DIRECT IO
3702 Description: unaligned directio crashes client with LASSERT
3703 Details : check for unaligned buffers before trying any requests.
3706 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3708 Description: stale data returned from RAID cache
3709 Details : If only a small amount of IO is done to the RAID device before
3710 reading it again it is possible to get stale data from the RAID
3711 cache instead of reading it from disk.
3714 Frequency : always for sles10 kernel
3716 Description: sles10 support
3717 Details : ll_follow_link: compile fixes and using of nd_set_link
3718 under newer kernels.
3721 Frequency : depends on arch, kernel and compiler version, always on sles10
3724 Description: recursive or deep enough symlinks cause stack overflow
3725 Details : getting rid of large stack-allocated variable in
3729 Frequency : depends on hardware
3731 Description: lustre write performance loss in the SLES10 kernel
3732 Details : the performance loss is caused by using of write barriers in the
3733 ext3 code. The SLES10 kernel turns barrier support on by
3734 default. The fix is to undo that change for ldiskfs.
3736 ------------------------------------------------------------------------------
3738 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3740 * Support for kernels:
3741 2.6.9-42.0.3EL (RHEL 4)
3742 2.6.5-7.276 (SLES 9)
3743 2.4.21-47.0.1.EL (RHEL 3)
3744 2.6.12.6 vanilla (kernel.org)
3748 Frequency : quota enabled and large files being deleted
3750 Description: releasing more than 4GB of quota at once hangs OST
3751 Details : If a user deletes more than 4GB of files on a single OST it
3752 will cause the OST to spin in an infinite loop. Release
3753 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3758 Description: statfs data retrieved from /proc may be stale or zero
3759 Details : When reading per-device statfs data from /proc, in the
3760 {kbytes,files}_{total,free,avail} files, it may appear
3761 as zero or be out of date.
3764 Frequency : systems with MD RAID1 external journal devices
3766 Description: lconf's call to blkid is confused by RAID1 journal devices
3767 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3768 of returning all block devices that match the journal UUID.
3771 Frequency : always, for aggregate stripe size over 4GB
3773 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3774 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3775 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3778 Frequency : always if "lfs find" used on a local file/directory
3780 Description: "lfs find" segfaults if used on a local file/directory
3781 Details : The case where a directory component was not specified wasn't
3782 handled correctly. Handle this properly.
3785 Frequency : always on ppc64
3787 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3788 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3789 perform endianness conversion.
3792 Frequency : rarely (truncate to non-zero file size after write under load)
3793 Bugzilla : 10730, 10687
3794 Description: Files padded with zeros to next 4K multiple
3795 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3796 it is possible that files that are truncated to a non-zero size
3797 immediately after being written are filled with zero bytes beyond
3798 the truncated size. No file data is lost.
3800 Severity : enhancement
3802 Description: Allow recovery/failover for liblustre clients.
3803 Details : liblustre clients were unaware of failover configurations until
3806 Severity : enhancement
3808 Description: user file locks should fail when not mounting with flock option
3809 Details : Set up an error-returning stub in ll_file_operations.lock field
3810 to prevent incorrect behaviour when client is mounted without
3811 flock option. Also, set up properly f_op->flock field for
3815 Frequency : always on ia64
3817 Description: "lfs df" loops on printing out MDS statfs information
3818 Details : The obd_ioctl_data was not initialized and in some systems
3819 this caused a failure during the ioctl that did not return
3820 an error. Initialize the struct and return an error on failure.
3823 Frequency : SLES 9 only
3825 Description: Error of copying files with lustre special EAs as root
3826 Details : Client side always return success for setxattr call for lustre
3827 special xattr (currently only "trusted.lov").
3830 Frequency : rarely on clusters with both ia64+i386 clients
3832 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3833 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3834 clients are doing concurrent writes to the same file it is
3835 possible that the ia64 clients may LASSERT because the OST
3836 extent locks are not PAGE_SIZE aligned. Ensure that grown
3837 locks are always aligned on the request boundary.
3840 Frequency : specific use, occasional
3842 Description: Overwriting in use executable truncates on-disk binary image
3843 Details : If one node attempts to overwrite an executable in use by
3844 another node, we now correctly return ETXTBSY instead of
3845 truncating the file.
3847 Severity : enhancement
3849 Description: Async OSC create to avoid the blocking unnecessarily.
3850 Details : If an OST has no remaining object, system will block on the creating
3851 when need to create a new object on this OST. Now, ways use
3852 pre-created objects when available, instead of blocking on an
3853 empty osc while others are not empty. If we must block, we block
3854 for the shortest possible period of time.
3859 Description: chmod on Lustre root is propagated to other clients
3860 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3861 invalid by the follow_mount time.
3864 Frequency : liblustre clients only
3866 Description: Race in 'instant cancel' lock handling could lead to such locks
3867 never to be granted in case of SMP MDS
3868 Details : Do not destroy not yet granted but cbpending locks in
3872 Frequency : replay/resend of open
3874 Description: non null lock assetion failure in mds_intent_policy
3875 Details : Trying to replay/resend lockless open requests resulted in
3876 mds_open() returning 0 with no lock. Now it sets a flag if
3877 a lock is going to be returned.
3879 Severity : enhancement
3881 Description: Checksum enhancements
3882 Details : New checksum enhancements allow for resending RPCs that failed
3885 Severity : enhancement
3887 Description: Tunables on number of dirty pages in cacche
3888 Details : Allow to set limit on number of dirty pages cached.
3893 Description: client crash on unmount - lock still has references
3894 Details : In some error handling cases it was possible to leak a lock
3895 reference on a client while accessing a file. This was not
3896 harmful to the client during operation, but would cause the
3897 client to crash when the filesystem is unmounted.
3900 Frequency : specific case, rare
3902 Description: ETXTBSY on mds though file not in use by client
3903 Details : ETXTBSY is no longer incorrectly returned when attempting to
3904 chmod or chown a directory that the user previously tried to
3905 execute or a currently-executing binary.
3908 Frequency : extremely rare except on liblustre-based clients
3910 Description: Lustre space not freed when files are deleted
3911 Details : Clean up open-unlinked files after client eviction. Previously
3912 the unlink was skipped and the files remained as orphans.
3917 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3918 Details : In some cases it was possible to send a blocking callback to
3919 a client doing a glimpse, even though that client didn't get
3920 a lock granted. When the glimpse lock is cancelled on the OST
3921 the freed lock is left on the waiting list and corrupted the list.
3924 Frequency : all core dumps
3926 Description: Broke core dumps to lustre
3927 Details : Negative dentry may be unhashed if parent does not have UPDATE
3928 lock, but some callers, e.g. do_coredump, expect dentry to be
3929 hashed after successful create, hash it in ll_create_it.
3931 ------------------------------------------------------------------------------
3933 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3935 * Support for kernels:
3936 2.6.9-42.0.2.EL (RHEL 4)
3937 2.6.5-7.276 (SLES 9)
3938 2.4.21-40.EL (RHEL 3)
3939 2.6.12.6 vanilla (kernel.org)
3943 Frequency : always on RHEL 3
3945 Description: Number of open files grows over time
3946 Details : The number of open files grows over time, whether or not
3947 Lustre is started. This was due to a filp leak introduced
3948 by one of our kernel patches.
3950 ------------------------------------------------------------------------------
3952 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3954 * Support for kernels:
3955 2.6.9-42.EL (RHEL 4)
3956 2.6.5-7.276 (SLES 9)
3957 2.4.21-40.EL (RHEL 3)
3958 2.6.12.6 vanilla (kernel.org)
3963 Bugzilla : 5719, 9635, 9792, 9684
3964 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3965 Details : If a server is under heavy load and cannot reply to new
3966 connection requests before the client resends the (re)connect,
3967 the connection handling code can behave badly if two service
3968 threads are concurrently handing separate (re)connections from
3969 the same client. Add better locking to the connection handling
3970 code, and ensure that only a single connection will be processed
3971 for a given client UUID, even if the lock is dropped.
3973 Severity : enhancement
3975 Description: add TCP zero-copy support to kernel
3976 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3977 sends if the hardware supports scatter-gather and checksumming.
3978 This allows socklnd to do client-write and server-read more
3979 efficiently and reduce CPU utilization from skbuf copying.
3982 Frequency : only if NFS exporting from client
3984 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3985 Details : If setting attributes on a file created under NFS that had
3986 never been opened it would be possible to oops the client
3987 if the file had no objects.
3990 Frequency : always for liblustre
3992 Description: liblustre client does MDS+OSTs setattr RPC for each write
3993 Details : When doing a write from a liblustre client, the client
3994 incorrectly issued an RPC to the MDS and each OST the file was
3995 striped over in order to update the timestamps. When writing
3996 with small chunks and many clients this could overwhelm the MDS
3997 with RPCs. In all cases it would slow down the write because
3998 these RPCs are unnecessary.
4000 Severity : enhancement
4002 Description: allow number of MDS service threads to be changed at module load
4003 Details : It is now possible to change the number of MDS service threads
4004 running. Adding "options mds mds_num_threads={N}" to the MDS's
4005 /etc/modprobe.conf will set the number of threads for the next
4006 time Lustre is restarted (assuming the "mds" module is also
4007 reloaded at that time). The default number of threads will
4008 stay the same, 32 for most systems.
4013 Description: OST crash if filesystem is unformatted or corrupt
4014 Details : If an OST is started on a device that has never been formatted
4015 or if the filesystem is corrupt and cannot even mount then the
4016 error handling cleanup routines would dereference a NULL pointer.
4021 Description: NULL pointer deref in llap_from_page.
4022 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4023 incorrect) mapping if the page was truncated/reclaimed while it was
4024 searched for. Check for this condition and skip such pages when
4025 doing readahead. Introduce extra check to llap_from_page() to
4026 verify page->mapping->host is non-NULL (so page is not anonymous).
4029 Frequency : Sometimes when using sys_sendfile
4031 Description: "page not covered by a lock" warnings from ll_readpage
4032 Details : sendfile called ll_readpage without right page locks present.
4033 Now we introduced ll_file_sendfile that does necessary locking
4034 around call to generic_file_sendfile() much like we do in
4038 Frequency : with certain MDS communication failures at client mount time
4040 Description: NULL pointer deref after failed client mount
4041 Details : a client connection request may delayed by the network layer
4042 and not be sent until after the PTLRPC layer has timed out the
4043 request. If the client fails the mount immediately it will try
4044 to clean up before the network times out the request. Add a
4045 reference from the request import to the obd device and delay
4046 the cleanup until the network drops the request.
4049 Frequency : occasionally during client (re)connect
4051 Description: assertion failure during client (re)connect
4052 Details : processing a client connection request may be delayed by the
4053 client or server longer than the client connect timeout. This
4054 causes the client to resend the connection request. If the
4055 original connection request is replied in this interval, the
4056 client may trip an assertion failure in ptlrpc_connect_interpret()
4057 which thought it would be the only running connect process.
4060 Frequency : only with obd_echo servers and clients that are rebooted
4062 Description: kernel BUG accessing uninitialized data structure
4063 Details : When running an obd_echo server it did not start the ping_evictor
4064 thread, and when a client was evicted an uninitialized data
4065 structure was accessed. Start the ping_evictor in the RPC
4066 service startup instead of the OBD startup.
4068 Severity : enhancement
4069 Bugzilla : 10193 (patchless)
4070 Description: Remove dependency on various unexported kernel interfaces.
4071 Details : No longer need reparent_to_init, exit_mm, exit_files,
4072 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4075 Frequency : rare (only users of deprecated and unsupported LDAP config)
4077 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4078 Details : LDAP apparently contains 'lustreName' attributes instead of
4079 'name'. A simple remapping of the name is sufficient.
4082 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4084 Description: waiting_locks_callback trips kernel BUG if client is evicted
4085 Details : Running with the dump_on_timeout debug flag turned on makes
4086 it possible that the waiting_locks_callback() can try to dump
4087 the Lustre kernel debug logs from an interrupt handler. Defer
4088 this log dumping to the expired_lock_main() thread.
4090 Severity : enhancement
4092 Description: Support NFS exporting on 2.6 kernels.
4093 Details : Implement non-rawops metadata methods for NFS server to use without
4094 changing NFS server code.
4097 Frequency : very rare (synthetic metadata workload only)
4099 Description: two racing renames might cause an MDS thread to deadlock
4100 Details : Running the "racer" program may cause one MDS thread to rename
4101 a file from being the source of a rename to being the target of
4102 a rename at exactly the same time that another thread is doing
4103 so, and the second thread has already enqueued these locks after
4104 doing a lookup of the target and is trying to relock them in
4105 order. Ensure that we don't try to re-lock the same resource.
4108 Frequency : only very large systems with liblustre clients
4110 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4111 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4112 all OSTs in parallel. This allows the eviction of stale liblustre
4113 clients to proceed much faster than if they were done in series,
4114 and also offers similar improvements for other set_info RPCs.
4119 Description: excessive CPU usage during initial read phase on client
4120 Details : During the initial read phase on a client, it would agressively
4121 retry readahead on the file, consuming too much CPU and impacting
4122 performance (since 1.4.5.8). Improve the readahead algorithm
4123 to avoid this, and also improve some other common cases (read
4124 of small files in particular, where "small" is files smaller than
4125 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4130 Description: MDS crash when receiving packet with unknown intent.
4131 Details : Do not LBUG in unknown intent case, just return -EFAULT
4133 Severity : enhancement
4134 Bugzilla : 9293, 9385
4135 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4136 Details : Do not serialize getattr (non-intent version) and statfs.
4139 Frequency : occasional, when OST network is overloaded/intermittent
4141 Description: client evicted by OST after bulk IO timeout
4142 Details : If a client sends a bulk IO request (read or write) the OST
4143 may evict the client if it is unresposive to its data GET/PUT
4144 request. This is incorrect if the network is overloaded (takes
4145 too long to transfer the RPC data) or dropped the OST GET/PUT
4146 request. There is no need to evict the client at all, since
4147 the pinger and/or lock callbacks will handle this, and the
4148 client can restart the bulk request.
4151 Frequency : Always when mmapping file with no objects
4153 Description: client crashes when mmapping file with no objects
4154 Details : Check that we actually have objects in a file before doing any
4155 operations on objects in ll_vm_open, ll_vm_close and
4161 Description: Request leak when working with deleted CWD
4162 Details : Introduce advanced request refcount tracking for requests
4163 referenced from lustre intent.
4165 Severity : Enhancement
4167 Description: Cache open file handles on client.
4168 Details : MDS now will return special lock along with openhandle, if
4169 requested and client is allowed to hold openhandle, even if unused,
4170 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4171 closing files for every read/write openration.
4173 Severity : Enhancement
4175 Description: Cache open negative dentries on client when possible.
4176 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4177 negative dentries on lock revocation.
4182 Description: Remounting a client read-only wasn't possible with a zconf mount
4183 Details : It wasn't possible to remount a client read-only with llmount.
4185 Severity : enhancement
4186 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4187 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4188 MPICH is included to add Lustre-specific ADIO interfaces.
4189 This is based closely on the UFS ADIO layer and only differs
4190 in file creation, in order to allow the OST striping to be set.
4191 This is user-contributed code and not supported by CFS.
4196 Description: extended inode attributes (immutable, append-only) work improperly
4197 when 2.4 and 2.6 kernels are used on client/server or vice versa
4198 Details : Introduce kernel-independent values for these flags.
4200 Severity : enhancement
4203 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4204 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4205 granularity. Now, Fractional MB granularity is be supported,
4206 this is very useful for low memory system.
4208 Severity : enhancement
4210 Description: Getattr by fid
4211 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4212 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4213 have locks and avoids getattr rpc requests for it on every stat.
4216 Frequency : Always, for filesystems larger than 2TB
4218 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4219 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4220 filesystems larger than 2TB to be mounted with mballoc enabled.
4223 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4225 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4226 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4227 filesystems larger than 2TB is mounted then the kernel will
4228 silently corrupt the start of the filesystem. CONFIG_LBD is
4229 enabled for all CFS-supported kernels, but the possibility of
4230 this happening with a modified kernel config exists.
4232 Severity : enhancement
4234 Description: add client O_DIRECT support for 2.6 kernels
4235 Details : It is now possible to do O_DIRECT reads and writes to files
4236 in the Lustre client mountpoint on 2.6 kernel clients.
4238 Severity : enhancement
4240 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4241 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4242 parallel, not waiting for response from every OST before sending
4243 a rpc to the next OST.
4248 Description: setattr vs write race when updating file timestamps
4249 Details : Client processes that update a file timestamp into the past
4250 right after writing to the file (e.g. tar) it is possible that
4251 the updated file modification time can be reset to the current
4252 time due to a race between processing the setattr and write RPC.
4254 Severity : enhancement
4256 Description: Bring 'lfs find' closer in line with regular Linux find.
4257 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4258 -print0 options and obtains all the needed info through the lustre
4261 Severity : enhancement
4263 Description: support up to 1024 configured devices on one node
4264 Details : change obd_dev array from statically allocated to dynamically
4265 allocated structs as they are first used to reduce memory usage
4270 Description: Flush dirty partially truncated pages during truncate
4271 Details : Immediatelly flush partially truncated pages in filter_setattr,
4272 this way we completely avoid having any pages in page cache on OST
4273 and can retire ugly workarounds during writes to flush such pages.
4278 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4279 Details : protect inode from truncation within vfs_unlink() context
4280 just take a reference before calling vfs_unlink() and release it
4281 when parent's i_sem is free.
4286 Description: last_id value checked outside lock on OST caused LASSERT failure
4287 Details : If there were multiple MDS->OST object precreate requests in
4288 flight, it was possible that the OST's last object id was checked
4289 outside a lock and incorrectly tripped an assertion. Move checks
4290 inside locks, and discard old precreate requests.
4293 Frequency : always, if extents are used on OSTs
4295 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4296 Details : OSTs using the extents format would not zero the high 16 bits of
4297 the index physical block number. This is not a problem for any
4298 OST filesystems smaller than 16TB, and no kernels support ext3
4299 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4300 new/modified files) and can be fixed for existing filesystems
4301 with e2fsprogs-1.39-cfs1.
4306 Description: import connection selection may be incorrect if timer wraps
4307 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4308 import connections to be ignored if the 32-bit jiffies counter
4309 wraps. Use a 64-bit jiffies counter.
4312 Frequency : very large clusters immediately after boot
4314 Description: LNET request buffers exhausted under heavy short-term load
4315 Details : If a large number of client requests are generated on a service
4316 that has previously never seen so many requests it is possible
4317 that the request buffer growth cannot keep up with the spike in
4318 demand. Instead of dropping incoming requests, they are held in
4319 the LND until the RPC service can accept more requests.
4322 Frequency : Sometimes during replay
4324 Description: Assertion failure in ll_local_open after replay.
4325 Details : If replay happened on an open request reply before we were able
4326 to set replay handler, reply will become not swabbed tripping the
4327 assertion in ll_local_open. Now we set the handler right after
4328 recognising of open request
4331 Frequency : very rare
4333 Description: kernel reports "badness in vsnprintf"
4334 Details : Reading from the "recovery_status" /proc file in small chunks
4335 may cause a negative length in lprocfs_obd_rd_recovery_status()
4336 call to vsnprintf() (which is otherwise harmless). Exit early
4337 if there is no more space in the output buffer.
4339 Severity : enhancement
4341 Description: clear OBD RPC statistics by writing to them
4342 Details : It is now possible to clear the OBD RPC statistics by writing
4343 to the "stats" file.
4348 Description: Client mtime is not the same on different clients after utimes
4349 Details : In some cases, the client was using the utimes() syscall on
4350 a file cached on another node. The clients now validate the
4351 ctime from the MDS + OSTs to determine which one is right.
4356 Description: Inability to activate failout mode
4357 Details : lconf script incorrectly assumed that in pythong string's numeric
4358 value is used in comparisons.
4361 Frequency : always with multiple stripes per file
4363 Description: Inefficient object allocation for mutli-stripe files
4364 Details : When selecting which OSTs to stripe files over, for files with
4365 a stripe count that divides evenly into the number of OSTs,
4366 the MDS is always picking the same starting OST for each file.
4367 Return the OST selection heuristic to the original design.
4372 Description: mount failures may take full timeout to return an error
4373 Details : Under some heavy load conditions it is possible that a
4374 failed mount can wait for the full obd_timeout interval,
4375 possibly several minutes, before reporting an error.
4376 Instead return an error as soon as the status is known.
4378 Frequency : quota enabled and large files being deleted
4380 Description: releasing more than 4GB of quota at once hangs OST
4381 Details : If a user deletes more than 4GB of files on a single OST it
4382 will cause the OST to spin in an infinite loop. Release
4383 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4388 Description: statfs data retrieved from /proc may be stale or zero
4389 Details : When reading per-device statfs data from /proc, in the
4390 {kbytes,files}_{total,free,avail} files, it may appear
4391 as zero or be out of date.
4394 Frequency : systems with MD RAID1 external journal devices
4396 Description: lconf's call to blkid is confused by RAID1 journal devices
4397 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4398 of returning all block devices that match the journal UUID.
4401 Frequency : always, for aggregate stripe size over 4GB
4403 Description: assertion fails when trying to use 4GB stripe size
4404 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4405 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4408 Frequency : always on ppc64
4410 Description: the first write on an ext3 filesystem with mballoc got stuck
4411 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4412 perform endianness conversion.
4414 ------------------------------------------------------------------------------
4416 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4418 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4419 INTEROPERATE with older versions automatically. Please read the
4420 user documentation before upgrading any part of a live system.
4421 * WARNING: Lustre networking configuration changes are required with
4422 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4425 * Support for kernels:
4426 2.6.9-22.0.2.EL (RHEL 4)
4427 2.6.5-7.244 (SLES 9)
4428 2.6.12.6 vanilla (kernel.org)
4431 Severity : enhancement
4432 Bugzilla : 7981/8208
4433 Description: Introduced Lustre Networking (LNET)
4434 Details : LNET is new networking infrastructure for Lustre, it includes
4435 a reorganized network configuration mode (see the user
4436 documentation for full details) as well as support for routing
4437 between different network fabrics. Lustre Networking Devices
4438 (LNDs) for the supported network fabrics have also been
4439 created for this new infrastructure.
4441 Severity : enhancement
4442 Description: Introduced Access control lists
4443 Details : clients can set ACLs on files and directories in order to have
4444 more fine-grained permissions than the standard Unix UGO+RWX.
4445 The MDS must be started with the "-o acl" mount option.
4447 Severity : enhancement
4448 Description: Introduced filesystem quotas
4449 Details : Administrators may now establish per-user quotas on the
4452 Severity : enhancement
4454 Description: Configuration change for the XT3
4455 The PTLLND is now used to run Lustre over Portals on the XT3
4456 The configure option(s) --with-cray-portals are no longer used.
4457 Rather --with-portals=<path-to-portals-includes> is used to
4458 enable building on the XT3. In addition to enable XT3 specific
4459 features the option --enable-cray-xt3 must be used.
4464 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4465 Details : A race between allocating a new llcd and re-getting the llcd_lock
4466 allowed another thread to grab newly-allocated llcd.
4468 Severity : enhancement
4470 Description: 2.6 OST async journal commit and locking fix to improve performance
4471 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4472 2.6 kernels are now async as they already were in 2.4 kernels so
4473 that they can commit concurrently with the network bulk transfer.
4474 For block-allocated files the filter allocation semaphore is held
4475 to avoid filesystem fragmentation during allocation. BKL lock
4476 removed for 2.6 xattr operations where it is no longer needed.
4481 Description: lconf incorrectly determined whether two IP networks could talk
4482 Details : In some more complicated routing and multiple-network
4483 configurations, lconf will avoid trying to make a network
4484 connection to a disjoint part of the IP space. It was doing the
4485 math incorrectly for one set of cases.
4490 Description: Fix for potential infinite loop processing records in an llog.
4491 Details : If an llog record is corrupted/zeroed, it is possible to loop
4492 forever in llog_process(). Validate the llog record length
4493 and skip the remainder of the block on error.
4496 Frequency : occasional (liblustre only)
4498 Description: liblustre could not open files whose last component is a symlink
4499 Details : sysio_path_walk() would incorrectly pass the open intent to
4500 intermediate path components.
4503 Frequency : rare (liblustre only with non-standard tuning)
4504 Bugzilla : 7201 (7350)
4505 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4506 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4507 after it has been released, only do this for referenced locks
4509 Severity : enhancement
4511 Description: specifying an (invalid) directory default stripe_size of -1
4512 would reset the directory default striping
4513 Details : stripe_size -1 was used internally to signal directory stripe
4514 removal, now use "all default" to signal dir stripe removal
4515 as a directory striping of "all default" is not useful
4518 Frequency : common for large clusters running liblustre clients
4520 Description: doing an ls when liblustre clients are running is slow
4521 Details : sending a glimpse AST to a liblustre client waits for every AST
4522 to time out, as liblustre clients will not respond. Since they
4523 cannot cache data we refresh the OST lock LVB from disk instead.
4525 Severity : enhancement
4527 Description: doing an ls at the same time as file IO can be slow
4528 Details : enqueue and other "small" requests can be blocked behind many
4529 large IO requests. Create a new OST IO portal for non-IO
4530 requests so they can be processed faster.
4533 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4535 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4536 created file with mode 000
4537 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4538 allow an ftruncate() as a non-root user to a file with mode 000.
4539 Lustre now respects this flag to disable mode checks when
4540 truncating a file owned by the user
4543 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4546 Description: Revoking locks from clients that went dead or catatonic might take
4548 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4549 cancellation of such locks instant on servers without waiting for
4550 any reply from clients. Clients drops these locks when cancel
4551 notification from server is received without replying.
4554 Frequency : liblustre-only, when liblustre client dies or becomes busy
4556 Description: Doing ls on Linux clients can take a long time with active
4558 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4559 granting such locks to it in the first place if possible. Locks
4560 are taken by proxy on the OST during the read or write and
4561 dropped immediately afterward. Add connect flags handling, do
4562 not grant locks to liblustre clients for glimpse ASTs.
4564 Severity : enhancement
4566 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4567 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4568 sequential reads and behaves badly with applications doing random
4569 reads. Improve it by reading ahead at least read region, and
4570 avoiding excessive large RPC for small reads.
4572 Severity : enhancement
4574 Description: Creating more than 1000 files for a single job may cause a load
4575 imbalance on the OSTs if there are also a large number of OSTs.
4576 Details : qos_prep_create() uses an OST index reseed value that is an
4577 even multiple of the number of available OSTs so that if the
4578 reseed happens in the middle of the object allocation it will
4579 still utilize the OSTs as uniformly as possible.
4584 Description: OST or MDS may oops in ping_evictor_main()
4585 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4586 but doesn't restart at beginning of obd_exports_timed list
4589 Severity : enhancement
4591 Description: improve by-nid export eviction on the MDS and OST
4592 Details : allow multiple exports with the same NID to be evicted at one
4593 time without re-searching the exports list.
4596 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4598 Description: MDS may oops in groups_free()
4599 Details : in rare race conditions a newly allocated group_info struct is
4600 freed again, and this can be NULL. The 2.4 compatibility code
4601 for groups_free() checked for a NULL pointer, but 2.6 did not.
4604 Frequency : common for liblustre clients doing little filesystem IO
4605 Bugzilla : 9352, 7313
4606 Description: server may evict liblustre clients accessing contended locks
4607 Details : if a client is granted a lock or receives a completion AST
4608 with a blocking AST already set it would not reply to the AST
4609 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4612 Frequency : lfs setstripe, only systems with more than 160 OSTs
4614 Description: unable to set striping with a starting offset beyond OST 160
4615 Details : llapi_create_file() incorrectly limited the starting stripe
4616 index to the maximum single-file stripe count.
4619 Frequency : LDAP users only
4621 Description: lconf did not handle in-kernel recovery with LDAP properly
4622 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4624 Severity : enhancement
4626 Description: bind OST threads to NUMA nodes to improve performance
4627 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4628 node and do their allocations there to localize memory access
4630 Severity : enhancement
4632 Description: llmount can determine client NID directly from Myrinet (GM)
4633 Details : the client NID code from gmnalnid was moved directly into
4634 llmount, removing the need to use this or specifying the
4635 client NID explicitly when mounting GM clients with zeroconf
4638 Frequency : if client is started with down MDS
4640 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4641 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4642 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4643 ptlrpc_import_delayed_req() because we didn't check if the
4644 request was interrupted, and we also didn't break out of the
4645 event loop if there was no timeout
4650 Description: data loss during non-page-aligned writes to a single file from
4651 both multiple nodes and multiple threads on one node at same time
4652 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4653 inconsistency led to false short-reads, that were cached and later
4654 used by ->prepare_write() to fill in partially written page,
4655 leading to data loss.
4658 Frequency : always, if lconf --abort_recovery used
4660 Description: lconf --abort_recovery fails with 'Operation not supported'
4661 Details : lconf was attempting to abort recovery on the MDT device and not
4664 Severity : enhancement
4666 Description: remove cleanup logs
4667 Details : replace lconf-generated cleanup logs with lustre internal
4668 cleanup routines. Eliminates the need for client-cleanup and
4671 Severity : enhancement
4673 Description: add support for EAs (user and system) on lustre filesystems
4674 Details : it is now possible to store extended attributes in the Lustre
4675 client filesystem, and with the user_xattr mount option it
4676 is possible to allow users to store EAs on their files also
4678 Severity : enhancement
4680 Description: Add possibility (config option) to show minimal available OST free
4682 Details : When compiled with --enable-mindf configure option, statfs(2)
4683 (and so, df) will return least minimal free space available from
4684 all OSTs as amount of free space on FS, instead of summary of
4685 free spaces of all OSTs.
4687 Severity : enhancement
4689 Description: do not expand extent locks acquired on OST-side
4690 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4691 by server: they are not cached anyway.
4694 Frequency : when mmap is used/binaries executed from Lustre
4696 Description: Unmmap pages before throwing them away from read cache.
4697 Details : llap_shrink cache now attempts to unmap pages before discarding
4698 them (if unmapping failed - do not discard). SLES9 kernel has
4699 extra checks that trigger if this unmapping is not done first.
4704 Description: lconf didn't resolve symlinks before checking to see whether a
4705 given mountpoint was already in use
4708 Frequency : when migrating failover services
4709 Bugzilla : 6395, 9514
4710 Description: When migrating a subset of services from a node (e.g. failback
4711 from a failover service node) the remaining services would
4712 time out and evict clients.
4713 Details : lconf --force (implied by --failover) sets the global obd_timeout
4714 to 5 seconds in order to quickly disconnect, but this caused
4715 other RPCs to time out too quickly. Do not change the global
4716 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4718 Severity : enhancement
4719 Frequency : if MDS is started with down OST
4720 Bugzilla : 9439,5706
4721 Description: Allow startup/shutdown of an MDS without depending on the
4722 availability of the OSTs.
4723 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4724 Add appropriate locking and lov-osc refcounts for safe
4725 cleaning. Add osc abort_inflight calls in case the
4726 synchronize never started.
4729 Frequency : occasional (Cray XT3 only)
4731 Description: root not authorized to access files in CRAY_PORTALS environment
4732 Details : The client process capabilities were not honoured on the MDS in
4733 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4734 been accessed by an authorized user then root was able to access
4735 the file on the local client also. The root user capabilities
4736 are now allowed on the MDS, as this environment has secure UID.
4739 Frequency : occasional
4741 Description: ldiskfs "too long searching" message happens too often
4742 Details : A debugging message (otherwise harmless) prints too often on
4743 the OST console. This has been reduced to only happen when
4744 there are fragmentation problems on the filesystem.
4749 Description: Division by zero in statfs when all OSCs are inactive
4750 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4751 lov_statfs divides by value returned by lov_get_stripecnt().
4755 Bugzilla : 9489, 3273
4756 Description: First write from each client to each OST was only 4kB in size,
4757 to initialize client writeback cache, which caused sub-optimal
4758 RPCs and poor layout on disk for the first written file.
4759 Details : Clients now request an initial cache grant at (re)connect time
4760 and so that they can start streaming writes to the cache right
4761 away and always do full-sized RPCs if there is enough data.
4762 If the OST is rebooted the client also re-establishes its grant
4763 so that client cached writes will be honoured under the grant.
4768 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4769 Details : Now I/O RPCs go to different portal number and (presumably) fast
4770 lock requests (and glimses) and other RPCs get their own service
4771 threads pool that should be able to service those RPCs
4774 Severity : enhancement
4776 Description: Ability to exchange lustre version between client and servers and
4777 issue warnings at client side if client is too old. Also for
4778 liblustre clients there is ability to refuse connection of too old
4780 Details : New 'version' field is added to connect data structure that is
4781 filled with version info. That info is later checked by server and
4785 Frequency : rare, liblustre only.
4786 Bugzilla : 9296, 9581
4787 Description: Two simultaneous writes from liblustre at offset within same page
4788 might proceed at the same time overwriting eachother with stale
4790 Details : I/O lock withing llu_file_prwv was released too early, before data
4791 actually was hitting the wire. Extended lock-holding time until
4792 server acknowledges receiving data.
4795 Frequency : extremely rare. Never observed in practice.
4797 Description: avoid generating lustre_handle cookie of 0.
4798 Details : class_handle_hash() generates handle cookies by incrementing
4799 global counter, and can hit 0 occasionaly (this is unlikely, but
4800 not impossible, because initial value of cookie counter is
4801 selected randonly). Value of 0 is used as a sentinel meaning
4802 "unassigned handle" --- avoid it. Also coalesce two critical
4803 sections in this function into one.
4805 Severity : enhancement
4807 Description: allow liblustre clients to delegate truncate locking to OST
4808 Details : To avoid overhead of locking, liblustre client instructs OST to
4809 take extent lock in ost_punch() on client's behalf. New connection
4810 flag is added to handle backward compatibility.
4812 Severity : enhancement
4813 Bugzilla : 4928, 7341, 9758
4814 Description: allow number of OST service threads to be specified
4815 Details : a module parameter allows the number of OST service threads
4816 to be specified via "options ost ost_num_threads={N}" in the
4817 OSS's /etc/modules.conf or /etc/modprobe.conf.
4821 Bugzilla : 6146, 9635, 9895
4822 Description: servers crash with bad pointer in target_handle_connect()
4823 Details : In rare cases when a client is reconnecting it was possible that
4824 the connection request was the last reference for that export.
4825 We would temporarily drop the export reference and get a new
4826 one, but this may have been the last reference and the export
4827 was just destroyed. Get new reference before dropping old one.
4829 Severity : enhancement
4830 Frequency : if client is started with failover MDS
4832 Description: Allow multiple MDS hostnames in the mount command
4833 Details : Try to read the configuration from all specified MDS
4834 hostnames during a client mount in case the "primary"
4837 Severity : enhancement
4839 Description: Stop sending data to evicted clients as soon as possible.
4840 Details : Check if the client we are about to send or are sending data to
4841 was evicted already. (Check is done every second of waiting,
4842 for which l_wait_event interface was extended to allow checking
4843 of exit condition at specified intervals).
4846 Frequency : rare, normally only when NFS exporting is done from client
4848 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4850 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4853 Frequency : always, if srand() is called before liblustre initialization
4855 Description: Liblustre uses system PRNG disturbing its usage by user application
4856 Details : Introduce internal to lustre fast and high-quality PRNG for
4857 lustre usage and make liblustre and some other places in generic
4858 lustre code to use it.
4860 Severity : enhancement
4861 Bugzilla : 9477, 9557, 9870
4862 Description: Verify that the MDS configuration logs are updated when xml is
4863 Details : Check if the .xml configuration logs are newer than the config
4864 logs stored on the MDS and report an error if this is the case.
4865 Request --write-conf, or allow starting with --old_conf.
4867 Severity : enhancement
4869 Description: Handle symlinks in the path when checking if Lustre is mounted.
4870 Details : Resolve intermediate symlinks when checking if a client has
4871 mounted a filesystem to avoid duplicate client mounts.
4876 Description: lconf can hit an error exception but still return success.
4877 Details : The lconf command catches the Command error exception at the top
4878 level script context and will exit with the associated exit
4879 status, but doesn't ensure that this exit status is non-zero.
4884 Description: failure of ptlrpc thread startup can cause oops
4885 Details : Starting a ptlrpc service thread can fail if there are a large
4886 number of threads or the server memory is very fragmented.
4887 Handle this without oopsing.
4890 Frequency : always, only if liblustre and non-default acceptor port was used
4892 Description: liblustre cannot connect to servers with non-default acceptor port
4893 Details : tcpnal_set_default_params() was not called and was therefore
4894 ignoring the environment varaible TCPNAL_PORT, as well as other
4895 TCPNAL_ environment variables
4900 Description: two objects could be created on the same OST for a single file
4901 Details : If an OST is down, in some cases it was possible to create two
4902 objects on a single OST for a single file. No problems other
4903 than potential performance impact and spurious error messages.
4907 Bugzilla : 5681, 9562
4908 Description: Client may oops in ll_unhash_aliases
4909 Details : Client dcache may become inconsistent in race condition.
4910 In some cases "getcwd" can fail if the current directory is
4916 Description: Inode refcounting problems in NFS export code
4917 Details : link_raw functions used to call d_instantiate without obtaining
4918 extra inode reference first.
4922 Bugzilla : 9942, 9903
4923 Description: Referencing freed requests leading to crash, memleaks with NFS.
4924 Details : We used to require that call to ll_revalidate_it was always
4925 followed by ll_lookup_it. Also with revalidate_special() it is
4926 possible to call ll_revalidate_it() twice for the same dentry
4927 even if first occurence returned success. This fix changes semantic
4928 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4929 reference on a request referred from the intent.
4930 ll_intent_release() then releases such a request.
4933 Frequency : rare, normally benchmark loads only
4935 Description: unlinked inodes were kept in memory on the client
4936 Details : If a client is repeatedly creating and unlinking files it
4937 can accumulate a lot of stale inodes in the inode slab cache.
4938 If there is no other client load running this can cause the
4939 client node to run out of memory. Instead flush old inodes
4940 from client cache that have the same inode number as a new inode.
4943 Frequency : SLES9 2.6.5 kernel and long filenames only
4944 Bugzilla : 9969, 10379
4945 Description: utime reports stale NFS file handle
4946 Details : SLES9 uses out-of-dentry names in some cases, which confused
4947 the lustre dentry revalidation. Change it to always use the
4951 Frequency : rare, unless heavy write-truncate concurrency is continuous
4952 Bugzilla : 4180, 6984, 7171, 9963, 9331
4953 Description: OST becomes very slow and/or deadlocked during object unlink
4954 Details : filter_destroy() was holding onto the parent directory lock
4955 while truncating+unlinking objects. For very large objects this
4956 may block other threads for a long time and slow overall OST
4957 responsiveness. It may also be possible to get a lock ordering
4958 deadlock in this case, or run out of journal credits because of
4959 the combined truncate+unlink. Solution is to do object truncate
4960 first in one transaction without parent lock, and then do the
4961 final unlink in a new transaction with the parent lock. This
4962 reduces the lock hold time dramatically.
4965 Frequency : rare, 2.4 kernels only
4967 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4968 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4969 before unmounting their filesystem in order to avoid deadlock.
4970 The kernel_locked() function in 2.4 kernels only checks whether
4971 the kernel lock is held, not whether it is this process that is
4972 holding it as 2.6 kernels do.
4977 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4978 Details : The client ptlrpc code may be trying to reconnect to a down
4979 server before a previous connection attempt has timed out.
4980 Increase the reconnect interval to be longer than the connection
4981 timeout interval to avoid sending duplicate connections to
4985 Frequency : echo_client brw_test command
4987 Description: fix echo_client to work with OST preallocated code
4988 Details : OST preallocation code (5137) didn't take echo_client IO path
4989 into account: echo_client calls filter methods outside of any
4990 OST thread and, hence, there is no per-thread preallocated
4991 pages and buffers to use. Solution: hijack pga pages for IO. As
4992 a byproduct, this avoids unnecessary data copying.
4996 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4997 Description: Client can oops in mdc_commit_close() after open replay
4998 Details : It was possible for the MDS to return an open request with no
4999 transaction number in mds_finish_transno() if the client was
5000 evicted, but without actually returning an error. Clients
5001 would later try to replay that open and may trip an assertion
5002 Simplify the client close codepath, and always return an error
5003 from the MDS in case the open is not successful.
5006 Frequency : rare, 2.6 OSTs only
5008 Description: OST may deadlock under high load on fragmented files
5009 Details : If there was a heavy load and highly-fragmented OST filesystems
5010 it was possible to have all the OST threads deadlock waiting on
5011 allocation of biovecs, because the biovecs were not released
5012 until the entire RPC IO was completed. Instead, release biovecs
5013 as soon as they are complete to ensure forward IO progress.
5015 Severity : enhancement
5017 Description: Support for specifying external journal device at mount
5018 Details : If an OST or MDS device is formatted with an external journal
5019 device, this device major/minor is stored in the ext3 superblock
5020 and may not be valid for failover. Allow detecting and
5021 specifying the external journal at mount time.
5026 Description: Mounting an MDS with pending unlinked files may cause oops
5027 Details : target_finish_recovery() calls mds_postrecov() which returned
5028 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5029 considers this an error and immediately begins cleaning up the
5030 lov, just after starting the mds_lov process
5032 Severity : enhancement
5034 Description: Implement 'lfs df' to report actual free space on per-OST basis
5035 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5036 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5037 usage of objects; '-h' to report in human readable format.
5039 ------------------------------------------------------------------------------
5041 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5048 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5049 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5050 at startup during orphan unlinking. Instead, skip buddy update
5051 before setup, as it will be regenerated after recovery is complete.
5056 Description: If an OST is inactive, its locks might reference stale inodes.
5057 Details : lov_change_cbdata() must iterate over all namespaces, even if
5058 they are inactive to clear inode references from the lock.
5060 Severity : enhancement
5061 Frequency : occasional, if non-standard max_dirty_mb used
5063 Description: Client will block write RPCs if not enough grant
5064 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5065 then the client will block writes while waiting for another RPC
5066 to complete instead of consuming its dirty limit. With change
5067 we get improved performance when max_dirty_mb is small.
5069 Severity : enhancement
5070 Bugzilla : 3389, 6253
5071 Description: Add support for supplementary groups on the MDS.
5072 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5073 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5074 lookups for user supplementary groups into a cache.
5078 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5079 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5080 for liblustre O_EXCL is enforced by MDS.
5084 Description: readdir never returns NULL in liblustre.
5085 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5086 dentry in struct dirent.
5090 Frequency : liblustre only, depends on application IO pattern
5091 Description: liblustre clients evicted if not contacting servers
5092 Details : Don't put liblustre clients into the ping_evictor list, so
5093 they will not be evicted by the pinger ever.
5095 Severity : enhancement
5097 Description: Add ability to evict clients by NID from MDS.
5098 Details : By echoing "nid:$NID" string into
5099 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5100 $NID would be instantly evicted from this MDS and from all active
5101 OSTs connected to it.
5105 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5106 Details : lookup_it_finish() used to query file size from OSTs that was not
5111 Description: service threads change working directory to that of init
5112 Details : Starting lustre service threads may pin the working directory
5113 of the parent thread, making that filesystem busy. Threads
5114 now change to the working directory of init to avoid this.
5118 Frequency : during shutdown only
5119 Description: shutdown with a failed MDS or OST can cause unmount to hang
5120 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5125 Frequency : 2.6 only
5126 Description: chmod/chown may include an extra supplementary group
5127 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5128 supplementary group and if none is specified this is used.
5131 Bugzilla : 5479 (6816)
5133 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5134 Details : If lookup is in progress on a file that is unlinked we might try
5135 to revalidate the inode and fail in revalidate after lookup is
5136 complete and ll_file_open() enqueues the open again but
5137 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5140 Frequency : always, if lconf --abort_recovery used
5142 Description: lconf --abort_recovery fails with 'Operation not supported'
5143 Details : lconf was attempting to abort recovery on the MDT device and not
5146 ------------------------------------------------------------------------------
5148 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5153 Frequency : rare (only unsupported configurations with a node running as an
5155 Bugzilla : 6514, 5137
5156 Description: Mounting a Lustre file system on a node running as an OST could
5158 Details : OSTs now preallocates memory needed to write out data at
5159 startup, instead of when needed, to avoid having to
5160 allocate memory in possibly low memory situations.
5161 Specifically, if the file system is mounted on on OST,
5162 memory pressure could force it to try to write out data,
5163 which it needed to allocate memory to do. Due to the low
5164 memory, it would be unable to do so and the node would
5165 become unresponsive.
5167 Severity : enhancement
5169 Description: Addition of lconf --service command line option
5170 Details : lconf now accepts a '--service <arg>' option, which is
5171 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5173 Severity : enhancement
5175 Description: Failover mode is now the default for OSTs.
5176 Details : By default, OSTs will now run in failover mode. To return to
5177 the old behaviour, add '--failout' to the lmc line for OSTs.
5179 Severity : enhancement
5181 Description: Health checks are now provided for MDS and OSTs
5182 Details : Additional detailed health check information on MSD and OSTs
5183 is now provided through the procfs health_check value.
5186 Frequency : occasional, depends on IO load
5188 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5189 after numerous create/delete cycles
5190 Details : The ext3 inode allocation policy would not allocate new inodes
5191 very well on the OSTs because there are no new directories
5192 being created. Instead we look for groups with free space if
5193 the parent directories are nearly full.
5198 Description: Network or server problems during mount may cause partially
5199 mounted clients instead of returning an error.
5200 Details : The config llog parsing code may overwrite the error return
5201 code during mount error handling, returning success instead
5207 Description: MDS can fail to allocate large reply buffers
5208 Details : After long uptimes the MDS can fail to allocate large reply
5209 buffers (e.g. zconf client mount config records) due to memory
5210 fragmentation or consumption by the buffer cache. Preallocate
5211 some large reply buffers so that these replies can be sent even
5212 under memory pressure.
5216 Frequency : rare (liblustre)
5217 Description: fsx running with liblustre complained that using truncate() to
5218 extend the file doesn't work. This patch corrects that issue.
5219 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5220 ATTR_SIZE and lsm use in llu_setattr_raw.
5224 Frequency : rare, only 2.6 kernels
5225 Description: Unusual file access patterns on the MDS may result in inode
5226 data being lost in very rare circumstances.
5227 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5228 code in the 2.6 kernel caused the inode and/or EA data not to
5229 be read from disk, causing single-file corruption.
5233 Frequency : rare, only 2.6 filesystems using extents
5234 Description: Heavy concurrent write and delete load may cause data corruption.
5235 Details : It was possible under high-load situations to have an extent
5236 metadata block in the block device cache from a just-unlinked
5237 file overwrite a newly-allocated data block. We now unmap any
5238 metadata buffers that alias just-allocated data blocks.
5242 Frequency : filesystems with default stripe_count larger than 77
5243 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5244 Details : lconf specifies an inode size of 4096 bytes when the default
5245 stripe_count is larger than 77. This conflicts with the default
5246 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5247 this case to avoid pinning too much memory for large EAs.
5249 ------------------------------------------------------------------------------
5251 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5256 Frequency : rare (extremely heavy IO load with hundreds of clients)
5258 Description: Client is evicted, gets IO error writing to file
5259 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5260 caused clients to be evicted for AST timeouts. The fixes in
5261 bug 5192 mean we no longer need to have such short AST timeouts
5262 so ldlm_timeout has been increased.
5265 Frequency : occasional during --force or --failover shutdown under load
5266 Bugzilla : 5949, 4834
5267 Description: Server oops/LBUG if stopped with --force or --failover under load
5268 Details : a collection of import/export refcount and cleanup ordering
5269 issues fixed for safer force cleanup
5272 Frequency : only filesystems larger than 120 OSTs
5273 Bugzilla : 5990, 6223
5274 Description: lfs getstripe would oops on a very large filesystem
5275 Details : lov_getconfig used kfree on vmalloc'd memory
5278 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5279 Bugzilla : 6242, 6243
5280 Description: reading from files that had been truncated to a non-zero size
5281 but never opened returned no data
5282 Details : ll_file_read() reads zeros from no-object files to EOF
5287 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5288 Details : There's little harm in aborting MDS/OSS recovery and letting it
5289 try again, so I removed the LASSERT and return an error instead.
5291 Severity : enhancement
5293 Description: New debugging infrastructure for tracking down data corruption
5294 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5295 (b) cover more of the client-side code path, and (c) try to narrow
5296 down where problems occurred
5300 Bugzilla : 3819, 4364, 4397, 6313
5301 Description: Racing close and eviction MDS could cause assertion in mds_close
5302 Details : It was possible to get multiple mfd references during close and
5303 client eviction, leading to one thread referencing a freed mfd.
5305 Severity: : enhancement
5306 Bugzilla : 3262, 6359
5307 Description: Attempts to reconnect to servers are now more aggressive.
5308 Details : This builds on the enhanced upcall-less recovery that was added
5309 in 1.4.2. When trying to reconnect to servers, clients will
5310 now try each server in the failover group every 10 seconds. By
5311 default, clients would previously try one server every 25 seconds.
5316 Description: After recovery, certain operations trigger a failed
5317 assertion on a client.
5318 Details : Failing over an mds, using lconf -d --failover, while a
5319 client was doing a readdir() call would cause the client to
5320 LBUG after recovery completed and the readdir() was resent.
5322 Severity : enhancement
5324 Description: Default groups are now added by lconf
5325 Details : You can now run lconf --group <servicename> without having to
5326 manually add groups with lmc.
5329 Frequency : occasional
5331 Description: Nodes with an elan id of 0 trigger a failed assertion
5334 Frequency : always when accessing e.g. tty/console device nodes
5336 Description: tty and some other devices nodes cannot be used on lustre
5337 Details : file's private_data field is used by device data and lustre
5338 values in there got lost. New field was added to struct file to
5339 store fs-specific private data.
5342 Frequency : when exporting Lustre via NFS
5344 Description: NFSD failed occasionally when looking up a path component
5345 Details : NFSD is looking up ".." which was broken in ext3 directories
5346 that had grown large enough to become hashed.
5349 Frequency : Clusters with multiple interfaces not on the same subnet
5351 Description: Nodes will repeatedly try to reconnect to an interface which it
5352 cannot reach and report an error to the log.
5353 Details : Extra peer list entries will be created by lconf with some peers
5354 unreachable. lconf now validates the peer before adding it.
5357 Frequency : Only if a default stripe is set on the filesystem root.
5359 Description: Setting a default stripe on the filesystem root prevented the
5360 filesystem from being remounted.
5361 Details : The client was sending extra request flags in the root getattr
5362 request and did not allocate a reply buffer for the dir EA.
5365 Frequency : occasional, higher if lots of files are accessed by one client
5366 Bugzilla : 6159, 6097
5367 Description: Client trips assertion regarding lsm mismatch/magic
5368 Details : While revalidating inodes the VFS looks up inodes with ifind()
5369 and in rare cases can find an inode that is being freed.
5370 The ll_test_inode() code will free the lsm during ifind()
5371 when it finds an existing inode and then the VFS later attaches
5372 this free lsm to a new inode.
5376 Bugzilla : 6422, 7030
5377 Description: MDS deadlock between mkdir and client eviction
5378 Details : Creating a new file via mkdir or mknod (starting a transaction
5379 and getting the ns lock) can deadlock with client eviction
5380 (gets ns lock and trying to finish a synchronous transaction).
5383 Frequency : occasional
5384 Description: While starting a server, the fsfilt_ext3 module could not be
5386 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5387 kernels. Previously, lconf would still use the ext3 name
5388 when trying to load modules. Now, it will correctly use
5389 ext3 on 2.4 and ldiskfs on 2.6.
5391 Severity : enhancement
5392 Description: The default stripe count has been changed to 1
5393 Details : The interpretation of the default stripe count (0, to lfs
5394 or lmc) has been changed to mean striping across a single
5395 OST, rather than all available. For general usage we have
5396 found a stripe count of 1 or 2 works best.
5398 Severity : enhancement
5399 Description: Add support for compiling against Cray portals.
5400 Details : Conditional compiling for some areas that are different
5404 Frequency : occasional
5405 Bugzilla : 6409, 6834
5406 Description: Creating files with an explicit stripe count may lead to
5407 a failed assertion on the MDS
5408 Details : If some OSTs are full or unavailable, creating files may
5409 trigger a failed assertion on the MDS. Now, Lustre will
5410 try to use other servers or return an error to the
5414 Frequency : occasional
5416 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5417 caused lock timeout during readahead (since 1.4.2).
5418 Details : Processes doing readahead might match a lock that hasn't been
5419 granted yet if there are overlapping and conflicting lock
5420 requests. The readahead process waits on ungranted lock
5421 (original lock is CBPENDING), while OST waits for that process
5422 to cancel CBPENDING read lock and eventually evicts client.
5424 Severity : enhancement
5426 Description: Initial enabling of flock support for clients
5427 Details : Implements fcntl advisory locking and file status functions.
5428 This feature is provided as an optional mount flag (default
5429 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5430 locking are implemented yet, and those that are are not guaranteed
5431 to be completely correct in production environments.
5432 mount -t lustre -o [flock|noflock] ...
5435 Frequency : occasional
5437 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5438 assertion in the ext3 JBD (journaling) layer.
5439 Details : The b_committed_data struct is protected by the big kernel lock
5440 in 2.4 kernels, serializing journal_commit_transaction() and
5441 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5442 access to this struct. In 2.6 kernels there is finer grained
5443 locking to improve SMP performance of the JBD layer.
5447 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5452 Description: merge and cleanup kernel patches.
5455 -----------------------------------------------------------------------------
5457 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5459 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5460 versions of Lustre. Please update all servers and clients to
5461 version 1.4.2 or later at the same time. You must also run
5462 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5463 to update the configuration logs.
5465 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5466 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5467 - swab llog records of type '0' so we get proper header size/idx (5861)
5468 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5469 - fix rename of one directory over another leaking an inode (5953)
5470 - avoid SetPageDirty on 2.6 (5981)
5471 - don't re-add just-being-destroyed locks to the waiting list (5653)
5472 - when creating new directories, inherit the parent's custom
5473 striping settings if present parent (3048)
5474 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5475 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5476 - don't hold client locks on temporary worklist from l_lru (5666)
5477 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5478 - automatically evict dead clients (5921)
5479 - Update file size properly in create+truncate+fstat case (6196)
5480 - Do not unhash mountpoint dentries, do not allow removal of
5482 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5483 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5484 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5485 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5486 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5487 - Correctly handle reads of files with no objects (6243)
5488 - lctl recover will also mark a device active if deactivate used (5933)
5490 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5491 - allow --write-conf on an MDS with different nettype than client (5619)
5492 - don't write config llogs to MDS for mounts not from that MDS (5617)
5493 - lconf should create multiple TCP connections from a client (5201)
5494 - init scripts are now turned off by default; run chkconfig --on
5495 lustre and chkconfig --on lustrefs to use them
5496 - upcalls are no longer needed for clients to recover to failover
5498 - add --abort-recovery option to lconf to abort recovery on device
5500 - add support for an arbitrary number of OSTs (3026)
5501 - Quota support protocol changes.
5502 - forward compatibility changes to wire structs (6007)
5503 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5504 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5505 - improved "lustre status" script
5506 - initialize blocksize for non-regular files (6062)
5507 - added --disable-server and --disable-client configure options (5782)
5508 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5509 - Vanilla 2.4.29 support
5510 - increase maximum number of obd devices to 520 (6242)
5511 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5512 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5513 SLES 9 2.6.5 kernels
5514 - limit stripes per file to 160 (the maximum EA size) (6093)
5516 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5519 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5520 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5521 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5522 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5523 - small ext3 extents cleanups and fixes (5733)
5524 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5525 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5526 - avoid needless client->OST connect, fix handle mismatch (5317)
5527 - fix DLM error path that led to out-of-sync client, long delays (5779)
5528 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5529 - fix several locking issues related to i_size (5492,5624,5654,5672)
5530 - don't move pending lock onto export if it is already evicted (5683)
5531 - fix kernel oops when creating .foo in unlinked directory (5548)
5532 - fix deadlock in obdfilter statistics vs. object create (5811)
5533 - use time_{before,after} to avoid timer jiffies wrap (5882)
5534 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5535 - Do not leak request if server was not able to process it (5154)
5536 - If mds_open unable to find parent dir, make that negative lookup(5154)
5537 - don't create new directories with extent-mapping (5909, 5936)
5539 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5540 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5541 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5542 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5543 - support for RHEL4 kernel on i686 (5773)
5544 - provide error messages when incompatible logs are encountered (5898)
5546 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5547 * version 1.4.0.10 (1.4.1 release candidate 1)
5549 - don't keep a lock reference when lock is not granted (4238)
5550 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5551 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5552 - drop import inflight refcount on signal_completed_replay error (5255)
5553 - unlock page after async write error during send (3677)
5554 - handle missing objects in filter_preprw_read properly (5265)
5555 - no transno return for symlink open, don't save no-trasno open (3440)
5556 - don't try to complete elan receive that already failed (4012)
5557 - free RPC server reply state on error (5406)
5558 - clean up thread from ptlrpc_start_thread() on error (5160)
5559 - readahead could read extra page into cache that wasn't ejected (5388)
5560 - prevent races in class_attach/setup/cleanup/detach (5260)
5561 - don't dereference de->d_inode after l_dput of de (5458)
5562 - use "int" for stripe value returned from lock_to_stripe (5544)
5563 - mballoc allocation and error-checking fixes in 2.6 (5504)
5564 - block device patches to fix I/O request sizes in 2.6 (5482)
5565 - look up hostnames for IB nals (5602)
5566 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5567 - don't start multiple acceptors for the same port (5277)
5568 - fix incorrect LASSERT in mds_getattr_name (5635)
5569 - export a proc file for general "ping" checking (5628)
5570 - fix "lfs check" to not block when the MDS is down (5628)
5572 - service request history (4965)
5573 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5574 - create an "evict_client" /proc entry on OSTs, like the MDS has
5575 - fix mount usage message, return errors per mount(8) (5168)
5576 - change grep [] to grep "[]" in tests so they work in more UMLs
5577 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5578 - remove ancient LOV_MAGIC_V0 EA support (5047)
5579 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5580 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5581 - allow readahead allocations to fail when low on memory (5383)
5582 - mmap locking landed again, after considerable improvement (2828)
5583 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5585 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5588 - send OST transaction number in read/write reply to free req (4966)
5589 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5590 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5591 - the watchdog thread now runs as interruptible (5246)
5592 - flock/lockf fixes (but it's still disabled, pending 5135)
5593 - don't use EXT3 constants in llite code (5094)
5594 - memory shortage at startup could cause assertion (5176)
5596 - reorganization of lov code
5597 - single portals codebase
5599 - add extents/mballoc support (5025)
5600 - direct I/O reads in the obdfilter (4048)
5601 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5603 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5606 - send OST transaction number in read/write reply to free req (4966)
5607 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5608 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5609 - the watchdog thread now runs as interruptible (5246)
5610 - handle missing objects in filter_preprw_read properly (5265)
5611 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5612 - drop import inflight refcount on signal_completed_replay error (5255)
5613 - unlock page after async write error during send (3677)
5614 - return original error code on reconstructed replies (3761)
5615 - no transno return for symlink open, don't save no-trasno open (3440)
5617 - add pid to ldlm debugging output (4922)
5618 - bump the watchdog timeouts -- we can't handle 30sec yet
5619 - extra debugging for orphan dentry/inode bug (5259)
5621 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5624 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5625 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5626 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5627 - fix NULL dereference in /proc/sys/portals/routes (4827)
5628 - allow failed mdc_close() operations to be interrupted (4561)
5629 - stop precreate on OST before MDS would time out on it (4778)
5630 - don't send partial-page writes before EOF from client (4410)
5631 - discard client grant for sub-page writes on large-page clients (4520)
5632 - don't free dentries not owned by NFS code, check generation (4806)
5633 - fix lsm leak if mds_create_objects() fails (4801)
5634 - limit debug_daemon file size, always print CERROR messages (4789)
5635 - use transno after validating reply (3892)
5636 - process timed out requests if import state changes (3754)
5637 - update mtime on OST during writes, return in glimpse (4829)
5638 - add mkfsoptions to LDAP (4679)
5639 - use ->max_readahead method instead of zapping global ra (5039)
5640 - don't interrupt __l_wait_event() during strace
5642 - add software watchdogs to catch hung threads quickly (4941)
5643 - make lustrefs init script start after nfs is mounted
5644 - fix CWARN/ERROR duplication (4930)
5645 - return async write errors to application if possible (2248)
5646 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5647 - print NAL number in %x format (4645)
5648 - update barely-supported suse-2.4.21-171 series (4842)
5649 - support for sles 9 %post scripts
5650 - support for building 2.6 kernel-source packages
5651 - support for sles km_* packages
5653 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5656 - ignore -ENOENT errors in osc_destroy (3639)
5657 - notify osc create thread that OSC is being cleaned up (4600)
5658 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5659 - reconstruct ost_handle() like mds_handle() (4657)
5660 - create a new thread to do import eviction to avoid deadlock (3969)
5661 - let lconf resolve symlinked-to devices (4629)
5662 - don't unlink "objects" from directory with default EA (4554)
5663 - hold socknal file ref over connect in case target is down (4394)
5664 - allow more than 32000 subdirectories in a single directory (3244)
5665 - fix blocks count for O_DIRECT writes (3751)
5666 - OST returns ENOSPC from object create when no space left (4539)
5667 - don't send truncate RPC if file size isn't changing (4410)
5668 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5669 - bind to privileged port in socknal and tcpnal (3689)
5671 - rate limit CERROR/CWARN console message to avoid overload (4519)
5672 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5673 - basic mmap support (3918)
5674 - kernel patch series update from b1_4 (4711)
5676 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5679 - avoid crash during MDS cleanup with OST shut down (2775)
5680 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5681 - don't use bad inodes on the MDS (3744)
5682 - dynamic object preallocation to improve recovery speed (4236)
5683 - don't hold spinlock over lock dumping or change debug flags (4401)
5684 - don't zero obd_dev when it is force cleaned (3651)
5685 - print grants to console if they go negative (4431)
5686 - "lctl deactivate" will stop automatic recovery attempts (3406)
5687 - look for existing locks in ldlm_handle_enqueue() (3764)
5688 - don't resolve lock handle twice in recovery avoiding race (4401)
5689 - revalidate should check working dir is a directory (4134)
5691 - don't always mark "slow" obdfilter messages as errors (4418)
5693 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5696 - don't close LustreDB during write_conf until it is done (3860)
5697 - fix typo in lconf for_each_profile (3821)
5698 - allow dumping logs from multiple threads at one time (3820)
5699 - don't allow multiple threads in OSC recovery (3812)
5700 - fix debug_size parameters (3864)
5701 - fix mds_postrecov to initialize import for llog ctxt (3121)
5702 - replace config semaphore with spinlock (3306)
5703 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5704 - don't allow enqueue to complete on a destroyed export (3822)
5705 - down write_lock before checking llog header bitmap (3825)
5706 - recover from lock replay timeout (3764)
5707 - up llog sem before sending rpc (3652)
5708 - reduce ns lock hold times when setting kms (3267)
5709 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5710 - fix NULL deref and obd_dev leak on setup error (3312)
5711 - replace some LBUG about llog ops with error handling (3841)
5712 - don't match INVALID dentries from d_lookup and spin (3784)
5713 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5714 - fix invalid assertion in ptlrpc_set_wait (3880)
5716 - add libwrap support for the TCP acceptor (3996)
5717 - add /proc/sys/portals/routes for non-root route listing (3994)
5718 - allow setting MDS UUID in .xml (2580)
5719 - print the stack of a process that LBUGs (4228)
5721 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5724 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5725 - make sure to unset replay flag from failed open requests (3440)
5726 - if default stripe count is 0, use OST count for inode size (3636)
5727 - update parent mtime/ctime on client for create/unlink (2611)
5728 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5729 - free recovery state on server during a forced cleanup (3571)
5730 - unregister_reply for resent reqs (3063)
5731 - loop back devices mounting and status check on 2.6 (3563)
5732 - fix resource-creation race that can provoke i_size == 0 (3513)
5733 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5734 - more debugging for page-accounting assertion (3746)
5735 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5736 - avoid deadlock after precreation failure (3758)
5737 - fix race and lock order deadlock in orphan handling (3450, 3750)
5738 - add validity checks when grabbing inodes from l_ast_data (3599)
5740 - add /proc/.../recovery_status to obdfilter (3428)
5741 - lightweight CDEBUG infrastructure, debug daemon (3668)
5742 - change default OSC RPC parameters to be better on small clusters
5743 - turn off OST read cache for files smaller than 32MB
5744 - install man pages and include them in rpms (3100)
5745 - add new init script for (un)mounting lustre filesystems (2593)
5746 - run chkconfig in %post for init scripts (3701)
5747 - drop scimac NAL (unmaintained)
5749 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5752 - clean kiobufs before and after use (3485)
5753 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5754 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5755 - change init script to fail more clearly if not run as root (1528)
5756 - allow clients to reconnect during replay (1742)
5757 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5758 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5759 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5760 - fix cond_resched (3554)
5761 - only evict unfinished clients after recovery (3515)
5762 - allow bulk resend, prevent data loss (3570)
5763 - dynamic ptlrpc request buffer allocation (2102)
5764 - don't allow unlinking open directory if it isn't empty (2904)
5765 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5766 - remove extraneous obd dereference causing LASSERT failure (3334)
5767 - don't use get_cycles() when creating temp. files on the mds (3156)
5768 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5769 - handle EEXIST for set-stripe, set proper directory name (3336)
5771 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5772 - fix ksocknal_fmb_callback() error messages (2918)
5774 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5777 - don't copy lvb into (possibly NULL) reply on error (2983)
5778 - don't deref dentry after dput, don't free lvb on error (2922)
5779 - use the kms to determine writeback rpc length (2947)
5780 - increment oti_logcookies when osc is inactive (2948)
5781 - update client's i_blocks count via lvb messages (2543)
5782 - handle intent open/close of special files properly (1557)
5783 - mount MDS with errors=remount-ro, like obdfilter (2009)
5784 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5785 - don't use cancelling-locks' kms values (2947)
5786 - use highest lock extent for kms, not last one (2925)
5787 - don't dereference ERR_PTR() dentry in error handling path (3107)
5788 - fix thread race in portals_debug_dumplog() (3122)
5789 - create lprocfs device entries at setup instead of at attach (1519)
5790 - common AST error handler, don't evict client on completion race (3145)
5791 - zero nameidata in detach_mnt in 2.6 (3118)
5792 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5793 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5794 - initialize RPC timeout timer earlier for 2.6 (3219)
5795 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5796 - print nal/nid for unknown nid (3258)
5797 - additional checks for oscc recovery before doing precreate (3284)
5798 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5799 - don't crash in mdc_close for bad permissions on open (3285)
5800 - zero i_rdev for non-device files (3147)
5801 - clear page->private before handing to FS, better assertion (3119)
5802 - tune the read pipeline (3236)
5803 - fix incorrect decref of invalidated dentry (2350)
5804 - provide read-ahead stats and refine rpc in flight stats (3328)
5805 - don't hold journal transaction open across create RPC (3313)
5806 - update atime on MDS at close time (3265)
5807 - close LDAP connection when recovering to avoid server load (3315)
5808 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5809 - don't leak open file on MDS after open resend (3325)
5810 - serialize filter_precreate and filter_destroy_precreated (3329)
5811 - loop device shouldn't call sync_dev() for nul device (3092)
5812 - clear page cache after eviction (2766)
5813 - resynchronize MDS->OST in background (2824)
5814 - refuse to mount the same filesystem twice on same mountpoint (3394)
5815 - allow llmount to create routes for mounting behind routers (3320)
5816 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5817 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5818 - fix rare problem with rename on htree directories (3417)
5820 - allow default OST striping configuration per directory (1414)
5821 - fix compilation for qswnal for 2.6 kernels (3125)
5822 - increase maximum number of MDS request buffers for large systems
5823 - change liblustreapi to be useful for external progs like lfsck (3098)
5824 - increase local configuration timeout for slow disks (3353)
5825 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5827 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5830 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5831 - don't overwrite extent policy data in reply if lock was blocked (2901)
5832 - drop filter export grants atomically with removal from device (2663)
5833 - del obd_self_export from work_list in class_disconnect_exports (2908)
5834 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5835 - swab reply message in mdc_close, other PPC fixes (2464)
5836 - fix destroying of named logs (2325)
5837 - overwrite old logs when running lconf --write_conf (2264)
5838 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5839 - fix race in target_handle_connect (2898)
5840 - mds_reint_create() should take same inode create lock (2926)
5841 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5842 - don't close files for self_export to avoid uninitialized obd (2936)
5843 - allow MDS with the same name as client node (2939)
5844 - hold dentry reference for closed log files for unlink (2325)
5845 - reserve space for all logs during transactions (2059)
5846 - don't evict page beyond end of stripe extent (2925)
5847 - don't oops on a deleted current working directory (2399)
5848 - handle hard links to targets without a parent properly (2517)
5849 - don't dereference NULL lock when racing during eviction (2867)
5850 - don't grow lock extents when lots of conflicting locks (2919)
5852 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5855 - account for cache space usage on clients to avoid data loss (974)
5856 - lfsck support in lustre kernel code (2349)
5857 - reduce journal credits needed for BRW writes (2370)
5858 - orphan handling to avoid losing space on client/server crashes
5859 - ptlrpcd can be blocked, stopping ALL progress (2477)
5860 - use lock value blocks to assist in proper KMS, faster stat (1021)
5861 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5862 - recovery for initial connections (2355)
5863 - fixes for mds_cleanup_orphans (1934)
5864 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5865 - block all file creations until orphan recovery completes (1901)
5866 - client remove rq_connection from request struct (2423)
5867 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5868 - recovery timer race (2670)
5869 - mdc_close recovey bug (2532)
5870 - ptlrpc cleanup bug (2710)
5871 - mds timeout on local locks (2588)
5872 - namespace lock held during RPCs (2431)
5873 - handle interrupted sync write properly (2503)
5874 - don't try to handle a message that hasn't been replied to (2699)
5875 - client assert failure during cleanup after abort recovery (2701)
5876 - leak mdc device after failed mount (2712)
5877 - ptlrpc_check_set allows timedout requests to complete (2714)
5878 - wait for inflight reqs when ptlrpcd finishes (2710)
5879 - make sure unregistered services are removed from the srv_list
5880 - reset bulk XID's when resending them (caught by 1138 test)
5881 - unregister_bulk after timeout
5882 - fix lconf error (2694)
5883 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5884 - readahead locks pages, leaves pending causing memory pressure (2673)
5885 - increase OST request buffers to 4096 on large machines (2729)
5886 - fix up permission of existing directories in simple_mkdir (2661)
5887 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5888 - don't assign transno to errored transactions (2742)
5889 - don't delete objects on OST if given a bogus objid from MDS (2751)
5890 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5891 - if rq_no_resend, then timeout request after recovery (2432)
5892 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5893 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5894 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5895 - MDS recovery shouldn't delete objects that it has given out (2730)
5896 - if enqueue arrives after completion, don't clobber LVB (2819)
5897 - don't unlock pages twice when trigger_group_io returns error (2814)
5898 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5899 - don't write pages to disk if there was an error (1450)
5900 - don't ping imports that have recovery disabled (2676)
5901 - take buffered bytes into account when balancing socknal conn (2817)
5902 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5903 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5904 - remove little-endian swabbing of llog records (1987)
5905 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5906 - retry reposting request buffers if they fail (1191)
5907 - grow extent at grant time to avoid granting a revoked lock (2809)
5908 - lock revoke doesn't evict page if covered by a second lock (2765)
5909 - disable VM readahead to avoid reading outside lock extents (2805)
5911 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5912 - updated kernel patches for hp-2.4.20 kernel (2681)
5914 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5917 - fix truncated write corruption (2366)
5918 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5920 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5921 - ensure len(uuid) < 37 in lmc (1171)
5922 - fix ia64 OOPS in llog_test (2255)
5923 - zero end of page at obdfilter for partial page writes (2648)
5924 - don't leave stale dentries around after renames (bug 2428)
5925 - fix timeouts when evicting a client with a single lock held (2642)
5926 - set deadline for the initial HELLO message to drain (2634)
5927 - print out dotted-quad IP addresses in the socknal (2302)
5929 - additional debugging for MDS client eviction problem (2443)
5930 - fix mkfsoptions support for osts (2603, 2604)
5932 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5935 - add series for the vanilla 2.6.0 kernel
5936 - add series for the vanilla 2.4.24 kernel
5937 - add series for a cray x86/64 UL kernel drop
5938 - fix xattr patches for the vanilla 2.4.19 series
5940 - generate true UUIDs in lmc (1171)
5941 - have portals stack dumping break in UML (2466)
5942 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5943 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5944 - wake callback waiting threads on client eviction (2460)
5945 - Add --ptldebug and --subsystem to lmc (1719)
5946 - update assertion to allow safe interrupt allocation
5947 - set rq_no_resend for cancel requests (2432)
5948 - recalculate ptlrpcd timeout after resend (2494)
5949 - call vfs_rmdir when removing pending directories (2368)
5950 - fix renaming a file to itself (2429)
5951 - lmc creates a default one-stripe lov (2454)
5952 - expand procfs space to handle large clusters (2326)
5953 - increase UML stack to avoid overflow
5954 - update lconf's list of debug and subsystem masks
5955 - fix lfs find --obd (2510)
5956 - /proc tunable for disabling filter read caching (2591)
5957 - stop rpm packages from altering slapd.conf (2301)
5958 - disable nagle in the socknal under 0conf (2578)
5959 - choose mds inode size based on stripe count (2572)
5960 - fix kernel-source rpm problems (2516)
5962 - add --disable-doc to avoid pdf generation (2421)
5963 - update documentation, tests, type-os, comments
5964 - avoid format warnings on ia64
5965 - remove the TOE NAL
5966 - tiny code cleanups by removing unused fields
5968 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5971 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5972 - remove the most blatant lies from BUILDING file (2371)
5973 - change default debug level to reasonable production setting
5974 - reduce client side cache size to reduce cache flush time
5975 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5976 - make TCP zerocopy and pinger support enabled by default (2476)
5977 - sync writes completed after process exits caused crashes (2319)
5978 - maintain correct mount count on the MDS (2356)
5979 - backout 1557, because 2316 wasn't really fixed
5980 - better file I/O statistics gathering in /proc
5981 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5982 - another kernel patch to fix zero-copy TCP function export
5983 - don't take duplicate lock when processing re-sent getattr (2420)
5984 - lctl uses obd_self_export instead of creating new conn (2353)
5985 - MDS/OST recovery case which requires object creation asserted (2425)
5986 - move lfs from /usr/sbin to /usr/bin in packages
5987 - fix race between mds_client_add and mds_client_free (2417)
5988 - use kmalloc instead of slabs in portals (2430)
5989 - don't create duplicate records when a failover MDS is present (2442)
5990 - remove unnecessary mount age check (2332)
5991 - don't remove directory inodes from locks prematurely (2451)
5992 - don't break if MDS service name is the same as hostname (2103)
5993 - fix races in client write RPC generation when cache full (2482)
5995 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5998 - remove now-unused request->rq_obd (278)
5999 - if an allocation fails, print out how much memory we've used (1933)
6000 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6001 - add the "configurable stack size" patch to most series files (1256)
6002 - ability to write large log records, for 100+ OST configs (2306)
6003 - fix NULL deref when filter_prep fails (2314)
6004 - fix operator precedence error in filter_sync
6005 - dynamic allocation of socknal TX descriptors (2315)
6006 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6007 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6008 - ensure that CWARN messages in Portals always get to the syslog
6009 - __init/__exit are not for prototype decls (ldlm_init/exit)
6010 - x86-64 compile warning fixes
6011 - fix gateway LMC keyword conflict (2318)
6012 - fix MDS lock inversions in getattr/reint paths (1844)
6013 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6014 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6015 - DLM race condition prevented some lock evictions (2328)
6016 - ENOMEM detection and retry on socknal sends (2230)
6017 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6018 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6019 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6020 - better error messages when a client is rejected during recovery (1505)
6021 - avoid cancelling locks which were never granted, after failure (2330)
6022 - fix i_sem/journal inversion in mds_client_add (2333)
6023 - fix truncate/getattr lock cycle deadlock (2334)
6024 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6025 - fix two rare exit paths which could leak an l_lock() ref (2321)
6026 - fencepost error in MDS/OST orphan recovery (2226)
6027 - make log record alignment 8 bytes (1988)
6028 - lstripe now fails when requested offset > ost_count (2237)
6029 - ensure that all kernel series have a complete list.h (1607)
6030 - fix crashes in special-file operations (2316)
6031 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6033 - allow configurable automake binary, for testing new versions
6034 - small update to the lfs documentation
6036 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6038 * fix negative export reference count in fsfilt_sync (2312)
6040 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6041 * release candidate 0.9.1
6043 - orphans are moved into the PENDING directory for possible recovery
6044 - replayed opens now open by fid for orphan/rename safety (1042)
6045 - last close of an orphan inode generates a transno (683)
6046 - chdir() and mount() now pin the directory entry (1020)
6047 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6048 - discard very old requests without processing them (1502)
6049 - remove some common, well-understood CERRORs (1505)
6050 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6051 - clear "grant" flags in OST replies until OST grant code lands (1644)
6052 - fix read performance by not clobbering i_blksize on client (1598)
6053 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6054 - make LRU size a /proc tunable, clears locks when reduced (707)
6055 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6056 - server locks take references on exports now (1558)
6057 - build fixes for 2.4.20-rh trees (1663)
6058 - return an error from lov_create if all OSCs are inactive (1751)
6059 - fix import levels when a reconnect happens without a timeout (1597)
6060 - exit early from mds_open if we get a lookup error (1749)
6061 - partial page read at EOF wouldn't wait for disk before sending (1642)
6062 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6063 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6064 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6065 - write/truncate lock inversion (1639)
6066 - Don't auto-load obdclass, portals modules during cleanup (1495)
6067 - fix timestamps from jumping to "now" (1763)
6068 - extra journal assertions (1648)
6069 - add an extra multiunlink test (1771)
6070 - fix read_record/write_record API (1776)
6071 - fix leak of offset_extent, possible incorrect i_size later (1772)
6072 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6073 - Debugging for the kqswnal_get_idle_tx problems (1820)
6074 - Allow recovery to be attempted multiple times (1536)
6075 - Write out MDS last_rcvd file after it is first created (1600)
6076 - Fix tx_descriptor leak in failed transmit situations (1827)
6077 - ext3 journaling fixes for assertion failure after IO error (1871)
6078 - class_export_put() on freed export after completion AST error (1896)
6079 - Fix revalidate looping in VFS (1322)
6080 - Don't access a freed export during MDS_REINT timeout (1521)
6081 - Add open-unlink recovery support on the MDS (1673,1764)
6082 - Return an error if no MDS data was read from last_rcvd (1946)
6083 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6084 - Don't setup a disk device that doesn't match exported UUID (317)
6085 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6086 - avoid committing NULL handle in force close
6087 - local.sh is now a one-stripe LOV configuration
6088 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6089 - fix ext3 htree duplicate directory entry corruption (1516)
6090 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6091 - update to new LOV EA format (2097)
6092 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6094 2003-06-15 Phil Schwan <phil@clusterfs.com>
6097 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6098 - per-import/export recovery handling (958, 931, 959)
6099 - multiple last-rcvd slots, for serving multiple FSes (949)
6100 - connections are again shared between multiple imp/exports (963, 964)
6101 - "umount -f" would hang if any requests needed to be sent (393, 978)
6102 - avoid pinning large req buffer by copying for queued messages (989)
6103 - add "uuid" to "lctl device" command to help upcalls (991)
6104 - "open" RPCs with transnos would confuse recovery counters (1037)
6105 - do proper endian conversion of all wire messages (288, 340, 891)
6106 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6107 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6108 - fix ptlrpc_connection leak in target_handle_connect (1174)
6109 - fix import refcounting bug in OST and MDS cleanup (1134)
6110 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6111 - fix very unlikely obd_types race condition (501)
6112 - remove osc_open hack for echo_client (1187)
6113 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6114 - a failure in read_inode2 leads to deadlock (1139)
6115 - cancel ack-locks as soon as transaction is committed (1072)
6116 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6117 - make sure to commitrw after any preprw to avoid deadlock (1162)
6118 - failing to execute a file in a lustre FS would lock inode (1203)
6119 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6120 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6121 - obd_uuid2tgt would walk off the end of the list (1255)
6122 - on IA64 the osc would give portals incorrect bulk size (1258)
6123 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6124 - fix lock inversion caused by new llite matching code (1282)
6125 - limit the number of dirty pages on a client to 10MB (1286)
6126 - timed out locks were not being corrected cancelled (1289)
6127 - fix O_DIRECT above 4GB on IA-32 (1292)
6128 * major user-visible changes
6129 - fail out/fail over policy now controlled by the upcall (993)
6131 - add OBD_PING to check server availability and failure (954)
6132 - lustre messages are now sent in sending host order (288, 340, 891)
6133 - add eadatalen to MDS getattr reply (340)
6134 - OST read replies may contain second buffer, with per-page status (593)
6136 2003-03-11 Phil Schwan <phil@clusterfs.com>
6139 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6140 - failed open()s could cause deadlock; fixed (867, 869)
6141 - stop cancelling OST locks when files are closed (481)
6142 - overlapping XID spaces caused network corruption (851, 853)
6143 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6144 - setattr_raw added, to do single-RPC, server-side setattrs
6145 - lmc/lconf syntax change for OST UUIDs
6146 - fix crashy race condition between ptlrpc_free_req and osc_close
6147 - don't use request in mdc_enqueue if we hit a timeout (889)
6148 - don't set the inode i_size for regular files from the MDS (896)
6149 - handle out of order completion AST (842)
6150 - don't LBUG if a lock request times out after receiving AST (913)
6151 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6152 - if a bad lock AST arrives, send an error instead of dropping entirely
6153 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6154 - fix leak in bulk IO when only partially completed (899, 900, 926)
6155 - fix O_DIRECT for ia64 (55)
6156 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6157 - C-z after timeout could hang a process forever; fixed (977)
6159 - client-side I/O cache (678, 924, 929, 941, 970)
6161 - READPAGE and SETATTRs which don't take server-side locks get
6164 2003-02-11 Phil Schwan <phil@clusterfs.com>
6167 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6168 - Fix fsfilt_extN_readpage() to read a full page of directory
6169 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6170 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6171 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6172 - Most intent operations were removed, in favour of a new RPC mode
6173 that does a single RPC to the server and bypasses most of the VFS
6174 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6175 - Aggressively cancel local locks on DLM servers
6176 - mds_reint_unlink sends EA to the client if it's the last nlink.
6177 client uses that EA to unlink OST objects.
6178 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6179 - recursive symlinks were fixed (439)
6180 - fixed NULL deref in DEBUG_REQ
6181 - filter_update_lastobjid no longer calls sync, which annoyed extN
6182 - fixed multi-client small-writes to a single file problem (445)
6183 - fixed mtime updates during file writes (607)
6184 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6185 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6186 - fixed Linux OST size reporting problem (444, 656)
6187 - OST now updates object mtime with writes or setattr (607, 619)
6188 - client verifies file size before zeroing page past EOF (445)
6189 - OST now writes last allocated objid to disk with allocation (108)
6190 - LOV on echo now works (409)
6192 - mds_reint_unlink sends a new buffer, with the EA included. this
6193 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6194 set if a regular file was being unlinked, and it was the last link
6195 - use PtlGet from the target for bulk writes (315)
6196 - OST now updates object mtime with writes or setattr (607, 619)
6197 - LDLM now has a grant-time callback to revalidate locked items, if
6199 - Many MDS operations were reorganized to combat race conditions
6201 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6202 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6204 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6207 - Fully reactivate OST imports after reconnection (512, others)
6208 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6209 - More graceful error handling for truncating on dead OST (515)
6210 - Don't error out unless we're actually accessing dead stripes (474)
6211 - Fix garbage sizes when stripes are missing (410)
6212 - LRU counters were broken, causing constant lock purge (433, 432)
6213 - garbage on read from stripes with failed OSTs (441)
6214 - mark OSCs as active before reconnecting during recovery (438)
6215 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6216 - lfind did not preserve OST order in output (443)
6217 - symlinks cause hung clients, incorrect data (439)
6218 - stop dereferencing request after dropping refcount (457)
6219 - don't LASSERT(spin_is_locked) on non-SMP (455)
6220 - fixes for many rename() bugs
6221 - fstat didn't correctly synchronize attributes (399)
6222 - server must handle lock cancellation during blocking AST prep (487)
6223 - bulk descriptors were free()d too soon (511)
6224 - fix paths in lconf, which would load incorrect modules (451, 507)
6225 - fix confusing lconf 'host not found' error message (386)
6226 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6227 - fix race condition in mdc_blocking_ast() for inode access (526)
6228 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6229 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6230 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6231 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6232 - fix lstripe in several ways (kernel side) (527)
6233 - fix request leak in ldlm_cli_enqueue (262)
6234 - incorrect OSC was marked inactive after OST failure
6235 - call mds_fs_cleanup before unmounting filesystem (524)
6236 - fix races between taking ns_lock and ldlm_lock_change_resource
6237 - fix races updating LOV export open file list
6238 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6239 - fix recovery NULL deref in ldlm_cli_cancel_unused
6240 - fix some DLM races by using new hash table for lock handles (419)
6241 - permit the client to specify desired inodes, at replay
6242 - duplicate requests when we queue them for replay reintegration
6243 - fix last_rcvd offset calculation
6244 - sync after each recovered transaction, so we always make progress
6245 - never, not always, ERESTART requests without transnos
6246 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6248 - skip replay if the MDS says that the client is already connected
6249 - don't check for a recovery-enabled export to match lctl's UUID
6250 - don't INC_USE_COUNT for phantom exports
6251 - don't crash when cleaning up phantom exports (567)
6252 - don't double-finish or set replay data for errored mdc_open requests
6253 - abort requests when they time out, so we don't get old replies
6254 - send/receive replies for AST messages again
6255 - if the client says that it doesn't have the lock, cancel it on the
6257 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6258 mark it as destroyed, it will all work out when decref is called
6259 - fix module use counts (22, 581)
6261 - ASTs now expect a reply (server cancels lock on error reply)
6263 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6266 - fix many simultaneous client startup (392)
6267 - fix dentry->d_it clobbering
6268 - credentials weren't being shipped for readdir/getattr operations
6269 - remove invalid assertions triggered during some concurrent MD
6271 - proper Lustre versions added (336, 389)
6272 - fix memory leak for create error case (398)
6273 - fix LOV locking bug that would get cli/srv out of sync
6274 - fix echo client over LOV (409)
6275 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6276 - fix double-O_EXCL intent crash (424)
6277 - avoid sending multiple lock CANCELs (352)
6279 - MDS can do multi-client recovery (modulo bugs in new code)
6281 - many updates, edits, cleanups
6283 2002-11-18 Phil Schwan <phil@clusterfs.com>
6286 - fix null d_it dereference (346)
6287 - fix full OST/dbench hang (333)
6288 - fix permission problem with file removal (286)
6289 - fix removal of OSCs from LOV when they fail
6290 - fix NULL deref during bulk timeout (214)
6291 - fix problems related to multiple filesystems on one MDS (241)
6292 - fixed serious subtle metadata locking bugs
6293 - free locks on clients when inodes are removed due to memory
6295 - fix inode pointer in lock data (285)
6296 - partial support for multiple MDS on a single host (241)
6297 - data locks weren't cancelled at clear_inode time (290, 311)
6298 - intent locks could lead to unbounded lock growth (205)
6299 - added a maximum lock count, an LRU list, and a flusher
6300 - fix multiple rename (365)
6301 - properly abstracted the echo client
6302 - OSC locked 1 byte too many; fixed
6303 - rewrote brw callback code:
6304 - fixed recovery bugs related to LOVs (306)
6305 - fixed too-many-pages-in-one-write crash (191)
6306 - fixed (again) crash in sync_io_timeout (214)
6307 - probably fixed callback-related race (385)
6309 - Add capability to MDS protocol
6310 - LDLM cancellations and callbacks on different portals
6312 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6315 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6316 - defer open object destruction to close time (601981, 138)
6317 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6318 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6319 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6320 - don't call obd_setattr in ll_file_release for destroyed objects
6321 * protocol change to lustre_msg: move |version| and add |flags|
6322 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6323 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6324 * added replay of create, unlink, link and rename operations during
6325 MDS failover; recovery should be much more robust now
6326 * remove failed OSCs from LOVs (only lov_create uses this so far)
6327 * the lustre-HOWTO was brought (more) up to date (582544)
6329 2002-10-23 Phil Schwan <phil@clusterfs.com>
6332 - in-use dentries weren't being reused properly (617851)
6333 - prevent multiple LDLM setup (599178)
6334 - fix LOV size calculations for truncate (617853)
6335 - fix client handling of MDS intent errors (POSIX)
6336 - fix permission bug in lovstripe.c test (624321)
6337 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6338 - truncate past end of file could corrupt data
6339 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6340 - a race in recovery could return ETIMEDOUT to apps (623947)
6341 - building outside the source directory was fixed
6342 * the lustre-HOWTO was brought (more) up to date (582544)
6343 * major progress was made on recovery functionality
6345 2002-10-10 Phil Schwan <phil@clusterfs.com>
6348 - recovery deadlock fix
6349 - rm -rf causes LBUG fix (617817)
6350 - file open by multiple tasks fix (618962)
6351 - directory permissions bugs (602707 and 620007)
6352 - journal_stop fixed with locking (611313)
6353 - O_APPEND failures resolved (618273, perhaps 614459)
6354 - lconf PATH fix (619770)
6355 - IA64 build fix (621450)
6356 - RPC buffer sizes scale with amount of memory
6358 2002-10-01 Phil Schwan <phil@clusterfs.com>
6361 - locks would be cancelled without throwing away data pages,
6362 resulting in inconsistent data (605627)
6363 - inode attributes were not always being refreshed (605627, 612449)
6364 - lconf now continues to cleanup after lctl reports an error
6365 - MDS now enforces user permissions (602707)
6366 - lprocfs cleanup fixed, but not yet enabled (614157)
6367 - fixed infinite server hang, should a client not respond to an AST
6368 - avoid going into recovery if user calls readlink() with a buffer
6369 that's too small (613941)
6370 - AST RPCs no longer require replies (614867) -- this may be changed
6371 - don't crash server if client sends an IOV that's too big (611336)
6372 - fixed lock conversion deadlock (611892)
6373 - fixed the following of symlinks (614622)
6374 * recovery: the server can remove locks from a client that dies, other
6375 clients can make progress
6376 * more extN patch fixes
6377 * compile-time configurable ptlrpc buffer allocations
6379 - collaborative read cache document
6380 - Lustre Lite Performance CDR document-in-progress
6382 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6385 - fix typo in patch-2.4.18
6387 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6390 - clear ptlrpc request each time in handle_incoming_request()
6391 - unlink of files now destroys the object on the OST
6393 2002-09-19 Peter Braam <braam@clusterfs.com>
6395 * add hard link support
6396 * change obdfile creation method
6397 * kernel patch changed
6399 2002-09-19 Peter Braam <braam@clusterfs.com>
6402 - stack overflow bug in extN fixed
6404 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6406 * documentation updates
6407 - add man pages for config tools
6408 - update tests/README to describe testing with new config tools
6409 - finish metadata API descriptions
6410 * bug fixes and cleanups
6411 - statfs workaround for 16TB limit
6412 - LOV stripe allocation improved, can stripe on subset of OSTs
6413 - LOV file size/IO offset was wrong for files > 4GB in size
6414 - object EA data was being dropped, caused files to be unreadable
6415 - memory overflow with non-LOV OST caused memory corruption
6416 - fixed regression tests to work with new config tools, obdfilter
6417 - fixed bug when directory size became larger than 1 block
6418 - fixed bug (for single client case) when PWD was deleted
6419 - invalidate local directory pages when doing intent-based ops
6420 - avoid LDLM oops when lock callback contained bad data
6422 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6424 * documentation updates
6425 * bug fixes and cleanups
6426 - configuration tools
6429 - 64-bit compile warnings
6430 - 64-bit internal statfs data
6432 * test_brw on persistent OST devices
6434 * lprocfs (disabled)
6436 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6438 * documentation updates
6439 * bug fixes and cleanups
6440 * configuration tools
6442 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6445 * many small fixes to 0_5_4
6446 * io/network handling
6447 * thinkos in MDS operations
6449 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6452 * crucial basic fixes to 0.5.3
6453 * IOR, Iozone work over Elan
6456 2002-08-07 Phil Schwan <phil@clusterfs.com>
6457 * version 0_5_3, our first alpha
6458 * we use the new Portals iovs
6459 * documentation updates
6460 * bug fixes and cleanups
6461 * small changes in the DLM wire protocol
6463 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6464 * version 0_5_1 with some initial stability,
6465 * locking on MD and file I/O.
6466 * documentation updates
6467 * several bug fixes since 0.5.0
6468 * small changes in wire protocol
6470 2002-07-18 Phil Schwan <phil@clusterfs.com>
6472 * delivered as Lustre Light Alpha
6473 * fixed a crash after handling invalid MDS requests
6474 * fixed directory pages for architectures with non-4k pages sizes
6476 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6477 * release version v0_4_4
6478 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6480 * Fixes a number of interruption problems with OST operations.
6481 * Update documentation for portals header changes
6482 * Move all wire protocol structs/defines to lustre_idl.h
6483 * Fixes symlink length bug.
6484 * Add tcpdump to repository.
6486 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6487 * release version v0_4_3
6488 * Fixes statfs for inodes on extN.
6489 * Fixes bug in runtests which would delete /etc/hosts.
6490 * Use 64-bit object IDs wherever possible (not into VFS though)
6491 Remove ost_get_info, which is unused by lustre, and out of date.
6493 2002-07-03 Peter Braam <braam@clusterfs.com>
6494 * release version v0_4_2 Fixes a lookup error (type not passed)
6495 * move forward to head of Portals
6496 * move forward to latest Lustre kernel
6498 2002-06-25 Peter Braam <braam@clusterfs.com>
6499 * release version v0_4_1. Hopefully stable on single node use.