1 tbd Sun Microsystems, Inc.
4 2.6.9-67.0.4.EL (RHEL 4),
5 2.6.16.54-0.2.5 (SLES 10),
6 2.6.18-53.1.14.el5 (RHEL 5).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.40.4-cfs1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
15 Severity : enhancement
17 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
20 Frequency : frequent on X2 node
22 Description: mdc_set_open_replay_data LBUG
23 Details : Set replay data for requests that are eligible for replay.
27 Description: lustre_mgs: operation 101 on unconnected MGS
28 Details : When MGC is disconnected from MGS long enough, MGS will evict the
29 MGC, and late on MGC cannot successfully connect to MGS and a lot
30 of the error messages complaining that MGS is not connected.
33 Frequency : on start mds
35 Description: Implement get_info(last_id) in obdfilter.
38 Frequency : occasional
40 Description: Correctly check stale fid, not start epoch if ost not support SOM
41 Details : open with flag O_CREATE need set old fid in op_fid3 because op_fid2
42 overwrited with new generated fid, but mds can anwer with one of these
43 two fids and both is not stale. setattr incorectly start epoch and
44 assume will be called done_writeting, but without SOM done_writing
48 Frequency : rare, depends on device drivers and load
50 Description: MDS or OSS nodes crash due to stack overflow
51 Details : Code changes in 1.8.0 increased the stack usage of some functions.
52 In some cases, in conjunction with device drivers that use a lot
53 of stack the MDS (or possibly OSS) service threads could overflow
54 the stack. One change which was identified to consume additional
55 stack has been reworked to avoid the extra stack usage.
58 Frequency : occasional
60 Description: Do not fail import if osc_interpret_create gets -EAGAIN
61 Details : If osc_interpret_create got -EAGAIN it immediately exits and
62 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
63 oscc_has_objects and see OSC has no objests and call
64 oscc_internal_create to resend create request.
66 Severity : enhancement
68 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
70 Severity : enhancement
72 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
77 Description: Fix replay if there is an un-replied request and open
78 Details : In some cases, older replay request will revert the
79 mcd->mcd_last_xid on MDS which is used to record the client's
82 Severity : enhancement
84 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
86 Severity : enhancement
88 Description: Add rhel5 support to HEAD.
90 Serverity : enhancement
92 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
97 Description: Don't allow skipping OSTs if index has been specified.
98 Details : Don't allow skipping OSTs if index has been specified, make locking
99 in internal create lots better.
103 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
104 Details : access to bitfield in structure is always rounded to long
105 and this produce problem with not atomic change any bit.
109 Description: Lustre make rpms failed.
110 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
111 builds Lustre rpms from tarball.
114 Frequency : rare on shutdown ost
116 Description: If llog cancel was not send before clean_exports phase, this can
117 produce deadlock in llog code.
118 Details : If llog thread has last reference to obd and call class_import_put
119 this produce deadlock because llog_cleanup_commit_master wait when
120 last llog_commit_thread exited, but this never success because was
121 called from llog_commit_thread.
125 Description: allow userland application know is lost one of stripes.
126 Details : fill lvb_blocks with error code on ost and return it to
127 application if error flag found.
131 Description: NULL lov_tgts causing MDS oops
132 Details : more safe checks for NULL lov_tgts for avoid oops.
134 Severity : enhancement
136 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
140 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
141 Details : If an OST is down the MDS will hang indefinitely in
142 obd_statfs_rqset() waiting for the statfs data. While for
143 MDS QOS usage of statfs, it should not stuck in waiting.
145 Severity : enhancement
147 Description: remote_acl support
148 Details : Support ACL-based permission check for remote user.
149 Support setfacl/getfacl for remote user with the utils
150 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
151 the system "{s,g}etfacl" utils.
153 Severity : enhancement
155 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
157 Severity : enhancement
159 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
163 Description: make mgs_setparam() handle fsname containing dash
164 Details : fsname containing a dash does not work with lctl conf_param
166 Severity : enhancement
168 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
170 Severity : enhancement
172 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
176 Frequency : rare, at shutdown
177 Description: access already free / zero obd_namespace.
178 Details : if client_disconnect_export was called without force flag set,
179 and exist connect request in flight, this can produce access to
180 NULL pointer (or already free pointer) when connect_interpret
181 store ocd flags in obd_namespace.
185 Frequency : only at startup
186 Description: not alloc memory with spinlock held.
187 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
188 if any spinlock held.
190 Severity : enhancement
192 Description: make lustre randomly fail allocating memory
193 Details : Make lustre randomly failed allocating memory for testing purpose.
195 Severity : enhancement
197 Description: lost problems with lov objid file
198 Details : Fixes some scability and access to not inited memory problems
199 in work with lov objdid file.
204 Description: lfs find does not continue on file error
205 Details : Continue other files processing when a file/dir is absent.
209 Description: Inconsistent usage of lustre_pack_reply()
210 Details : Standardize the usage of lustre_pack_reply() such that it
211 always generate a CERROR on failure.
214 Frequency : occasional
216 Description: MDS hang or stay in waiting lock
217 Details : If client receive lock with CBPENDING flag ldlm need send lock
218 cancel as separate rpc, to avoid situation when cancel request
219 can't processed due all i/o threads stay in wait lock.
223 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
227 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
231 Description: Update to SLES9 kernel-2.6.5-7.287.3.
235 Description: Update to RHEL4 latest kernel.
237 Severity : enhancement
239 Description: Build SLES10 patchless client fails
240 Details : The configure was broken by run ./configure with
241 --with-linux-obj=.... argument for patchless client. When the
242 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
243 can't search header adequately. Use absolute path such as
244 -I($LINUX)/include instead.
248 Description: interrupt oig_wait produce painc on resend.
249 Details : brw_redo_request can be used for resend requests from ptlrpcd and
250 private set, and this produce situation when rq_ptlrpcd_data not
251 copyed to new allocated request and triggered LBUG on assert
252 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
253 ptlrpcd set if request is changed and can be safety changed to use
258 Description: LASSERT_{REQ,REP}SWAB macros are buggy
259 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
260 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
261 these macros and replace them with their difinitions instead.
265 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
266 Details : Update which_patch & target file for SLES10 latest kernel.
269 Description: Remove client patches from SLES 10 kernel.
270 Details : This causes SLES 10 clients to behave as patchless clients
271 even on a Lustre-patched (server) kernel.
273 Severity : enhancement
275 Description: self-adjustable client's lru lists
276 Details : use adaptive algorithm for managing client cached locks lru
277 lists according to current server load, other client's work
278 pattern, memory activities, etc. Both, server and client
279 side namespaces provide number of proc tunables for controlling
282 Severity : enhancement
284 Description: light-weight GSS support
285 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
286 and reduce performance overhead.
288 Severity : enhancement
290 Description: Linux keyring support
291 Details : Support using service of Linux keyring for Lustre GSS internal
292 context refresh/cache mechanism.
296 Description: Fix errors in lfs documentation
297 Details : Fixes man pages
301 Description: don't use GFP_* in generic Lustre code.
302 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
306 Description: obdclass is limited by single OBD_ALLOC(idarray)
307 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
311 Description: warnings with build patchless client with vanila 2.6.19 and up
312 Details : change old ctl_table style and replace ctl_table/ctl_table_header
313 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
317 Frequency : Only for SLES
318 Description: sanity_quota fail test_1
319 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
320 file, which makes TSTID[2] inunique.
324 Frequency : Only for SLES9
325 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
326 Details : Update target/ChangeLog/which_patch .
330 Frequency : Only for test_10 in sanity-quota.sh
331 Description: a bug in quota test code
332 Details : checking a test flag in a wrong place
336 Frequency : Only for test_13 in sanity-quota.sh
337 Description: a bug in quota test script
338 Details : shouldn't check overfull allocated quota
342 Description: jbd statistics
343 Details : Port older jbd statistics patch for sles10
345 Severity : enhancement
347 Description: Add printing inode info into message about error in writepage.
352 Description: reply_lock_interpret crash due to race with it and lock cancel.
353 Details : Do not replay locks that are being cancelled. Do not reference
354 locks by their address during replay, just by their handle.
358 Frequency : When flocks are used.
359 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
360 Details : Flock locks might destroy just granted lock if it could be merged
361 with another existing flock, this is done in completion handler,
362 so teach ldlm_cli_enquque_fini that this is a valid case for
368 Description: Oops in read and write path when failing to allocate lock.
369 Details : Check if lock allocation failed and return error back.
373 Description: lstripe command fails for valid OST index
374 Details : The stripe offset is compared to lov->desc.ld_tgt_count
375 instead of lov->desc.ld_active_tgt_count.
379 Description: sanity.sh failed test 103
380 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
385 Description: df doesn't work properly if diskfs blocksize != 4K
386 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
390 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
391 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
395 Description: obd_health_check_timeout too short
396 Details : set obd_health_check_timeout as 1.5x of obd_timeout
400 Description: llapi_file_create() does not allow some changes
401 Details : add llapi_file_open() that allows specifying the mode and
402 open flags, and also returns an open file handle.
406 Description: merge and cleanup kernel patches.
407 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
411 Description: Add journal checksum support.(Kernel part)
412 Details : The journal checksum feature adds two new flags i.e
413 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
414 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
415 indicates that the commit block contains the checksum for
416 the blocks described by the descriptor blocks. Now commit
417 record can be sent to disk without waiting for descriptor
418 blocks to be written to disk. This behavior is controlled
419 using JBD2_FEATURE_ASYNC_COMMIT flag.
423 Description: OSS needs mutliple precreate threads
424 Details : Add ability to start more than one create thread per OSS.
428 Description: Sanity.sh test_65k failed: setstripe should have succeeded
429 Details : execute lfs setstripe on client
433 Description: mds_obd_create error creating tmp object
434 Details : When the user sets quota on root, llog will be affected and can't
435 create files and write files.
438 Frequency : Always on ia64 patchless client, and possibly others.
440 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
441 Details : This allows the patchless client to be loaded on architectures
446 Description: RedHat Update kernel for RHEL5
447 Details : Add the kernel config file for RHEL5.
451 Description: RedHat Update kernel for RHEL5
452 Details : Modify the kernel config file more closer RHEL5.
456 Description: Build failure against Centos5 (RHEL5)
457 Details : Define PAGE_SIZE when it isn't present.
461 Description: test_7 on sanity-quota.sh can't be used on separate machines
462 Details : add facet to handle it
466 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
467 Details : with NFS, the anon dentry's parent was set to itself in
468 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
469 corresponding dentry other than use rec->ur_name.
471 Severity : enhancement
473 Description: lfs setstripe enhancement
474 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
478 Description: enable data checksumming by default
479 Details : enable checksum by default, allow --disable-checksum
480 configure option and "-o nochecksum" mount option. Checksums
481 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
485 Description: lustre support for RHEL5
486 Details : Add support for RHEL5.
490 Description: Client eviction due to failover config
491 Details : after a connection loss, the lustre client should attempt to
492 reconnect to the last active server first before trying the
493 other potential connections.
497 Description: when mds and osts use different quota unit(32bit and 64bit),
498 quota will be released repeatly.
499 Details : void sending multiple quota reqs to mds, which will keep the
500 status between the reqs.
504 Description: osts not allocated evenly to files
505 Details : change the condition to increase offset_idx
508 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
510 Description: Data corruption for OSTs that are formatted larger than 2TB
512 Details : When generating the bio request for lustre file writes the
513 sector number would overflow a temporary variable before being
514 used for the IO. The data reads correctly from Lustre (which
515 will overflow in a similar manner) but other file data or
516 filesystem metadata may be corrupted in some cases.
520 Description: Tune the kernel for good SCSI performance.
521 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
522 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
527 Description: rewrite ext2-derived code in obdclass/uuid.c
528 Details : rewrite inherited code (uuid parsing code from ext2 utils)
529 from scratch preserving functionality.
533 Description: "lfs find -obd UUID" prints directories
534 Details : "lfs find -obd UUID" will return all directory names instead
535 of just file names. It is incorrect because the directories
536 do not reside on the OSTs.
540 Description: use i_size_read and i_size_write in 2.6 port
541 Details : replace inode->i_size access with i_size_read/write()
544 Frequency : when removing large files
546 Description: scheduling issue during removal of large Lustre files
547 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
548 It causes scheduling issues when removing large files (17TB in the
552 Frequency : only with liblustre clients on XT3
554 Description: evictions taking too long
555 Details : allow llrd to evict clients directly on OSTs
558 Frequency : only on ppc
560 Description: /proc/fs/lustre/devices broken on ppc
561 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
562 We should make sure the type of 'index' is loff_t and then cast
563 explicitly as needed below. Do not assign an explicitly cast
567 Frequency : only for rhel5
569 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
570 Details : Modify the target file & which_kernel.
572 Severity : enhancement
574 Description: omit set fsid for export NFS
575 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
576 as FS_REQUIRES_DEV for avoid problems with generate fsid.
580 Frequency : Always, for kernels after 2.6.16
581 Description: Fix warning idr_remove called for id=.. which is not allocated.
582 Details : Last kernels save old s_dev before kill super and not allow
583 to restore from callback - restore it before call kill_anon_super.
587 Description: buffer overruns could theoretically occur
588 Details : llapi_semantic_traverse() modifies the "path" argument by
589 appending values to the end of the origin string, and a
590 overrun may occur. Adding buffer overrun check in liblustreapi.
594 Description: Fix error on 'ls .' at the top of the Lustre mount.
595 Details : Don't revalidate dentry if it is a root dentry.
599 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
600 Details : Modify vm-tunables-rhel4.patch.
604 Description: Kernel config for 2.6.18-vanilla.
605 Details : Modify targets/2.6-vanilla.target.in.
606 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
607 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
608 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
609 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
613 Description: improve handling recoverable errors
614 Details : if request processig with error which can be recoverable on server
615 request should be resend, otherwise page released from cache and
620 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
621 Details : Modify target file & which_patch.
622 A flaw was found in the IA32 system call emulation provided
623 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
624 value could be stored in the %RAX register, which could trigger an
625 out-of-bounds system call table access. An untrusted local user
626 could exploit this flaw to run code in the kernel
627 (ie a root privilege escalation). (CVE-2007-4573).
631 Description: O_DIRECT bypasses client statistics.
632 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
633 read_ahead_stats were updated. copied the stats section from
634 osc_send_oap_rpc() into async_internal().
638 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
642 Description: change order of libsysio includes
643 Details : '#include sysio.h' should always come before '#include xtio.h'
647 Description: handle "serious error: objid * already exists" more gracefully
648 Details : If LAST_ID value on disk is smaller than the objects existing in
649 the O/0/d* directories, it indicates disk corruption and causes an
650 LBUG(). If the object is 0-length, then we should use the existing
651 object. This will help to avoid a full fsck in most cases.
653 Severity : enhancement
655 Description: adapt the lustre_config script to support the upgrade case
656 Details : Add "-u" option for lustre_config script to support upgrading 1.4
657 server targets to 1.6 in parallel.
661 Description: To avoid grant space > avaible space when the disk is almost
662 full. Without this patch you might see the error "grant XXXX >
663 available" or some LBUG about grant, when the disk is almost
665 Details : In filter_check_grant, for non_grant cache write, we should
666 check the left space by if (*left > ungranted + bytes), instead
667 of (*left > ungranted), because only we are sure the left space
668 is enough for another "bytes", then the ungrant space should be
669 increase. In client, we should update cl_avail_grant only there
670 is OBD_MD_FLGRANT in the reply.
674 Description: Update RHEL 4 kernel to fix local root privilege escalation.
675 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
676 described in CVE-2007-4573. This problem could allow untrusted
677 local users to gain root access.
680 Frequency : when using O_DIRECT and quotas
682 Description: Incorrect file ownership on O_DIRECT output files
683 Details : block usage reported by 'lfs quota' does not take into account
684 files that have been written with O_DIRECT.
689 Description: touch file failed when fs is not full
690 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
691 otherwise we can get ENOSP while fs is not full.
695 Description: parallel lock callbacks
696 Details : Instead of sending blocking and completion callbacks as separated
697 requests, adding them to a set and sending in parallel.
700 Frequency : only for Cray XT3
701 Bugzilla : 12829/13455
702 Description: Changing primary group doesn't change the group lustre assigns to
704 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
705 overridden with the primary group provided by the group upcall,
706 whereas the supplied fsgid can be trusted if it is in the list of
707 supplementary groups returned by the group upcall.
709 Severity : enhancement
711 Description: Allow masking D_WARNING, D_ERROR messages from console
712 Details : Console messages can now be disabled via lnet.printk.
716 Description: User code with malformed file open parameter crashes client node
717 Details : Before packing join_file req, all the related reference should be
718 checked carefully in case some malformed flags cause fake
719 join_file req on client.
723 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
724 Details : ldlm_completion_ast() assumes that a lock is granted when the req
725 mode is equal to the granted mode. However, it should also check
726 that LDLM_FL_CP_REQD is not set.
730 Description: Heavy nfs access might result in deadlocks
731 Details : After ELC code landed, it is now improper to enqueue any mds
732 locks under och_sem, because enqueue might want to decide to
733 cancel open locks for same inode we are holding och_sem for.
737 Description: Client eviction while running blogbench
738 Details : A lot of unlink operations with concurrent I/O can lead to a
739 deadlock causing evictions. To address the problem, the number of
740 oustanding OST_DESTROY requests is now throttled to
741 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
742 callbacks are processed in priority.
746 Description: enable ACLs on MDS by default
747 Details : ACLs must be enabled on MDS by default.
750 Frequency : PPC/PPC64 only
752 Description: conflicts between asm-ppc64/types.h and lustre_types.h
753 Details : fix duplicated definitions between asm-ppc64/types.h and
754 lustre_types.h on PPC.
757 Frequency : PPC/PPC64 only
759 Description: asm-ppc/segment.h does not exist
760 Details : fix compile issue on PPC.
764 Description: better handle error messages in extents code
767 Frequency : RHEL4 only
769 Description: mkfs is very slow on IA64/RHEL4
770 Details : A performance regression has been discovered in the MPT Fusion
771 driver between versions 3.02.73rh and 3.02.99.00rh. As a
772 consequence, we have downgraded the MPT Fusion driver in the RHEL4
773 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
775 Severity : enhancement
777 Description: SNMP support enhancement
778 Details : Adding total number of sampled request for an MDS node in snmp
781 Severity : enhancement
783 Description: Optimize ldlm waiting list processing for PR extent locks
784 Details : When processing waiting list for read extent lock and meeting read
785 lock that is same or wider to it that is not contended, skip
786 processing rest of the list and immediatelly return current
787 status of conflictness, since we are guaranteed there are no
788 conflicting locks in the rest of the list.
792 Description: Time out and refuse to reconnect
793 Details : When the failover node is the primary node, it is possible
794 to have two identical connections in imp_conn_list. We must
795 compare not conn's pointers but NIDs, otherwise we can defeat
796 connection throttling.
800 Description: port llog fixes from b1_6 into HEAD
801 Details : Port llog reference couting and some llog cleanups from b1_6
802 (bug 10800) into HEAD, for protect from panic and access to already
803 free llog structures.
807 Description: Detect stride IO mode in read-ahead
808 Details : When a client does stride read, read-ahead should detect that and
809 read-ahead pages according to the detected stride pattern.
813 Description: data checksumming impacts single node performance
814 Details : add support for several checksum algorithm. Currently, only CRC32
815 and Adler-32 are supported. The checksum type can be changed on
816 the fly via /proc/fs/lustre/osc/*/checksum_type.
820 Description: use adler32 for page checksums
821 Details : when available, use the Adler-32 algorithm instead of CRC32 for
824 --------------------------------------------------------------------------------
826 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
828 * Support for kernels:
829 2.6.5-7.283 (SLES 9),
830 2.6.9-55.EL (RHEL 4),
831 2.6.16.46-0.14 (SLES 10),
832 2.6.18.8 vanilla (kernel.org)
833 * Client support for unpatched kernels:
834 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
835 2.6.16 - 2.6.22 vanilla (kernel.org)
836 * Due to recently discovered recovery problems, we do not recommend
837 using patchless RHEL 4 clients with this or any earlier release.
838 * Recommended e2fsprogs version: 1.39.cfs8
839 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
840 * Starting with this release, the ldiskfs backing filesystem required
841 by Lustre is now in its own package, lustre-ldiskfs. This package
842 should be installed. It is versioned separately from Lustre and
843 may be released separately in future.
847 Description: block reactivating mgc import until all deactivates complete
848 Details : Fix race when failing back MDT/MGS to itself (testing)
851 Frequency : at statup only
853 Description: mds_lov_synchronize race leads to various problems
854 Details : simultaneous MDT->OST connections at startup can cause the
855 sync to abort, leaving the OSC in a bad state.
857 Severity : enhancement
859 Description: add optional extra BUILD_VERSION info
860 Details : add a new environment variable (namely LUSTRE_VERS) which allows
861 to override the lustre version.
864 Frequency : 2.6.18 servers only
866 Description: ll_kern_mount() doesn't release the module reference
867 Details : The ldiskfs module reference count never drops down to 0
868 because ll_kern_mount() doesn't release the module reference.
873 Description: server LBUG when using old ost_num_threads parameter
874 Details : Accept the old ost_num_threads parameter but warn that it
875 is deprecated, and fix an off-by-one error that caused an LBUG.
880 Description: Transient SCSI error results in persistent IO issue
881 Details : iobuf->dr_error is not reinitialized to 0 between two
885 Frequency : sometimes when underlying device returns I/O errors
887 Description: OSTs not going read-only during write failures
888 Details : OSTs are not remounted read-only when the journal commit threads
889 get I/O errors because fsfilt_ext3 calls journal_start/stop()
890 instead of the ext3 wrappers.
894 Description: poor connect scaling with increasing client count
895 Details : Don't run filter_grant_sanity_check for more than 100 exports
896 to improve scaling for large numbers of clients.
899 Frequency : SLES10 only
901 Description: sanity-quota.sh quotacheck failed: rc = -22
902 Details : Quotas cannot be enabled on SLES10.
905 Frequency : liblustre clients only
907 Description: getdirentries does not give error when run on compute nodes
908 Details : getdirentries does not fail when the size specified as an argument
909 is too small to contain at least one entry
911 Severity : enhancement
913 Description: Add LNET router traceability for debug purposes
914 Details : If a checksum failure occurs with a router as part of the
915 IO path, the NID of the last router that forwarded the bulk data
916 is printed so it can be identified.
921 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
922 Details : Due to a race condition, liblustre clients were occasionally
925 Severity : enhancement
927 Description: lfs setstripe use optional parameters instead of postional
930 Severity : enhancement
932 Description: Nanosecond timestamp support for ldiskfs
933 Details : The on-disk ldiskfs filesystem has added support for nanosecond
934 resolution timestamps. There is not yet support for this at
935 the Lustre filesystem level.
938 Frequency : during server recovery
940 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
941 Details : request with rq_no_resend flag not awake l_wait_event if they get a
945 Frequency : nfs export on patchless client
947 Description: connectathon hang when test nfs export over patchless client
948 Details : Disconnected dentry cannot be found with lookup, so we do not need
949 to unhash it or make it invalid
952 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
953 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
954 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
959 Description: open req refcounting wrong on reconnect
960 Details : If reconnect happened between getting open reply from server and
961 call to mdc_set_replay_data in ll_file_open, we will schedule
962 replay for unreferenced request that we are about to free.
963 Subsequent close will crash in variety of ways.
964 Check that request is still eligible for replay in
965 mdc_set_replay_data().
970 Description: disable writes to filesystem when reading health_check file
971 Details : the default for reading the health_check proc file has changed
972 to NOT do a journal transaction and write to disk, because this
973 can cause reads of the /proc file to hang and block HA state
974 checking on a healthy but otherwise heavily loaded system. It
975 is possible to return to the previous behaviour during configure
976 with --enable-health-write.
978 Severity : enhancement
980 Description: 64-bit inode version
981 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
982 to the inode. This will be required for version-based recovery.
987 Description: MDS fails to start if a duplicate client export is detected
988 Details : in some rare cases it was possible for a client to connect to
989 an MDS multiple times. Upon recovery the MDS would detect this
990 and fail during startup. Handle this more gracefully.
992 Severity : enhancement
994 Description: Add -o localflock option to simulate old noflock
996 Details : This will achieve local-only flock/fcntl locks
1002 Description: log_commit_thread vs filter_destroy race leads to crash
1003 Details : Take import reference before releasing llog record semaphore
1008 Description: Wrong request locking in request set processing
1009 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1010 imp_delayed_list, in this place should be used imp_lock.
1013 Frequency : when reconnection
1015 Description: Grant Leak when osc reconnect to OST
1016 Details : When osc reconnect ost, OST(filter) should check whether it
1017 should grant more space to client by comparing fed_grant and
1018 cl_avail_grant, and return the granted space to client instead
1019 of "new granted" space, because client will call osc_init_grant
1020 to update the client grant space info.
1023 Frequency : when client reconnect to OST
1025 Description: Grant Leak when osc do resend and replay bulk write
1026 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1027 bulk write request, because the grant info while be sync between
1028 OSC and OST when reconnect, and we should ignore the grant info
1029 these of resend/replay write req.
1034 Description: Grant space more than avaiable left space sometimes.
1035 Details : When then OST is about to be full, if two bulk writing from
1036 different clients came to OST. Accord the avaliable space of the
1037 OST, the first req should be permitted, and the second one
1038 should be denied by ENOSPC. But if the seconde arrived before
1039 the first one is commited. The OST might wrongly permit second
1040 writing, which will cause grant space > avaiable space.
1043 Frequency : when client is evicted
1045 Description: Grant might be wrongly erased when osc is evicted by OST
1046 Details : when the import is evicted by server, it will fork another
1047 thread ptlrpc_invalidate_import_thread to invalidate the
1048 import, where the grant will be set to 0. While the original
1049 thread will update the grant it got when connecting. So if
1050 the former happened latter, the grant will be wrongly errased
1051 because of this race.
1056 Description: Checking Stale with correct fid
1057 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1058 to check whether it is stale, because sometimes, we want the
1059 enqueue happened anyway, and op_data.fid2 will not be initialized.
1061 Severity : enhancement
1063 Description: update patchless client
1064 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1067 Frequency : only with 2.4 kernel
1069 Description: random memory corruption
1070 Details : size of struct ll_inode_info is to big for union inode.u and this
1071 can be cause of random memory corruption.
1076 Description: Memory leak in recovery
1077 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1078 It should also check obd_fail before fsfilt_start, otherwise if
1079 fsfilt_start return -EROFS,(failover mds during mds recovery).
1080 then the req will return with repmsg->transno = 0 and rc = EROFS.
1081 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1082 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1083 be freed no matter whether fsfilt_commit success or not.
1086 Frequency : only with huge count clients
1088 Description: Prevents from taking the superblock lock in llap_from_page for
1090 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1091 for llap_from_page call in ll_removepage prevents from taking the
1092 superblock lock for a soon died page.
1097 Description: Not check open intent error before release open handle
1098 Details : in some rare cases, the open intent error is not checked before
1099 release open handle, which may cause
1100 ASSERTION(open_req->rq_transno != 0), because it tries to release
1101 the failed open handle.
1106 Description: Set cat log bitmap only after create log success.
1107 Details : in some rare cases, the cat log bitmap is set too early. and it
1108 should be set only after create log success.
1112 Description: Accessing a block bevice can re-enable I/O when Lustre is
1113 tearing down a device.
1114 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1119 Description: service threads may hog cpus when there are a lot of requests
1120 Details : Insert cond_resched to give other threads a chance to use some CPU
1125 Description: the cat log was not initialized in recovery
1126 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1127 unlink log on mds will not be initialized until mds post
1128 recovery. And also in mds post recovery, the unlink log will
1129 initialization will be done asynchronausly, so there will be race
1130 between add unlink log and unlink log initialization.
1134 Description: brw_stats were being printed incorrectly
1135 Details : brw_stats were being printed as log2 but all of them were not
1136 recorded as log2. Also remove some code duplication arising from
1137 filter_tally_{read,write}.
1141 Frequency : rare, only in recovery.
1142 Description: ASSERTION(req->rq_type != LI_POISON) failed
1143 Details : imp_lock should be held while iterating over imp_sending_list for
1144 prevent destroy request after get timeout in ptlrpc_queue_wait.
1148 Description: replay-single.sh test 52 fails
1149 Details : A lock's skiplist need to be cleanup when it being unlinked
1150 from its resource list.
1154 Description: Short directio read returns full requested size rather than
1156 Details : Direct I/O operations should return actual amount of bytes
1157 transferred rather than requested size.
1159 Severity : enhancement
1161 Description: metadata RPC reduction (e.g. for rm performance)
1162 Details : decrease the amount of synchronous RPC between clients and servers
1163 by canceling conflicing lock before the operation on the client
1164 and packing thier handles into the main operation RPC to server.
1166 Severity : enhancement
1168 Description: Async OSC create to avoid the blocking unnecessarily.
1169 Details : If a OST has no remain object, system will block on the creating
1170 when need to create a new object on this OST. Now, ways use
1171 pre-created objects when available, instead of blocking on an
1172 empty osc while others are not empty. If we must block, we block
1173 for the shortest possible period of time.
1175 Severity : enhancement
1177 Description: refine locking for avoid write wrong info into lov_objid file
1178 Details : fix possible races with add new target and write/update data in
1181 --------------------------------------------------------------------------------
1183 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1188 Frequency : on some architectures
1190 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1191 Details : Uninitialized flags sometimes cause configuration commands to
1195 Frequency : patchless clients only
1197 Description: missing __iget() symbol export
1198 Details : The __iget() symbol export is missing. To avoid the need for
1199 this on patchless clients the deathrow inode reaper is turned
1200 off, and we depend on the VM to clean up old inodes. This
1201 dependency was during via the fix for bug 12181.
1203 --------------------------------------------------------------------------------
1205 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1207 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1208 INTEROPERATE with older versions automatically. In many cases a
1209 special upgrade step is needed. Please read the
1210 user documentation before upgrading any part of a live system.
1211 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
1212 version will not interoperate with 1.6 betas before beta5 (1.5.95).
1213 * WARNING: Lustre configuration and startup changes are required with
1214 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1220 Severity : enhancement
1222 Description: MountConf
1223 Details : Lustre configuration is now managed via mkfs and mount
1224 commands instead of lmc and lconf. New obd types (MGS, MGC)
1225 are added for dynamic configuration management. See
1226 https://mail.clusterfs.com/wikis/lustre/MountConf for
1229 Severity : enhancement
1231 Description: dynamic OST addition
1232 Details : OSTs can now be added to a live filesystem
1234 Severity : enhancement
1236 Description: startup order invariance
1237 Details : MDTs and OSTs can be started in any order. Clients only
1238 require the MDT to complete startup.
1240 Severity : enhancement
1242 Description: parallel, asynchronous orphan cleanup
1243 Details : orphan cleanup is now performed in separate threads for each
1244 OST, allowing parallel non-blocking operation.
1246 Severity : enhancement
1248 Description: optimized stripe assignment
1249 Details : stripe assignments are now made based on ost space available,
1250 ost previous usage, and OSS previous usage, in order to try
1251 to optimize storage space and networking resources.
1253 Severity : enhancement
1255 Description: Permanently set tunables
1256 Details : All writable /proc/fs/lustre tunables can now be permanently
1257 set on a per-server basis, at mkfs time or on a live system.
1259 Severity : enhancement
1261 Description: Lustre message v2
1262 Details : Add lustre message format v2.
1264 Severity : enhancement
1266 Description: client OST exclusion list
1267 Details : Clients can be started with a list of OSTs that should be
1268 declared "inactive" for known non-responsive OSTs.
1272 Description: SPEC SFS validation failure on NFS v2 over lustre.
1273 Details : Changes the blocksize for regular files to be 2x RPC size,
1274 and not depend on stripe size.
1276 Severity : enhancement
1278 Description: Multiple MD RPCs in flight.
1279 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1280 To avoid overly-overloading MDS, introduce a limit on number of
1281 MDS RPCs in flight for a single client and add /proc controls
1282 to adjust this limit.
1284 Severity : enhancement
1286 Description: client read/write statistics
1287 Details : Add client read/write call usage stats for performance
1288 analysis of user processes.
1289 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1290 file access. extents_stats shows chunk size distribution.
1291 extents_stats_per_process show chunk size distribution per
1294 Severity : enhancement
1296 Description: mds statistics
1297 Details : Add detailed mds operations statistics in
1298 /proc/fs/lustre/mds/*/stats.
1302 Description: Failure of copying files with lustre special EAs.
1303 Details : Client side always return success for setxattr call for lustre
1304 special xattr (currently only "trusted.lov").
1309 Description: Refcount LNET uuids
1310 Details : The global LNET uuid list grew linearly with every startup;
1311 refcount repeated list entries instead of always adding to
1314 Severity : enhancement
1316 Description: Dynamic service threads
1317 Details : Within a small range, start extra service threads
1318 automatically when the request queue builds up.
1321 Frequency : mixed-endian client/server environments
1323 Description: mixed-endian crashes
1324 Details : The new msg_v2 system had some failures in mixed-endian
1327 Severity : enhancement
1329 Description: Easy OST removal
1330 Details : OSTs can be permanently deactivated with e.g. 'lctl
1331 conf_param lustre-OST0001.osc.active=0'
1333 Severity : enhancement
1335 Description: MGS proc entries
1336 Details : Added basic proc entries for the MGS showing what filesystems
1339 Severity : enhancement
1341 Description: provide MGS failover
1342 Details : Added config lock reacquisition after MGS server failover.
1344 Severity : enhancement
1346 Description: add Linux 2.4 support
1347 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1351 Description: a large application tries to do I/O to the same resource and dies
1352 in the middle of it.
1353 Details : Check the req->rq_arrival time after the call to
1354 ost_brw_lock_get(), but before we do anything about
1355 processing it & sending the BULK transfer request. This
1356 should help move old stale pending locks off the queue as
1357 quickly as obd_timeout.
1360 Frequency : when an incorrect nid is specified during startup
1362 Description: ptlrpc connect to non-existant node causes kernel crash
1363 Details : LNET can't be re-entered from an event callback, which
1364 happened when we expire a message after the export has been
1365 cleaned up. Instead, hand the zombie cleanup off to another
1368 Severity : enhancement
1370 Description: plain/inodebits lock performance improvement
1371 Details : Grouping plain/inodebits in granted list by their request modes
1372 and bits policy, thus improving the performance of search through
1376 Frequency : only if OST filesystem is corrupted
1378 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1379 Details : for a short time RPCs with bulk IO are in the replay list,
1380 but replay of bulk IOs is unimplemented. If the OST filesystem
1381 is corrupted due to disk cache incoherency and then replay is
1382 started it is possible to trip an assertion. Avoid putting
1383 committed RPCs into the replay list at all to avoid this issue.
1386 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1388 Description: System hang on startup
1389 Details : This bug allowed the liblustre (e.g. catamount) client to
1390 return to the app before handling all startup RPCs. This
1391 could leave the node unresponsive to lustre network traffic
1392 and manifested as a server ptllnd timeout.
1394 Severity : enhancement
1396 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1397 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1398 boolean to control whether to print peer debug info when a
1399 client's RPC times out.
1402 Frequency : only for kernels with patches from Lustre below 1.4.3
1404 Description: Remove old rdonly API
1405 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1408 Frequency : only for devices with external journals
1410 Description: Set external device read-only also
1411 Details : During a commanded failover stop, we set the disk device
1412 read-only while the server shuts down. We now also set any
1413 external journal device read-only at the same time.
1416 Frequency : when upgrading from 1.4 while trying to change parameters
1418 Description: The wrong (new) MDC name was used when setting parameters for
1419 upgraded MDT's. Also allows changing of OSC (and MDC)
1420 parameters if --writeconf is specified at tunefs upgrade time.
1423 Frequency : when setting specific ost indicies
1425 Description: QOS code breaks on skipped indicies
1426 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1427 created with --index need not be sequential.
1429 Severity : enhancement
1431 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1432 Details : The uninit_groups feature works in conjunction with the kernel
1433 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1434 up the pass1 processing of e2fsck. This is a read-only feature
1435 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1436 mount filesystems that have had this feature enabled.
1438 Severity : enhancement
1440 Description: Improve multi-block allocation algorithm to avoid fragmentation
1441 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1442 allocation locality and avoid filesystem fragmentation.
1444 ------------------------------------------------------------------------------
1446 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1448 * Support for kernels:
1449 2.4.21-47.0.1.EL (RHEL 3)
1450 2.6.5-7.283 (SLES 9)
1451 2.6.9-42.0.10.EL (RHEL 4)
1452 2.6.12.6 vanilla (kernel.org)
1453 2.6.16.27-0.9 (SLES 10)
1454 * Recommended e2fsprogs version: 1.39.cfs5
1456 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1460 Frequency : occasional, depends on client load and configuration
1461 Bugzilla : 12181, 12203
1462 Description: data loss for recently-modified files
1464 Details : In some cases it is possible that recently written or created
1465 files may not be written to disk in a timely manner (this should
1466 normally be within 30s unless client IO load is very high).
1467 The problem appears as zero-length files or files that are a
1468 multiple of 1MB in size after a client crash or client eviction
1469 that are missing data at the end of the file.
1471 This problem is more likely to be hit on clients where files are
1472 repeatedly created and unlinked in the same directory, clients
1473 have a large amount of RAM, have many CPUs, the filesystem has
1474 many OSTs, the clients are rebooted frequently, and/or the files
1475 are not accessed by other nodes after being written.
1477 The presence of the problem can be detected by looking at
1478 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1479 smaller than the second (nr_unused) then dirty files will not
1480 be flushed automatically to disk. "sync; sleep 10" should be
1481 run several times on the node before unmounting it to update
1482 Lustre (this is also safe to run on nodes without this problem).
1484 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1485 that can cause this same problem, so customers using that kernel
1486 also need to update the kernel in addition to Lustre. In order
1487 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1489 It is normal that files written just before a client crash (less
1490 than 30s) may not yet have been flushed to disk, even for local
1494 Frequency : frequent on thin XT3 nodes
1496 Description: UUID collision on thin XT3 Linux nodes
1497 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1498 randomly, since we relied on an insufficiently-seeded PRNG.
1503 Description: OSS hangs after "All ost request buffers busy"
1504 Details : A deadlock between quota and journal operations caused OSS
1505 hangs after printing "All ost request buffers busy."
1508 Frequency : always on liblustre builds
1510 Description: Cleanup compiler warnings on liblustre
1513 Frequency : always on liblustre builds on XT3
1515 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1519 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1520 > 32000 subdirectories
1521 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1522 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1523 subdirectory count crosses 32000. This will aid e2fsck to
1524 correctly handle more than 32000 subdirectories.
1527 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1529 Description: System hang on startup
1530 Details : This bug allowed the liblustre (e.g. catamount) client to
1531 return to the app before handling all startup RPCs. This
1532 could leave the node unresponsive to lustre network traffic
1533 and manifested as a server ptllnd timeout.
1535 Severity : enhancement
1537 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1538 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1539 boolean to control whether to print peer debug info when a
1540 client's RPC times out.
1545 Description: make O_SYNC working on 2.6 kernels
1546 Details : 2.6 kernels use different method for mark pages for write,
1547 so need add a code to lustre for O_SYNC work.
1552 Description: Failure to close file and release space on NFS
1553 Details : Put inode details into lock acquired in ll_intent_file_open.
1554 Use mdc_intent_lock in ll_intent_open to properly
1555 detect all kind of errors unhandled by mdc_enqueue.
1560 Description: proc file read during shutdown sometimes raced obd removal,
1562 Details : Add lock to prevent obd access after proc file removal.
1565 Frequency : Only for files larger than 4GB on 32-bit clients.
1567 Description: improperly doing page alignment of locks
1568 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1569 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1574 Description: RPCs being resent when they shouldn't be
1575 Details : Some RPCs that should not be resent are being resent. This
1576 can cause inconsistencies in the RPC state machine. Do not
1577 resend such requests.
1580 Frequency : rare, only with NFS export
1582 Description: Crash on NFS re-export node
1583 Details : under very unusual load conditions an assertion is hit in
1584 ll_intent_file_open()
1587 Frequency : only if OST filesystem is corrupted
1589 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1590 Details : for a short time RPCs with bulk IO are in the replay list,
1591 but replay of bulk IOs is unimplemented. If the OST filesystem
1592 is corrupted due to disk cache incoherency and then replay is
1593 started it is possible to trip an assertion. Avoid putting
1594 committed RPCs into the replay list at all to avoid this issue.
1599 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1600 Details : Large single O_DIRECT read and write calls can fail to allocate
1601 a sufficiently large buffer to process the request. In case of
1602 allocation failure the allocation is retried with a smaller
1603 buffer and broken into smaller requests.
1605 Severity : enhancement
1607 Description: Add -o localflock option to simulate old noflock behaviour.
1608 Details : This will achieve local-only flock/fcntl locks coherentness.
1613 Description: versioning check is incomplete
1614 Details : Checking the version difference of client vs. server, report
1615 error if the gap is too big.
1620 Description: add support PG_writeback bit
1621 Details : add support for PG_writeback bit for Lustre, for more carefull
1622 work with page cache in 2.6 kernel. This also fix some deadlocks
1623 and remove hack for work O_SYNC with 2.6 kernel.
1625 Severity : enhancement
1627 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1628 Details : The uninit_groups feature works in conjunction with the kernel
1629 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1630 up the pass1 processing of e2fsck. This is a read-only feature
1631 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1632 mount filesystems that have had this feature enabled.
1634 Severity : enhancement
1636 Description: Improve multi-block allocation algorithm to avoid fragmentation
1637 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1638 allocation locality and avoid filesystem fragmentation.
1640 ------------------------------------------------------------------------------
1642 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1644 * Support for kernels:
1645 2.6.9-42.0.3.EL (RHEL 4)
1646 2.6.5-7.276 (SLES 9)
1647 2.4.21-47.0.1.EL (RHEL 3)
1648 2.6.12.6 vanilla (kernel.org)
1649 2.6.16.21-0.8 (SLES10)
1650 * Recommended e2fsprogs version: 1.39.cfs2-0
1652 * The backwards-compatible /proc/sys/portals symlink has been removed
1653 in this release. Before upgrading, please ensure that you change
1654 any configuration scripts or /etc/sysctl.conf files that access
1655 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1656 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1657 in advance of the upgrade on any system running Lustre 1.4.6 or
1658 newer, since /proc/sys/lnet was added in that version.
1659 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1663 Frequency : only when quota is used
1665 Description: avoid scanning export list for quota master
1666 Details : Change the algorithms to avoid scanning export list in order
1667 to improve the efficiency.
1670 Frequency : MDS failover only, very rarely
1672 Description: "went back in time" messages on mds failover
1673 Details : The greatest transno may be lost when the current operation
1674 finishes with an error (transno==0) and the client's last_rcvd
1675 record is over-written. Save the greatest transno in the
1676 mds_last_transno for this case.
1679 Frequency : always for specific kernels and striping counts
1681 Description: client may get "Matching packet too big" without ACL support
1682 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1683 when trying to access files in certain configurations. The
1684 clients should in fact be denied when mounting because they do
1685 not understand ACLs.
1688 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1690 Description: many clients connecting with IO in progress causes connect timeouts
1691 Details : Avoid synchronous journal commits to avoid delays caused by many
1692 clients connecting/disconnecting when bulk IO is in progress.
1693 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1694 OST_IO_PORTAL to avoid delays behind potentially many pending
1698 Frequency : occasionally with multiple writers to a single file
1700 Description: shared writes to file may result in wrong size reported by stat()
1701 Details : Allow growing of kms when extent lock is cancelled
1704 Frequency : always with random mmap IO to multi-striped file
1706 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1707 Details : Only if the hole is at the end of OST object so that kms is too
1708 small. Fix is to increase kms accordingly in ll_nopage.
1711 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1713 Description: writes to a missing object would leak memory on the OST
1714 Details : If there is an inconsistency between the MDS and OST filesystems,
1715 such that the MDS references an object that doesn't exist, writes
1716 to that object will leak memory due to incorrect cleanup in the
1717 error handling path, eventually running out of memory on the OST.
1722 Description: Creating too long symlink causes lustre errors
1723 Details : Check symlink and name lengths before sending requests to MDS.
1726 Frequency : only if flock is enabled (not on by default)
1728 Description: posix locks not released on fd closure on 2.6.9+
1729 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1730 kernels, this caused such locks not to be released on fd close and
1731 then assertions on fs unmount about still used locks.
1734 Frequency : MDS failover only, very rarely
1736 Description: clients may get ASSERTION(granted_lock != NULL)
1737 Details : When request was taking a long time, and a client was resending
1738 a getattr by name lock request. The were multiple lock requests
1739 with the same client lock handle and
1740 mds_getattr_name->fixup_handle_for_resent_request found one of the
1741 lock handles but later failed with ASSERTION(granted_lock != NULL).
1746 Description: handle->h_buffer_credits > 0, assertion failure
1747 Details : h_buffer_credits is zero after truncate, causing assertion
1748 failure. This patch extends the transaction or creates a new
1752 Frequency : NFS re-export or patchless client
1753 Bugzilla : 11179, 10796
1754 Description: Crash on NFS re-export node (__d_move)
1755 Details : We do not want to hash the dentry if we don't have a lock.
1756 But if this dentry is later used in d_move, we'd hit uninitialised
1757 list head d_hash, so we just do this to init d_hash field but
1758 leave dentry unhashed.
1761 Frequency : NFS re-export or patchless client
1763 Description: NFS exports has problem with symbolic link
1764 Details : lustre client didn't properly install dentry when re-exported
1765 to NFS or running patchless client.
1768 Frequency : NFS re-export or patchless client
1770 Description: Various nfs/patchless fixes.
1771 Details : fixes reuse disconected alias for lookup process - this fixes
1772 warning "find_exported_dentry: npd != pd",
1773 fix permission error with open files at nfs.
1774 fix apply umask when do revalidate.
1777 Frequency : occasional
1779 Description: Crash on NFS re-export node
1780 Details : calling clear_page() on the wrong pointer triggered oops in
1781 generic_mapping_read().
1784 Frequency : rarely, using O_DIRECT IO
1786 Description: unaligned directio crashes client with LASSERT
1787 Details : check for unaligned buffers before trying any requests.
1790 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1792 Description: stale data returned from RAID cache
1793 Details : If only a small amount of IO is done to the RAID device before
1794 reading it again it is possible to get stale data from the RAID
1795 cache instead of reading it from disk.
1798 Frequency : always for sles10 kernel
1800 Description: sles10 support
1801 Details : ll_follow_link: compile fixes and using of nd_set_link
1802 under newer kernels.
1805 Frequency : depends on arch, kernel and compiler version, always on sles10
1808 Description: recursive or deep enough symlinks cause stack overflow
1809 Details : getting rid of large stack-allocated variable in
1813 Frequency : depends on hardware
1815 Description: lustre write performance loss in the SLES10 kernel
1816 Details : the performance loss is caused by using of write barriers in the
1817 ext3 code. The SLES10 kernel turns barrier support on by
1818 default. The fix is to undo that change for ldiskfs.
1820 ------------------------------------------------------------------------------
1822 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1824 * Support for kernels:
1825 2.6.9-42.0.3EL (RHEL 4)
1826 2.6.5-7.276 (SLES 9)
1827 2.4.21-47.0.1.EL (RHEL 3)
1828 2.6.12.6 vanilla (kernel.org)
1832 Frequency : quota enabled and large files being deleted
1834 Description: releasing more than 4GB of quota at once hangs OST
1835 Details : If a user deletes more than 4GB of files on a single OST it
1836 will cause the OST to spin in an infinite loop. Release
1837 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1842 Description: statfs data retrieved from /proc may be stale or zero
1843 Details : When reading per-device statfs data from /proc, in the
1844 {kbytes,files}_{total,free,avail} files, it may appear
1845 as zero or be out of date.
1848 Frequency : systems with MD RAID1 external journal devices
1850 Description: lconf's call to blkid is confused by RAID1 journal devices
1851 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1852 of returning all block devices that match the journal UUID.
1855 Frequency : always, for aggregate stripe size over 4GB
1857 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1858 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1859 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1862 Frequency : always if "lfs find" used on a local file/directory
1864 Description: "lfs find" segfaults if used on a local file/directory
1865 Details : The case where a directory component was not specified wasn't
1866 handled correctly. Handle this properly.
1869 Frequency : always on ppc64
1871 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1872 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1873 perform endianness conversion.
1876 Frequency : rarely (truncate to non-zero file size after write under load)
1877 Bugzilla : 10730, 10687
1878 Description: Files padded with zeros to next 4K multiple
1879 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1880 it is possible that files that are truncated to a non-zero size
1881 immediately after being written are filled with zero bytes beyond
1882 the truncated size. No file data is lost.
1884 Severity : enhancement
1886 Description: Allow recovery/failover for liblustre clients.
1887 Details : liblustre clients were unaware of failover configurations until
1890 Severity : enhancement
1892 Description: user file locks should fail when not mounting with flock option
1893 Details : Set up an error-returning stub in ll_file_operations.lock field
1894 to prevent incorrect behaviour when client is mounted without
1895 flock option. Also, set up properly f_op->flock field for
1899 Frequency : always on ia64
1901 Description: "lfs df" loops on printing out MDS statfs information
1902 Details : The obd_ioctl_data was not initialized and in some systems
1903 this caused a failure during the ioctl that did not return
1904 an error. Initialize the struct and return an error on failure.
1907 Frequency : SLES 9 only
1909 Description: Error of copying files with lustre special EAs as root
1910 Details : Client side always return success for setxattr call for lustre
1911 special xattr (currently only "trusted.lov").
1914 Frequency : rarely on clusters with both ia64+i386 clients
1916 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1917 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1918 clients are doing concurrent writes to the same file it is
1919 possible that the ia64 clients may LASSERT because the OST
1920 extent locks are not PAGE_SIZE aligned. Ensure that grown
1921 locks are always aligned on the request boundary.
1924 Frequency : specific use, occasional
1926 Description: Overwriting in use executable truncates on-disk binary image
1927 Details : If one node attempts to overwrite an executable in use by
1928 another node, we now correctly return ETXTBSY instead of
1929 truncating the file.
1931 Severity : enhancement
1933 Description: Async OSC create to avoid the blocking unnecessarily.
1934 Details : If a OST has no remain object, system will block on the creating
1935 when need to create a new object on this OST. Now, ways use
1936 pre-created objects when available, instead of blocking on an
1937 empty osc while others are not empty. If we must block, we block
1938 for the shortest possible period of time.
1943 Description: chmod on Lustre root is propagated to other clients
1944 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1945 invalid by the follow_mount time.
1948 Frequency : liblustre clients only
1950 Description: Race in 'instant cancel' lock handling could lead to such locks
1951 never to be granted in case of SMP MDS
1952 Details : Do not destroy not yet granted but cbpending locks in
1956 Frequency : replay/resend of open
1958 Description: non null lock assetion failure in mds_intent_policy
1959 Details : Trying to replay/resend lockless open requests resulted in
1960 mds_open() returning 0 with no lock. Now it sets a flag if
1961 a lock is going to be returned.
1963 Severity : enhancement
1965 Description: Checksum enhancements
1966 Details : New checksum enhancements allow for resending RPCs that failed
1969 Severity : enhancement
1971 Description: Tunables on number of dirty pages in cacche
1972 Details : Allow to set limit on number of dirty pages cached.
1977 Description: client crash on unmount - lock still has references
1978 Details : In some error handling cases it was possible to leak a lock
1979 reference on a client while accessing a file. This was not
1980 harmful to the client during operation, but would cause the
1981 client to crash when the filesystem is unmounted.
1984 Frequency : specific case, rare
1986 Description: ETXTBSY on mds though file not in use by client
1987 Details : ETXTBSY is no longer incorrectly returned when attempting to
1988 chmod or chown a directory that the user previously tried to
1989 execute or a currently-executing binary.
1992 Frequency : extremely rare except on liblustre-based clients
1994 Description: Lustre space not freed when files are deleted
1995 Details : Clean up open-unlinked files after client eviction. Previously
1996 the unlink was skipped and the files remained as orphans.
2001 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2002 Details : In some cases it was possible to send a blocking callback to
2003 a client doing a glimpse, even though that client didn't get
2004 a lock granted. When the glimpse lock is cancelled on the OST
2005 the freed lock is left on the waiting list and corrupted the list.
2008 Frequency : all core dumps
2010 Description: Broke core dumps to lustre
2011 Details : Negative dentry may be unhashed if parent does not have UPDATE
2012 lock, but some callers, e.g. do_coredump, expect dentry to be
2013 hashed after successful create, hash it in ll_create_it.
2015 ------------------------------------------------------------------------------
2017 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2019 * Support for kernels:
2020 2.6.9-42.0.2.EL (RHEL 4)
2021 2.6.5-7.276 (SLES 9)
2022 2.4.21-40.EL (RHEL 3)
2023 2.6.12.6 vanilla (kernel.org)
2027 Frequency : always on RHEL 3
2029 Description: Number of open files grows over time
2030 Details : The number of open files grows over time, whether or not
2031 Lustre is started. This was due to a filp leak introduced
2032 by one of our kernel patches.
2034 ------------------------------------------------------------------------------
2036 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2038 * Support for kernels:
2039 2.6.9-42.EL (RHEL 4)
2040 2.6.5-7.276 (SLES 9)
2041 2.4.21-40.EL (RHEL 3)
2042 2.6.12.6 vanilla (kernel.org)
2047 Bugzilla : 5719, 9635, 9792, 9684
2048 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2049 Details : If a server is under heavy load and cannot reply to new
2050 connection requests before the client resends the (re)connect,
2051 the connection handling code can behave badly if two service
2052 threads are concurrently handing separate (re)connections from
2053 the same client. Add better locking to the connection handling
2054 code, and ensure that only a single connection will be processed
2055 for a given client UUID, even if the lock is dropped.
2057 Severity : enhancement
2059 Description: add TCP zero-copy support to kernel
2060 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2061 sends if the hardware supports scatter-gather and checksumming.
2062 This allows socklnd to do client-write and server-read more
2063 efficiently and reduce CPU utilization from skbuf copying.
2066 Frequency : only if NFS exporting from client
2068 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2069 Details : If setting attributes on a file created under NFS that had
2070 never been opened it would be possible to oops the client
2071 if the file had no objects.
2074 Frequency : always for liblustre
2076 Description: liblustre client does MDS+OSTs setattr RPC for each write
2077 Details : When doing a write from a liblustre client, the client
2078 incorrectly issued an RPC to the MDS and each OST the file was
2079 striped over in order to update the timestamps. When writing
2080 with small chunks and many clients this could overwhelm the MDS
2081 with RPCs. In all cases it would slow down the write because
2082 these RPCs are unnecessary.
2084 Severity : enhancement
2086 Description: allow number of MDS service threads to be changed at module load
2087 Details : It is now possible to change the number of MDS service threads
2088 running. Adding "options mds mds_num_threads={N}" to the MDS's
2089 /etc/modprobe.conf will set the number of threads for the next
2090 time Lustre is restarted (assuming the "mds" module is also
2091 reloaded at that time). The default number of threads will
2092 stay the same, 32 for most systems.
2097 Description: OST crash if filesystem is unformatted or corrupt
2098 Details : If an OST is started on a device that has never been formatted
2099 or if the filesystem is corrupt and cannot even mount then the
2100 error handling cleanup routines would dereference a NULL pointer.
2105 Description: NULL pointer deref in llap_from_page.
2106 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2107 incorrect) mapping if the page was truncated/reclaimed while it was
2108 searched for. Check for this condition and skip such pages when
2109 doing readahead. Introduce extra check to llap_from_page() to
2110 verify page->mapping->host is non-NULL (so page is not anonymous).
2113 Frequency : Sometimes when using sys_sendfile
2115 Description: "page not covered by a lock" warnings from ll_readpage
2116 Details : sendfile called ll_readpage without right page locks present.
2117 Now we introduced ll_file_sendfile that does necessary locking
2118 around call to generic_file_sendfile() much like we do in
2122 Frequency : with certain MDS communication failures at client mount time
2124 Description: NULL pointer deref after failed client mount
2125 Details : a client connection request may delayed by the network layer
2126 and not be sent until after the PTLRPC layer has timed out the
2127 request. If the client fails the mount immediately it will try
2128 to clean up before the network times out the request. Add a
2129 reference from the request import to the obd device and delay
2130 the cleanup until the network drops the request.
2133 Frequency : occasionally during client (re)connect
2135 Description: assertion failure during client (re)connect
2136 Details : processing a client connection request may be delayed by the
2137 client or server longer than the client connect timeout. This
2138 causes the client to resend the connection request. If the
2139 original connection request is replied in this interval, the
2140 client may trip an assertion failure in ptlrpc_connect_interpret()
2141 which thought it would be the only running connect process.
2144 Frequency : only with obd_echo servers and clients that are rebooted
2146 Description: kernel BUG accessing uninitialized data structure
2147 Details : When running an obd_echo server it did not start the ping_evictor
2148 thread, and when a client was evicted an uninitialized data
2149 structure was accessed. Start the ping_evictor in the RPC
2150 service startup instead of the OBD startup.
2152 Severity : enhancement
2153 Bugzilla : 10193 (patchless)
2154 Description: Remove dependency on various unexported kernel interfaces.
2155 Details : No longer need reparent_to_init, exit_mm, exit_files,
2156 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2159 Frequency : rare (only users of deprecated and unsupported LDAP config)
2161 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2162 Details : LDAP apparently contains 'lustreName' attributes instead of
2163 'name'. A simple remapping of the name is sufficient.
2166 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2168 Description: waiting_locks_callback trips kernel BUG if client is evicted
2169 Details : Running with the dump_on_timeout debug flag turned on makes
2170 it possible that the waiting_locks_callback() can try to dump
2171 the Lustre kernel debug logs from an interrupt handler. Defer
2172 this log dumping to the expired_lock_main() thread.
2174 Severity : enhancement
2176 Description: Support NFS exporting on 2.6 kernels.
2177 Details : Implement non-rawops metadata methods for NFS server to use without
2178 changing NFS server code.
2181 Frequency : very rare (synthetic metadata workload only)
2183 Description: two racing renames might cause an MDS thread to deadlock
2184 Details : Running the "racer" program may cause one MDS thread to rename
2185 a file from being the source of a rename to being the target of
2186 a rename at exactly the same time that another thread is doing
2187 so, and the second thread has already enqueued these locks after
2188 doing a lookup of the target and is trying to relock them in
2189 order. Ensure that we don't try to re-lock the same resource.
2192 Frequency : only very large systems with liblustre clients
2194 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2195 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2196 all OSTs in parallel. This allows the eviction of stale liblustre
2197 clients to proceed much faster than if they were done in series,
2198 and also offers similar improvements for other set_info RPCs.
2203 Description: excessive CPU usage during initial read phase on client
2204 Details : During the initial read phase on a client, it would agressively
2205 retry readahead on the file, consuming too much CPU and impacting
2206 performance (since 1.4.5.8). Improve the readahead algorithm
2207 to avoid this, and also improve some other common cases (read
2208 of small files in particular, where "small" is files smaller than
2209 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2214 Description: MDS crash when receiving packet with unknown intent.
2215 Details : Do not LBUG in unknown intent case, just return -EFAULT
2217 Severity : enhancement
2218 Bugzilla : 9293, 9385
2219 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2220 Details : Do not serialize getattr (non-intent version) and statfs.
2223 Frequency : occasional, when OST network is overloaded/intermittent
2225 Description: client evicted by OST after bulk IO timeout
2226 Details : If a client sends a bulk IO request (read or write) the OST
2227 may evict the client if it is unresposive to its data GET/PUT
2228 request. This is incorrect if the network is overloaded (takes
2229 too long to transfer the RPC data) or dropped the OST GET/PUT
2230 request. There is no need to evict the client at all, since
2231 the pinger and/or lock callbacks will handle this, and the
2232 client can restart the bulk request.
2235 Frequency : Always when mmapping file with no objects
2237 Description: client crashes when mmapping file with no objects
2238 Details : Check that we actually have objects in a file before doing any
2239 operations on objects in ll_vm_open, ll_vm_close and
2245 Description: Request leak when working with deleted CWD
2246 Details : Introduce advanced request refcount tracking for requests
2247 referenced from lustre intent.
2249 Severity : Enhancement
2251 Description: Cache open file handles on client.
2252 Details : MDS now will return special lock along with openhandle, if
2253 requested and client is allowed to hold openhandle, even if unused,
2254 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2255 closing files for every read/write openration.
2257 Severity : Enhancement
2259 Description: Cache open negative dentries on client when possible.
2260 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2261 negative dentries on lock revocation.
2266 Description: Remounting a client read-only wasn't possible with a zconf mount
2267 Details : It wasn't possible to remount a client read-only with llmount.
2269 Severity : enhancement
2270 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2271 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2272 MPICH is included to add Lustre-specific ADIO interfaces.
2273 This is based closely on the UFS ADIO layer and only differs
2274 in file creation, in order to allow the OST striping to be set.
2275 This is user-contributed code and not supported by CFS.
2280 Description: extended inode attributes (immutable, append-only) work improperly
2281 when 2.4 and 2.6 kernels are used on client/server or vice versa
2282 Details : Introduce kernel-independent values for these flags.
2284 Severity : enhancement
2287 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2288 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2289 granularity. Now, Fractional MB granularity is be supported,
2290 this is very useful for low memory system.
2292 Severity : enhancement
2294 Description: Getattr by fid
2295 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2296 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2297 have locks and avoids getattr rpc requests for it on every stat.
2300 Frequency : Always, for filesystems larger than 2TB
2302 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2303 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2304 filesystems larger than 2TB to be mounted with mballoc enabled.
2307 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2309 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2310 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2311 filesystems larger than 2TB is mounted then the kernel will
2312 silently corrupt the start of the filesystem. CONFIG_LBD is
2313 enabled for all CFS-supported kernels, but the possibility of
2314 this happening with a modified kernel config exists.
2316 Severity : enhancement
2318 Description: add client O_DIRECT support for 2.6 kernels
2319 Details : It is now possible to do O_DIRECT reads and writes to files
2320 in the Lustre client mountpoint on 2.6 kernel clients.
2322 Severity : enhancement
2324 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2325 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2326 parallel, not waiting for response from every OST before sending
2327 a rpc to the next OST.
2332 Description: setattr vs write race when updating file timestamps
2333 Details : Client processes that update a file timestamp into the past
2334 right after writing to the file (e.g. tar) it is possible that
2335 the updated file modification time can be reset to the current
2336 time due to a race between processing the setattr and write RPC.
2338 Severity : enhancement
2340 Description: Bring 'lfs find' closer in line with regular Linux find.
2341 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2342 -print0 options and obtains all the needed info through the lustre
2345 Severity : enhancement
2347 Description: support up to 1024 configured devices on one node
2348 Details : change obd_dev array from statically allocated to dynamically
2349 allocated structs as they are first used to reduce memory usage
2354 Description: Flush dirty partially truncated pages during truncate
2355 Details : Immediatelly flush partially truncated pages in filter_setattr,
2356 this way we completely avoid having any pages in page cache on OST
2357 and can retire ugly workarounds during writes to flush such pages.
2362 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2363 Details : protect inode from truncation within vfs_unlink() context
2364 just take a reference before calling vfs_unlink() and release it
2365 when parent's i_sem is free.
2370 Description: last_id value checked outside lock on OST caused LASSERT failure
2371 Details : If there were multiple MDS->OST object precreate requests in
2372 flight, it was possible that the OST's last object id was checked
2373 outside a lock and incorrectly tripped an assertion. Move checks
2374 inside locks, and discard old precreate requests.
2377 Frequency : always, if extents are used on OSTs
2379 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2380 Details : OSTs using the extents format would not zero the high 16 bits of
2381 the index physical block number. This is not a problem for any
2382 OST filesystems smaller than 16TB, and no kernels support ext3
2383 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2384 new/modified files) and can be fixed for existing filesystems
2385 with e2fsprogs-1.39-cfs1.
2390 Description: import connection selection may be incorrect if timer wraps
2391 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2392 import connections to be ignored if the 32-bit jiffies counter
2393 wraps. Use a 64-bit jiffies counter.
2396 Frequency : very large clusters immediately after boot
2398 Description: LNET request buffers exhausted under heavy short-term load
2399 Details : If a large number of client requests are generated on a service
2400 that has previously never seen so many requests it is possible
2401 that the request buffer growth cannot keep up with the spike in
2402 demand. Instead of dropping incoming requests, they are held in
2403 the LND until the RPC service can accept more requests.
2406 Frequency : Sometimes during replay
2408 Description: Assertion failure in ll_local_open after replay.
2409 Details : If replay happened on an open request reply before we were able
2410 to set replay handler, reply will become not swabbed tripping the
2411 assertion in ll_local_open. Now we set the handler right after
2412 recognising of open request
2415 Frequency : very rare
2417 Description: kernel reports "badness in vsnprintf"
2418 Details : Reading from the "recovery_status" /proc file in small chunks
2419 may cause a negative length in lprocfs_obd_rd_recovery_status()
2420 call to vsnprintf() (which is otherwise harmless). Exit early
2421 if there is no more space in the output buffer.
2423 Severity : enhancement
2425 Description: clear OBD RPC statistics by writing to them
2426 Details : It is now possible to clear the OBD RPC statistics by writing
2427 to the "stats" file.
2432 Description: Client mtime is not the same on different clients after utimes
2433 Details : In some cases, the client was using the utimes() syscall on
2434 a file cached on another node. The clients now validate the
2435 ctime from the MDS + OSTs to determine which one is right.
2440 Description: Inability to activate failout mode
2441 Details : lconf script incorrectly assumed that in pythong string's numeric
2442 value is used in comparisons.
2445 Frequency : always with multiple stripes per file
2447 Description: Inefficient object allocation for mutli-stripe files
2448 Details : When selecting which OSTs to stripe files over, for files with
2449 a stripe count that divides evenly into the number of OSTs,
2450 the MDS is always picking the same starting OST for each file.
2451 Return the OST selection heuristic to the original design.
2456 Description: mount failures may take full timeout to return an error
2457 Details : Under some heavy load conditions it is possible that a
2458 failed mount can wait for the full obd_timeout interval,
2459 possibly several minutes, before reporting an error.
2460 Instead return an error as soon as the status is known.
2462 Frequency : quota enabled and large files being deleted
2464 Description: releasing more than 4GB of quota at once hangs OST
2465 Details : If a user deletes more than 4GB of files on a single OST it
2466 will cause the OST to spin in an infinite loop. Release
2467 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2472 Description: statfs data retrieved from /proc may be stale or zero
2473 Details : When reading per-device statfs data from /proc, in the
2474 {kbytes,files}_{total,free,avail} files, it may appear
2475 as zero or be out of date.
2478 Frequency : systems with MD RAID1 external journal devices
2480 Description: lconf's call to blkid is confused by RAID1 journal devices
2481 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2482 of returning all block devices that match the journal UUID.
2485 Frequency : always, for aggregate stripe size over 4GB
2487 Description: assertion fails when trying to use 4GB stripe size
2488 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2489 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2492 Frequency : always on ppc64
2494 Description: the first write on an ext3 filesystem with mballoc got stuck
2495 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2496 perform endianness conversion.
2498 ------------------------------------------------------------------------------
2500 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2502 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2503 INTEROPERATE with older versions automatically. Please read the
2504 user documentation before upgrading any part of a live system.
2505 * WARNING: Lustre networking configuration changes are required with
2506 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2509 * Support for kernels:
2510 2.6.9-22.0.2.EL (RHEL 4)
2511 2.6.5-7.244 (SLES 9)
2512 2.6.12.6 vanilla (kernel.org)
2515 Severity : enhancement
2516 Bugzilla : 7981/8208
2517 Description: Introduced Lustre Networking (LNET)
2518 Details : LNET is new networking infrastructure for Lustre, it includes
2519 a reorganized network configuration mode (see the user
2520 documentation for full details) as well as support for routing
2521 between different network fabrics. Lustre Networking Devices
2522 (LNDs) for the supported network fabrics have also been
2523 created for this new infrastructure.
2525 Severity : enhancement
2526 Description: Introduced Access control lists
2527 Details : clients can set ACLs on files and directories in order to have
2528 more fine-grained permissions than the standard Unix UGO+RWX.
2529 The MDS must be started with the "-o acl" mount option.
2531 Severity : enhancement
2532 Description: Introduced filesystem quotas
2533 Details : Administrators may now establish per-user quotas on the
2536 Severity : enhancement
2538 Description: Configuration change for the XT3
2539 The PTLLND is now used to run Lustre over Portals on the XT3
2540 The configure option(s) --with-cray-portals are no longer used.
2541 Rather --with-portals=<path-to-portals-includes> is used to
2542 enable building on the XT3. In addition to enable XT3 specific
2543 features the option --enable-cray-xt3 must be used.
2548 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2549 Details : A race between allocating a new llcd and re-getting the llcd_lock
2550 allowed another thread to grab newly-allocated llcd.
2552 Severity : enhancement
2554 Description: 2.6 OST async journal commit and locking fix to improve performance
2555 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2556 2.6 kernels are now async as they already were in 2.4 kernels so
2557 that they can commit concurrently with the network bulk transfer.
2558 For block-allocated files the filter allocation semaphore is held
2559 to avoid filesystem fragmentation during allocation. BKL lock
2560 removed for 2.6 xattr operations where it is no longer needed.
2565 Description: lconf incorrectly determined whether two IP networks could talk
2566 Details : In some more complicated routing and multiple-network
2567 configurations, lconf will avoid trying to make a network
2568 connection to a disjoint part of the IP space. It was doing the
2569 math incorrectly for one set of cases.
2574 Description: Fix for potential infinite loop processing records in an llog.
2575 Details : If an llog record is corrupted/zeroed, it is possible to loop
2576 forever in llog_process(). Validate the llog record length
2577 and skip the remainder of the block on error.
2580 Frequency : occasional (liblustre only)
2582 Description: liblustre could not open files whose last component is a symlink
2583 Details : sysio_path_walk() would incorrectly pass the open intent to
2584 intermediate path components.
2587 Frequency : rare (liblustre only with non-standard tuning)
2588 Bugzilla : 7201 (7350)
2589 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2590 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2591 after it has been released, only do this for referenced locks
2593 Severity : enhancement
2595 Description: specifying an (invalid) directory default stripe_size of -1
2596 would reset the directory default striping
2597 Details : stripe_size -1 was used internally to signal directory stripe
2598 removal, now use "all default" to signal dir stripe removal
2599 as a directory striping of "all default" is not useful
2602 Frequency : common for large clusters running liblustre clients
2604 Description: doing an ls when liblustre clients are running is slow
2605 Details : sending a glimpse AST to a liblustre client waits for every AST
2606 to time out, as liblustre clients will not respond. Since they
2607 cannot cache data we refresh the OST lock LVB from disk instead.
2609 Severity : enhancement
2611 Description: doing an ls at the same time as file IO can be slow
2612 Details : enqueue and other "small" requests can be blocked behind many
2613 large IO requests. Create a new OST IO portal for non-IO
2614 requests so they can be processed faster.
2617 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2619 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2620 created file with mode 000
2621 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2622 allow an ftruncate() as a non-root user to a file with mode 000.
2623 Lustre now respects this flag to disable mode checks when
2624 truncating a file owned by the user
2627 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2630 Description: Revoking locks from clients that went dead or catatonic might take
2632 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2633 cancellation of such locks instant on servers without waiting for
2634 any reply from clients. Clients drops these locks when cancel
2635 notification from server is received without replying.
2638 Frequency : liblustre-only, when liblustre client dies or becomes busy
2640 Description: Doing ls on Linux clients can take a long time with active
2642 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2643 granting such locks to it in the first place if possible. Locks
2644 are taken by proxy on the OST during the read or write and
2645 dropped immediately afterward. Add connect flags handling, do
2646 not grant locks to liblustre clients for glimpse ASTs.
2648 Severity : enhancement
2650 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2651 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2652 sequential reads and behaves badly with applications doing random
2653 reads. Improve it by reading ahead at least read region, and
2654 avoiding excessive large RPC for small reads.
2656 Severity : enhancement
2658 Description: Creating more than 1000 files for a single job may cause a load
2659 imbalance on the OSTs if there are also a large number of OSTs.
2660 Details : qos_prep_create() uses an OST index reseed value that is an
2661 even multiple of the number of available OSTs so that if the
2662 reseed happens in the middle of the object allocation it will
2663 still utilize the OSTs as uniformly as possible.
2668 Description: OST or MDS may oops in ping_evictor_main()
2669 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2670 but doesn't restart at beginning of obd_exports_timed list
2673 Severity : enhancement
2675 Description: improve by-nid export eviction on the MDS and OST
2676 Details : allow multiple exports with the same NID to be evicted at one
2677 time without re-searching the exports list.
2680 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2682 Description: MDS may oops in groups_free()
2683 Details : in rare race conditions a newly allocated group_info struct is
2684 freed again, and this can be NULL. The 2.4 compatibility code
2685 for groups_free() checked for a NULL pointer, but 2.6 did not.
2688 Frequency : common for liblustre clients doing little filesystem IO
2689 Bugzilla : 9352, 7313
2690 Description: server may evict liblustre clients accessing contended locks
2691 Details : if a client is granted a lock or receives a completion AST
2692 with a blocking AST already set it would not reply to the AST
2693 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2696 Frequency : lfs setstripe, only systems with more than 160 OSTs
2698 Description: unable to set striping with a starting offset beyond OST 160
2699 Details : llapi_create_file() incorrectly limited the starting stripe
2700 index to the maximum single-file stripe count.
2703 Frequency : LDAP users only
2705 Description: lconf did not handle in-kernel recovery with LDAP properly
2706 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2708 Severity : enhancement
2710 Description: bind OST threads to NUMA nodes to improve performance
2711 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2712 node and do their allocations there to localize memory access
2714 Severity : enhancement
2716 Description: llmount can determine client NID directly from Myrinet (GM)
2717 Details : the client NID code from gmnalnid was moved directly into
2718 llmount, removing the need to use this or specifying the
2719 client NID explicitly when mounting GM clients with zeroconf
2722 Frequency : if client is started with down MDS
2724 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2725 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2726 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2727 ptlrpc_import_delayed_req() because we didn't check if the
2728 request was interrupted, and we also didn't break out of the
2729 event loop if there was no timeout
2734 Description: data loss during non-page-aligned writes to a single file from
2735 both multiple nodes and multiple threads on one node at same time
2736 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2737 inconsistency led to false short-reads, that were cached and later
2738 used by ->prepare_write() to fill in partially written page,
2739 leading to data loss.
2742 Frequency : always, if lconf --abort_recovery used
2744 Description: lconf --abort_recovery fails with 'Operation not supported'
2745 Details : lconf was attempting to abort recovery on the MDT device and not
2748 Severity : enhancement
2750 Description: remove cleanup logs
2751 Details : replace lconf-generated cleanup logs with lustre internal
2752 cleanup routines. Eliminates the need for client-cleanup and
2755 Severity : enhancement
2757 Description: add support for EAs (user and system) on lustre filesystems
2758 Details : it is now possible to store extended attributes in the Lustre
2759 client filesystem, and with the user_xattr mount option it
2760 is possible to allow users to store EAs on their files also
2762 Severity : enhancement
2764 Description: Add possibility (config option) to show minimal available OST free
2766 Details : When compiled with --enable-mindf configure option, statfs(2)
2767 (and so, df) will return least minimal free space available from
2768 all OSTs as amount of free space on FS, instead of summary of
2769 free spaces of all OSTs.
2771 Severity : enhancement
2773 Description: do not expand extent locks acquired on OST-side
2774 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2775 by server: they are not cached anyway.
2778 Frequency : when mmap is used/binaries executed from Lustre
2780 Description: Unmmap pages before throwing them away from read cache.
2781 Details : llap_shrink cache now attempts to unmap pages before discarding
2782 them (if unmapping failed - do not discard). SLES9 kernel has
2783 extra checks that trigger if this unmapping is not done first.
2788 Description: lconf didn't resolve symlinks before checking to see whether a
2789 given mountpoint was already in use
2792 Frequency : when migrating failover services
2793 Bugzilla : 6395, 9514
2794 Description: When migrating a subset of services from a node (e.g. failback
2795 from a failover service node) the remaining services would
2796 time out and evict clients.
2797 Details : lconf --force (implied by --failover) sets the global obd_timeout
2798 to 5 seconds in order to quickly disconnect, but this caused
2799 other RPCs to time out too quickly. Do not change the global
2800 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2802 Severity : enhancement
2803 Frequency : if MDS is started with down OST
2804 Bugzilla : 9439,5706
2805 Description: Allow startup/shutdown of an MDS without depending on the
2806 availability of the OSTs.
2807 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2808 Add appropriate locking and lov-osc refcounts for safe
2809 cleaning. Add osc abort_inflight calls in case the
2810 synchronize never started.
2813 Frequency : occasional (Cray XT3 only)
2815 Description: root not authorized to access files in CRAY_PORTALS environment
2816 Details : The client process capabilities were not honoured on the MDS in
2817 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2818 been accessed by an authorized user then root was able to access
2819 the file on the local client also. The root user capabilities
2820 are now allowed on the MDS, as this environment has secure UID.
2823 Frequency : occasional
2825 Description: ldiskfs "too long searching" message happens too often
2826 Details : A debugging message (otherwise harmless) prints too often on
2827 the OST console. This has been reduced to only happen when
2828 there are fragmentation problems on the filesystem.
2833 Description: Division by zero in statfs when all OSCs are inactive
2834 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2835 lov_statfs divides by value returned by lov_get_stripecnt().
2839 Bugzilla : 9489, 3273
2840 Description: First write from each client to each OST was only 4kB in size,
2841 to initialize client writeback cache, which caused sub-optimal
2842 RPCs and poor layout on disk for the first writen file.
2843 Details : Clients now request an initial cache grant at (re)connect time
2844 and so that they can start streaming writes to the cache right
2845 away and always do full-sized RPCs if there is enough data.
2846 If the OST is rebooted the client also re-establishes its grant
2847 so that client cached writes will be honoured under the grant.
2852 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2853 Details : Now I/O RPCs go to different portal number and (presumably) fast
2854 lock requests (and glimses) and other RPCs get their own service
2855 threads pool that should be able to service those RPCs
2858 Severity : enhancement
2860 Description: Ability to exchange lustre version between client and servers and
2861 issue warnings at client side if client is too old. Also for
2862 liblustre clients there is ability to refuse connection of too old
2864 Details : New 'version' field is added to connect data structure that is
2865 filled with version info. That info is later checked by server and
2869 Frequency : rare, liblustre only.
2870 Bugzilla : 9296, 9581
2871 Description: Two simultaneous writes from liblustre at offset within same page
2872 might proceed at the same time overwriting eachother with stale
2874 Details : I/O lock withing llu_file_prwv was released too early, before data
2875 actually was hitting the wire. Extended lock-holding time until
2876 server acknowledges receiving data.
2879 Frequency : extremely rare. Never observed in practice.
2881 Description: avoid generating lustre_handle cookie of 0.
2882 Details : class_handle_hash() generates handle cookies by incrementing
2883 global counter, and can hit 0 occasionaly (this is unlikely, but
2884 not impossible, because initial value of cookie counter is
2885 selected randonly). Value of 0 is used as a sentinel meaning
2886 "unassigned handle" --- avoid it. Also coalesce two critical
2887 sections in this function into one.
2889 Severity : enhancement
2891 Description: allow liblustre clients to delegate truncate locking to OST
2892 Details : To avoid overhead of locking, liblustre client instructs OST to
2893 take extent lock in ost_punch() on client's behalf. New connection
2894 flag is added to handle backward compatibility.
2896 Severity : enhancement
2897 Bugzilla : 4928, 7341, 9758
2898 Description: allow number of OST service threads to be specified
2899 Details : a module parameter allows the number of OST service threads
2900 to be specified via "options ost ost_num_threads={N}" in the
2901 OSS's /etc/modules.conf or /etc/modprobe.conf.
2905 Bugzilla : 6146, 9635, 9895
2906 Description: servers crash with bad pointer in target_handle_connect()
2907 Details : In rare cases when a client is reconnecting it was possible that
2908 the connection request was the last reference for that export.
2909 We would temporarily drop the export reference and get a new
2910 one, but this may have been the last reference and the export
2911 was just destroyed. Get new reference before dropping old one.
2913 Severity : enhancement
2914 Frequency : if client is started with failover MDS
2916 Description: Allow multiple MDS hostnames in the mount command
2917 Details : Try to read the configuration from all specified MDS
2918 hostnames during a client mount in case the "primary"
2921 Severity : enhancement
2923 Description: Stop sending data to evicted clients as soon as possible.
2924 Details : Check if the client we are about to send or are sending data to
2925 was evicted already. (Check is done every second of waiting,
2926 for which l_wait_event interface was extended to allow checking
2927 of exit condition at specified intervals).
2930 Frequency : rare, normally only when NFS exporting is done from client
2932 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2934 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2937 Frequency : always, if srand() is called before liblustre initialization
2939 Description: Liblustre uses system PRNG disturbing its usage by user application
2940 Details : Introduce internal to lustre fast and high-quality PRNG for
2941 lustre usage and make liblustre and some other places in generic
2942 lustre code to use it.
2944 Severity : enhancement
2945 Bugzilla : 9477, 9557, 9870
2946 Description: Verify that the MDS configuration logs are updated when xml is
2947 Details : Check if the .xml configuration logs are newer than the config
2948 logs stored on the MDS and report an error if this is the case.
2949 Request --write-conf, or allow starting with --old_conf.
2951 Severity : enhancement
2953 Description: Handle symlinks in the path when checking if Lustre is mounted.
2954 Details : Resolve intermediate symlinks when checking if a client has
2955 mounted a filesystem to avoid duplicate client mounts.
2960 Description: lconf can hit an error exception but still return success.
2961 Details : The lconf command catches the Command error exception at the top
2962 level script context and will exit with the associated exit
2963 status, but doesn't ensure that this exit status is non-zero.
2968 Description: failure of ptlrpc thread startup can cause oops
2969 Details : Starting a ptlrpc service thread can fail if there are a large
2970 number of threads or the server memory is very fragmented.
2971 Handle this without oopsing.
2974 Frequency : always, only if liblustre and non-default acceptor port was used
2976 Description: liblustre cannot connect to servers with non-default acceptor port
2977 Details : tcpnal_set_default_params() was not called and was therefore
2978 ignoring the environment varaible TCPNAL_PORT, as well as other
2979 TCPNAL_ environment variables
2984 Description: two objects could be created on the same OST for a single file
2985 Details : If an OST is down, in some cases it was possible to create two
2986 objects on a single OST for a single file. No problems other
2987 than potential performance impact and spurious error messages.
2991 Bugzilla : 5681, 9562
2992 Description: Client may oops in ll_unhash_aliases
2993 Details : Client dcache may become inconsistent in race condition.
2994 In some cases "getcwd" can fail if the current directory is
3000 Description: Inode refcounting problems in NFS export code
3001 Details : link_raw functions used to call d_instantiate without obtaining
3002 extra inode reference first.
3006 Bugzilla : 9942, 9903
3007 Description: Referencing freed requests leading to crash, memleaks with NFS.
3008 Details : We used to require that call to ll_revalidate_it was always
3009 followed by ll_lookup_it. Also with revalidate_special() it is
3010 possible to call ll_revalidate_it() twice for the same dentry
3011 even if first occurence returned success. This fix changes semantic
3012 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3013 reference on a request referred from the intent.
3014 ll_intent_release() then releases such a request.
3017 Frequency : rare, normally benchmark loads only
3019 Description: unlinked inodes were kept in memory on the client
3020 Details : If a client is repeatedly creating and unlinking files it
3021 can accumulate a lot of stale inodes in the inode slab cache.
3022 If there is no other client load running this can cause the
3023 client node to run out of memory. Instead flush old inodes
3024 from client cache that have the same inode number as a new inode.
3027 Frequency : SLES9 2.6.5 kernel and long filenames only
3028 Bugzilla : 9969, 10379
3029 Description: utime reports stale NFS file handle
3030 Details : SLES9 uses out-of-dentry names in some cases, which confused
3031 the lustre dentry revalidation. Change it to always use the
3035 Frequency : rare, unless heavy write-truncate concurrency is continuous
3036 Bugzilla : 4180, 6984, 7171, 9963, 9331
3037 Description: OST becomes very slow and/or deadlocked during object unlink
3038 Details : filter_destroy() was holding onto the parent directory lock
3039 while truncating+unlinking objects. For very large objects this
3040 may block other threads for a long time and slow overall OST
3041 responsiveness. It may also be possible to get a lock ordering
3042 deadlock in this case, or run out of journal credits because of
3043 the combined truncate+unlink. Solution is to do object truncate
3044 first in one transaction without parent lock, and then do the
3045 final unlink in a new transaction with the parent lock. This
3046 reduces the lock hold time dramatically.
3049 Frequency : rare, 2.4 kernels only
3051 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3052 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3053 before unmounting their filesystem in order to avoid deadlock.
3054 The kernel_locked() function in 2.4 kernels only checks whether
3055 the kernel lock is held, not whether it is this process that is
3056 holding it as 2.6 kernels do.
3061 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3062 Details : The client ptlrpc code may be trying to reconnect to a down
3063 server before a previous connection attempt has timed out.
3064 Increase the reconnect interval to be longer than the connection
3065 timeout interval to avoid sending duplicate connections to
3069 Frequency : echo_client brw_test command
3071 Description: fix echo_client to work with OST preallocated code
3072 Details : OST preallocation code (5137) didn't take echo_client IO path
3073 into account: echo_client calls filter methods outside of any
3074 OST thread and, hence, there is no per-thread preallocated
3075 pages and buffers to use. Solution: hijack pga pages for IO. As
3076 a byproduct, this avoids unnecessary data copying.
3080 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3081 Description: Client can oops in mdc_commit_close() after open replay
3082 Details : It was possible for the MDS to return an open request with no
3083 transaction number in mds_finish_transno() if the client was
3084 evicted, but without actually returning an error. Clients
3085 would later try to replay that open and may trip an assertion
3086 Simplify the client close codepath, and always return an error
3087 from the MDS in case the open is not successful.
3090 Frequency : rare, 2.6 OSTs only
3092 Description: OST may deadlock under high load on fragmented files
3093 Details : If there was a heavy load and highly-fragmented OST filesystems
3094 it was possible to have all the OST threads deadlock waiting on
3095 allocation of biovecs, because the biovecs were not released
3096 until the entire RPC IO was completed. Instead, release biovecs
3097 as soon as they are complete to ensure forward IO progress.
3099 Severity : enhancement
3101 Description: Support for specifying external journal device at mount
3102 Details : If an OST or MDS device is formatted with an external journal
3103 device, this device major/minor is stored in the ext3 superblock
3104 and may not be valid for failover. Allow detecting and
3105 specifying the external journal at mount time.
3110 Description: Mounting an MDS with pending unlinked files may cause oops
3111 Details : target_finish_recovery() calls mds_postrecov() which returned
3112 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3113 considers this an error and immediately begins cleaning up the
3114 lov, just after starting the mds_lov process
3116 Severity : enhancement
3118 Description: Implement 'lfs df' to report actual free space on per-OST basis
3119 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3120 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3121 usage of objects; '-h' to report in human readable format.
3123 ------------------------------------------------------------------------------
3125 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
3132 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3133 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3134 at startup during orphan unlinking. Instead, skip buddy update
3135 before setup, as it will be regenerated after recovery is complete.
3140 Description: If an OST is inactive, its locks might reference stale inodes.
3141 Details : lov_change_cbdata() must iterate over all namespaces, even if
3142 they are inactive to clear inode references from the lock.
3144 Severity : enhancement
3145 Frequency : occasional, if non-standard max_dirty_mb used
3147 Description: Client will block write RPCs if not enough grant
3148 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3149 then the client will block writes while waiting for another RPC
3150 to complete instead of consuming its dirty limit. With change
3151 we get improved performance when max_dirty_mb is small.
3153 Severity : enhancement
3154 Bugzilla : 3389, 6253
3155 Description: Add support for supplementary groups on the MDS.
3156 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3157 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3158 lookups for user supplementary groups into a cache.
3162 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3163 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3164 for liblustre O_EXCL is enforced by MDS.
3168 Description: readdir never returns NULL in liblustre.
3169 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3170 dentry in struct dirent.
3174 Frequency : liblustre only, depends on application IO pattern
3175 Description: liblustre clients evicted if not contacting servers
3176 Details : Don't put liblustre clients into the ping_evictor list, so
3177 they will not be evicted by the pinger ever.
3179 Severity : enhancement
3181 Description: Add ability to evict clients by NID from MDS.
3182 Details : By echoing "nid:$NID" string into
3183 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3184 $NID would be instantly evicted from this MDS and from all active
3185 OSTs connected to it.
3189 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3190 Details : lookup_it_finish() used to query file size from OSTs that was not
3195 Description: service threads change working directory to that of init
3196 Details : Starting lustre service threads may pin the working directory
3197 of the parent thread, making that filesystem busy. Threads
3198 now change to the working directory of init to avoid this.
3202 Frequency : during shutdown only
3203 Description: shutdown with a failed MDS or OST can cause unmount to hang
3204 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3209 Frequency : 2.6 only
3210 Description: chmod/chown may include an extra supplementary group
3211 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3212 supplementary group and if none is specified this is used.
3215 Bugzilla : 5479 (6816)
3217 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3218 Details : If lookup is in progress on a file that is unlinked we might try
3219 to revalidate the inode and fail in revalidate after lookup is
3220 complete and ll_file_open() enqueues the open again but
3221 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3224 Frequency : always, if lconf --abort_recovery used
3226 Description: lconf --abort_recovery fails with 'Operation not supported'
3227 Details : lconf was attempting to abort recovery on the MDT device and not
3230 ------------------------------------------------------------------------------
3232 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3237 Frequency : rare (only unsupported configurations with a node running as an
3239 Bugzilla : 6514, 5137
3240 Description: Mounting a Lustre file system on a node running as an OST could
3242 Details : OSTs now preallocates memory needed to write out data at
3243 startup, instead of when needed, to avoid having to
3244 allocate memory in possibly low memory situations.
3245 Specifically, if the file system is mounted on on OST,
3246 memory pressure could force it to try to write out data,
3247 which it needed to allocate memory to do. Due to the low
3248 memory, it would be unable to do so and the node would
3249 become unresponsive.
3251 Severity : enhancement
3253 Description: Addition of lconf --service command line option
3254 Details : lconf now accepts a '--service <arg>' option, which is
3255 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3257 Severity : enhancement
3259 Description: Failover mode is now the default for OSTs.
3260 Details : By default, OSTs will now run in failover mode. To return to
3261 the old behaviour, add '--failout' to the lmc line for OSTs.
3263 Severity : enhancement
3265 Description: Health checks are now provided for MDS and OSTs
3266 Details : Additional detailed health check information on MSD and OSTs
3267 is now provided through the procfs health_check value.
3270 Frequency : occasional, depends on IO load
3272 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3273 after numerous create/delete cycles
3274 Details : The ext3 inode allocation policy would not allocate new inodes
3275 very well on the OSTs because there are no new directories
3276 being created. Instead we look for groups with free space if
3277 the parent directories are nearly full.
3282 Description: Network or server problems during mount may cause partially
3283 mounted clients instead of returning an error.
3284 Details : The config llog parsing code may overwrite the error return
3285 code during mount error handling, returning success instead
3291 Description: MDS can fail to allocate large reply buffers
3292 Details : After long uptimes the MDS can fail to allocate large reply
3293 buffers (e.g. zconf client mount config records) due to memory
3294 fragmentation or consumption by the buffer cache. Preallocate
3295 some large reply buffers so that these replies can be sent even
3296 under memory pressure.
3300 Frequency : rare (liblustre)
3301 Description: fsx running with liblustre complained that using truncate() to
3302 extend the file doesn't work. This patch corrects that issue.
3303 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3304 ATTR_SIZE and lsm use in llu_setattr_raw.
3308 Frequency : rare, only 2.6 kernels
3309 Description: Unusual file access patterns on the MDS may result in inode
3310 data being lost in very rare circumstances.
3311 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3312 code in the 2.6 kernel caused the inode and/or EA data not to
3313 be read from disk, causing single-file corruption.
3317 Frequency : rare, only 2.6 filesystems using extents
3318 Description: Heavy concurrent write and delete load may cause data corruption.
3319 Details : It was possible under high-load situations to have an extent
3320 metadata block in the block device cache from a just-unlinked
3321 file overwrite a newly-allocated data block. We now unmap any
3322 metadata buffers that alias just-allocated data blocks.
3326 Frequency : filesystems with default stripe_count larger than 77
3327 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3328 Details : lconf specifies an inode size of 4096 bytes when the default
3329 stripe_count is larger than 77. This conflicts with the default
3330 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3331 this case to avoid pinning too much memory for large EAs.
3333 ------------------------------------------------------------------------------
3335 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3340 Frequency : rare (extremely heavy IO load with hundreds of clients)
3342 Description: Client is evicted, gets IO error writing to file
3343 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3344 caused clients to be evicted for AST timeouts. The fixes in
3345 bug 5192 mean we no longer need to have such short AST timeouts
3346 so ldlm_timeout has been increased.
3349 Frequency : occasional during --force or --failover shutdown under load
3350 Bugzilla : 5949, 4834
3351 Description: Server oops/LBUG if stopped with --force or --failover under load
3352 Details : a collection of import/export refcount and cleanup ordering
3353 issues fixed for safer force cleanup
3356 Frequency : only filesystems larger than 120 OSTs
3357 Bugzilla : 5990, 6223
3358 Description: lfs getstripe would oops on a very large filesystem
3359 Details : lov_getconfig used kfree on vmalloc'd memory
3362 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3363 Bugzilla : 6242, 6243
3364 Description: reading from files that had been truncated to a non-zero size
3365 but never opened returned no data
3366 Details : ll_file_read() reads zeros from no-object files to EOF
3371 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3372 Details : There's little harm in aborting MDS/OSS recovery and letting it
3373 try again, so I removed the LASSERT and return an error instead.
3375 Severity : enhancement
3377 Description: New debugging infrastructure for tracking down data corruption
3378 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3379 (b) cover more of the client-side code path, and (c) try to narrow
3380 down where problems occurred
3384 Bugzilla : 3819, 4364, 4397, 6313
3385 Description: Racing close and eviction MDS could cause assertion in mds_close
3386 Details : It was possible to get multiple mfd references during close and
3387 client eviction, leading to one thread referencing a freed mfd.
3389 Severity: : enhancement
3390 Bugzilla : 3262, 6359
3391 Description: Attempts to reconnect to servers are now more aggressive.
3392 Details : This builds on the enhanced upcall-less recovery that was added
3393 in 1.4.2. When trying to reconnect to servers, clients will
3394 now try each server in the failover group every 10 seconds. By
3395 default, clients would previously try one server every 25 seconds.
3400 Description: After recovery, certain operations trigger a failed
3401 assertion on a client.
3402 Details : Failing over an mds, using lconf -d --failover, while a
3403 client was doing a readdir() call would cause the client to
3404 LBUG after recovery completed and the readdir() was resent.
3406 Severity : enhancement
3408 Description: Default groups are now added by lconf
3409 Details : You can now run lconf --group <servicename> without having to
3410 manually add groups with lmc.
3413 Frequency : occasional
3415 Description: Nodes with an elan id of 0 trigger a failed assertion
3418 Frequency : always when accessing e.g. tty/console device nodes
3420 Description: tty and some other devices nodes cannot be used on lustre
3421 Details : file's private_data field is used by device data and lustre
3422 values in there got lost. New field was added to struct file to
3423 store fs-specific private data.
3426 Frequency : when exporting Lustre via NFS
3428 Description: NFSD failed occasionally when looking up a path component
3429 Details : NFSD is looking up ".." which was broken in ext3 directories
3430 that had grown large enough to become hashed.
3433 Frequency : Clusters with multiple interfaces not on the same subnet
3435 Description: Nodes will repeatedly try to reconnect to an interface which it
3436 cannot reach and report an error to the log.
3437 Details : Extra peer list entries will be created by lconf with some peers
3438 unreachable. lconf now validates the peer before adding it.
3441 Frequency : Only if a default stripe is set on the filesystem root.
3443 Description: Setting a default stripe on the filesystem root prevented the
3444 filesystem from being remounted.
3445 Details : The client was sending extra request flags in the root getattr
3446 request and did not allocate a reply buffer for the dir EA.
3449 Frequency : occasional, higher if lots of files are accessed by one client
3450 Bugzilla : 6159, 6097
3451 Description: Client trips assertion regarding lsm mismatch/magic
3452 Details : While revalidating inodes the VFS looks up inodes with ifind()
3453 and in rare cases can find an inode that is being freed.
3454 The ll_test_inode() code will free the lsm during ifind()
3455 when it finds an existing inode and then the VFS later attaches
3456 this free lsm to a new inode.
3460 Bugzilla : 6422, 7030
3461 Description: MDS deadlock between mkdir and client eviction
3462 Details : Creating a new file via mkdir or mknod (starting a transaction
3463 and getting the ns lock) can deadlock with client eviction
3464 (gets ns lock and trying to finish a synchronous transaction).
3467 Frequency : occasional
3468 Description: While starting a server, the fsfilt_ext3 module could not be
3470 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3471 kernels. Previously, lconf would still use the ext3 name
3472 when trying to load modules. Now, it will correctly use
3473 ext3 on 2.4 and ldiskfs on 2.6.
3475 Severity : enhancement
3476 Description: The default stripe count has been changed to 1
3477 Details : The interpretation of the default stripe count (0, to lfs
3478 or lmc) has been changed to mean striping across a single
3479 OST, rather than all available. For general usage we have
3480 found a stripe count of 1 or 2 works best.
3482 Severity : enhancement
3483 Description: Add support for compiling against Cray portals.
3484 Details : Conditional compiling for some areas that are different
3488 Frequency : occasional
3489 Bugzilla : 6409, 6834
3490 Description: Creating files with an explicit stripe count may lead to
3491 a failed assertion on the MDS
3492 Details : If some OSTs are full or unavailable, creating files may
3493 trigger a failed assertion on the MDS. Now, Lustre will
3494 try to use other servers or return an error to the
3498 Frequency : occasional
3500 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3501 caused lock timeout during readahead (since 1.4.2).
3502 Details : Processes doing readahead might match a lock that hasn't been
3503 granted yet if there are overlapping and conflicting lock
3504 requests. The readahead process waits on ungranted lock
3505 (original lock is CBPENDING), while OST waits for that process
3506 to cancel CBPENDING read lock and eventually evicts client.
3508 Severity : enhancement
3510 Description: Initial enabling of flock support for clients
3511 Details : Implements fcntl advisory locking and file status functions.
3512 This feature is provided as an optional mount flag (default
3513 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3514 locking are implemented yet, and those that are are not guaranteed
3515 to be completely correct in production environments.
3516 mount -t lustre -o [flock|noflock] ...
3519 Frequency : occasional
3521 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3522 assertion in the ext3 JBD (journaling) layer.
3523 Details : The b_committed_data struct is protected by the big kernel lock
3524 in 2.4 kernels, serializing journal_commit_transaction() and
3525 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3526 access to this struct. In 2.6 kernels there is finer grained
3527 locking to improve SMP performance of the JBD layer.
3531 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3536 Description: merge and cleanup kernel patches.
3539 -----------------------------------------------------------------------------
3541 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3543 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3544 versions of Lustre. Please update all servers and clients to
3545 version 1.4.2 or later at the same time. You must also run
3546 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3547 to update the configuration logs.
3549 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3550 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3551 - swab llog records of type '0' so we get proper header size/idx (5861)
3552 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3553 - fix rename of one directory over another leaking an inode (5953)
3554 - avoid SetPageDirty on 2.6 (5981)
3555 - don't re-add just-being-destroyed locks to the waiting list (5653)
3556 - when creating new directories, inherit the parent's custom
3557 striping settings if present parent (3048)
3558 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3559 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3560 - don't hold client locks on temporary worklist from l_lru (5666)
3561 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3562 - automatically evict dead clients (5921)
3563 - Update file size properly in create+truncate+fstat case (6196)
3564 - Do not unhash mountpoint dentries, do not allow removal of
3566 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3567 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3568 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3569 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3570 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3571 - Correctly handle reads of files with no objects (6243)
3572 - lctl recover will also mark a device active if deactivate used (5933)
3574 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3575 - allow --write-conf on an MDS with different nettype than client (5619)
3576 - don't write config llogs to MDS for mounts not from that MDS (5617)
3577 - lconf should create multiple TCP connections from a client (5201)
3578 - init scripts are now turned off by default; run chkconfig --on
3579 lustre and chkconfig --on lustrefs to use them
3580 - upcalls are no longer needed for clients to recover to failover
3582 - add --abort-recovery option to lconf to abort recovery on device
3584 - add support for an arbitrary number of OSTs (3026)
3585 - Quota support protocol changes.
3586 - forward compatibility changes to wire structs (6007)
3587 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3588 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3589 - improved "lustre status" script
3590 - initialize blocksize for non-regular files (6062)
3591 - added --disable-server and --disable-client configure options (5782)
3592 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3593 - Vanilla 2.4.29 support
3594 - increase maximum number of obd devices to 520 (6242)
3595 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3596 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3597 SLES 9 2.6.5 kernels
3598 - limit stripes per file to 160 (the maximum EA size) (6093)
3600 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3603 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3604 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3605 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3606 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3607 - small ext3 extents cleanups and fixes (5733)
3608 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3609 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3610 - avoid needless client->OST connect, fix handle mismatch (5317)
3611 - fix DLM error path that led to out-of-sync client, long delays (5779)
3612 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3613 - fix several locking issues related to i_size (5492,5624,5654,5672)
3614 - don't move pending lock onto export if it is already evicted (5683)
3615 - fix kernel oops when creating .foo in unlinked directory (5548)
3616 - fix deadlock in obdfilter statistics vs. object create (5811)
3617 - use time_{before,after} to avoid timer jiffies wrap (5882)
3618 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3619 - Do not leak request if server was not able to process it (5154)
3620 - If mds_open unable to find parent dir, make that negative lookup(5154)
3621 - don't create new directories with extent-mapping (5909, 5936)
3623 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3624 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3625 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3626 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3627 - support for RHEL4 kernel on i686 (5773)
3628 - provide error messages when incompatible logs are encountered (5898)
3630 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3631 * version 1.4.0.10 (1.4.1 release candidate 1)
3633 - don't keep a lock reference when lock is not granted (4238)
3634 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3635 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3636 - drop import inflight refcount on signal_completed_replay error (5255)
3637 - unlock page after async write error during send (3677)
3638 - handle missing objects in filter_preprw_read properly (5265)
3639 - no transno return for symlink open, don't save no-trasno open (3440)
3640 - don't try to complete elan receive that already failed (4012)
3641 - free RPC server reply state on error (5406)
3642 - clean up thread from ptlrpc_start_thread() on error (5160)
3643 - readahead could read extra page into cache that wasn't ejected (5388)
3644 - prevent races in class_attach/setup/cleanup/detach (5260)
3645 - don't dereference de->d_inode after l_dput of de (5458)
3646 - use "int" for stripe value returned from lock_to_stripe (5544)
3647 - mballoc allocation and error-checking fixes in 2.6 (5504)
3648 - block device patches to fix I/O request sizes in 2.6 (5482)
3649 - look up hostnames for IB nals (5602)
3650 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3651 - don't start multiple acceptors for the same port (5277)
3652 - fix incorrect LASSERT in mds_getattr_name (5635)
3653 - export a proc file for general "ping" checking (5628)
3654 - fix "lfs check" to not block when the MDS is down (5628)
3656 - service request history (4965)
3657 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3658 - create an "evict_client" /proc entry on OSTs, like the MDS has
3659 - fix mount usage message, return errors per mount(8) (5168)
3660 - change grep [] to grep "[]" in tests so they work in more UMLs
3661 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3662 - remove ancient LOV_MAGIC_V0 EA support (5047)
3663 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3664 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3665 - allow readahead allocations to fail when low on memory (5383)
3666 - mmap locking landed again, after considerable improvement (2828)
3667 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3669 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3672 - send OST transaction number in read/write reply to free req (4966)
3673 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3674 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3675 - the watchdog thread now runs as interruptible (5246)
3676 - flock/lockf fixes (but it's still disabled, pending 5135)
3677 - don't use EXT3 constants in llite code (5094)
3678 - memory shortage at startup could cause assertion (5176)
3680 - reorganization of lov code
3681 - single portals codebase
3683 - add extents/mballoc support (5025)
3684 - direct I/O reads in the obdfilter (4048)
3685 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3687 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3690 - send OST transaction number in read/write reply to free req (4966)
3691 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3692 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3693 - the watchdog thread now runs as interruptible (5246)
3694 - handle missing objects in filter_preprw_read properly (5265)
3695 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3696 - drop import inflight refcount on signal_completed_replay error (5255)
3697 - unlock page after async write error during send (3677)
3698 - return original error code on reconstructed replies (3761)
3699 - no transno return for symlink open, don't save no-trasno open (3440)
3701 - add pid to ldlm debugging output (4922)
3702 - bump the watchdog timeouts -- we can't handle 30sec yet
3703 - extra debugging for orphan dentry/inode bug (5259)
3705 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3708 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3709 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3710 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3711 - fix NULL dereference in /proc/sys/portals/routes (4827)
3712 - allow failed mdc_close() operations to be interrupted (4561)
3713 - stop precreate on OST before MDS would time out on it (4778)
3714 - don't send partial-page writes before EOF from client (4410)
3715 - discard client grant for sub-page writes on large-page clients (4520)
3716 - don't free dentries not owned by NFS code, check generation (4806)
3717 - fix lsm leak if mds_create_objects() fails (4801)
3718 - limit debug_daemon file size, always print CERROR messages (4789)
3719 - use transno after validating reply (3892)
3720 - process timed out requests if import state changes (3754)
3721 - update mtime on OST during writes, return in glimpse (4829)
3722 - add mkfsoptions to LDAP (4679)
3723 - use ->max_readahead method instead of zapping global ra (5039)
3724 - don't interrupt __l_wait_event() during strace
3726 - add software watchdogs to catch hung threads quickly (4941)
3727 - make lustrefs init script start after nfs is mounted
3728 - fix CWARN/ERROR duplication (4930)
3729 - return async write errors to application if possible (2248)
3730 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3731 - print NAL number in %x format (4645)
3732 - update barely-supported suse-2.4.21-171 series (4842)
3733 - support for sles 9 %post scripts
3734 - support for building 2.6 kernel-source packages
3735 - support for sles km_* packages
3737 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3740 - ignore -ENOENT errors in osc_destroy (3639)
3741 - notify osc create thread that OSC is being cleaned up (4600)
3742 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3743 - reconstruct ost_handle() like mds_handle() (4657)
3744 - create a new thread to do import eviction to avoid deadlock (3969)
3745 - let lconf resolve symlinked-to devices (4629)
3746 - don't unlink "objects" from directory with default EA (4554)
3747 - hold socknal file ref over connect in case target is down (4394)
3748 - allow more than 32000 subdirectories in a single directory (3244)
3749 - fix blocks count for O_DIRECT writes (3751)
3750 - OST returns ENOSPC from object create when no space left (4539)
3751 - don't send truncate RPC if file size isn't changing (4410)
3752 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3753 - bind to privileged port in socknal and tcpnal (3689)
3755 - rate limit CERROR/CWARN console message to avoid overload (4519)
3756 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3757 - basic mmap support (3918)
3758 - kernel patch series update from b1_4 (4711)
3760 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3763 - avoid crash during MDS cleanup with OST shut down (2775)
3764 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3765 - don't use bad inodes on the MDS (3744)
3766 - dynamic object preallocation to improve recovery speed (4236)
3767 - don't hold spinlock over lock dumping or change debug flags (4401)
3768 - don't zero obd_dev when it is force cleaned (3651)
3769 - print grants to console if they go negative (4431)
3770 - "lctl deactivate" will stop automatic recovery attempts (3406)
3771 - look for existing locks in ldlm_handle_enqueue() (3764)
3772 - don't resolve lock handle twice in recovery avoiding race (4401)
3773 - revalidate should check working dir is a directory (4134)
3775 - don't always mark "slow" obdfilter messages as errors (4418)
3777 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3780 - don't close LustreDB during write_conf until it is done (3860)
3781 - fix typo in lconf for_each_profile (3821)
3782 - allow dumping logs from multiple threads at one time (3820)
3783 - don't allow multiple threads in OSC recovery (3812)
3784 - fix debug_size parameters (3864)
3785 - fix mds_postrecov to initialize import for llog ctxt (3121)
3786 - replace config semaphore with spinlock (3306)
3787 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3788 - don't allow enqueue to complete on a destroyed export (3822)
3789 - down write_lock before checking llog header bitmap (3825)
3790 - recover from lock replay timeout (3764)
3791 - up llog sem before sending rpc (3652)
3792 - reduce ns lock hold times when setting kms (3267)
3793 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3794 - fix NULL deref and obd_dev leak on setup error (3312)
3795 - replace some LBUG about llog ops with error handling (3841)
3796 - don't match INVALID dentries from d_lookup and spin (3784)
3797 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3798 - fix invalid assertion in ptlrpc_set_wait (3880)
3800 - add libwrap support for the TCP acceptor (3996)
3801 - add /proc/sys/portals/routes for non-root route listing (3994)
3802 - allow setting MDS UUID in .xml (2580)
3803 - print the stack of a process that LBUGs (4228)
3805 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3808 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3809 - make sure to unset replay flag from failed open requests (3440)
3810 - if default stripe count is 0, use OST count for inode size (3636)
3811 - update parent mtime/ctime on client for create/unlink (2611)
3812 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3813 - free recovery state on server during a forced cleanup (3571)
3814 - unregister_reply for resent reqs (3063)
3815 - loop back devices mounting and status check on 2.6 (3563)
3816 - fix resource-creation race that can provoke i_size == 0 (3513)
3817 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3818 - more debugging for page-accounting assertion (3746)
3819 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3820 - avoid deadlock after precreation failure (3758)
3821 - fix race and lock order deadlock in orphan handling (3450, 3750)
3822 - add validity checks when grabbing inodes from l_ast_data (3599)
3824 - add /proc/.../recovery_status to obdfilter (3428)
3825 - lightweight CDEBUG infrastructure, debug daemon (3668)
3826 - change default OSC RPC parameters to be better on small clusters
3827 - turn off OST read cache for files smaller than 32MB
3828 - install man pages and include them in rpms (3100)
3829 - add new init script for (un)mounting lustre filesystems (2593)
3830 - run chkconfig in %post for init scripts (3701)
3831 - drop scimac NAL (unmaintained)
3833 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3836 - clean kiobufs before and after use (3485)
3837 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3838 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3839 - change init script to fail more clearly if not run as root (1528)
3840 - allow clients to reconnect during replay (1742)
3841 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3842 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3843 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3844 - fix cond_resched (3554)
3845 - only evict unfinished clients after recovery (3515)
3846 - allow bulk resend, prevent data loss (3570)
3847 - dynamic ptlrpc request buffer allocation (2102)
3848 - don't allow unlinking open directory if it isn't empty (2904)
3849 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3850 - remove extraneous obd dereference causing LASSERT failure (3334)
3851 - don't use get_cycles() when creating temp. files on the mds (3156)
3852 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3853 - handle EEXIST for set-stripe, set proper directory name (3336)
3855 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3856 - fix ksocknal_fmb_callback() error messages (2918)
3858 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3861 - don't copy lvb into (possibly NULL) reply on error (2983)
3862 - don't deref dentry after dput, don't free lvb on error (2922)
3863 - use the kms to determine writeback rpc length (2947)
3864 - increment oti_logcookies when osc is inactive (2948)
3865 - update client's i_blocks count via lvb messages (2543)
3866 - handle intent open/close of special files properly (1557)
3867 - mount MDS with errors=remount-ro, like obdfilter (2009)
3868 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3869 - don't use cancelling-locks' kms values (2947)
3870 - use highest lock extent for kms, not last one (2925)
3871 - don't dereference ERR_PTR() dentry in error handling path (3107)
3872 - fix thread race in portals_debug_dumplog() (3122)
3873 - create lprocfs device entries at setup instead of at attach (1519)
3874 - common AST error handler, don't evict client on completion race (3145)
3875 - zero nameidata in detach_mnt in 2.6 (3118)
3876 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3877 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3878 - initialize RPC timeout timer earlier for 2.6 (3219)
3879 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3880 - print nal/nid for unknown nid (3258)
3881 - additional checks for oscc recovery before doing precreate (3284)
3882 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3883 - don't crash in mdc_close for bad permissions on open (3285)
3884 - zero i_rdev for non-device files (3147)
3885 - clear page->private before handing to FS, better assertion (3119)
3886 - tune the read pipeline (3236)
3887 - fix incorrect decref of invalidated dentry (2350)
3888 - provide read-ahead stats and refine rpc in flight stats (3328)
3889 - don't hold journal transaction open across create RPC (3313)
3890 - update atime on MDS at close time (3265)
3891 - close LDAP connection when recovering to avoid server load (3315)
3892 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3893 - don't leak open file on MDS after open resend (3325)
3894 - serialize filter_precreate and filter_destroy_precreated (3329)
3895 - loop device shouldn't call sync_dev() for nul device (3092)
3896 - clear page cache after eviction (2766)
3897 - resynchronize MDS->OST in background (2824)
3898 - refuse to mount the same filesystem twice on same mountpoint (3394)
3899 - allow llmount to create routes for mounting behind routers (3320)
3900 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3901 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3902 - fix rare problem with rename on htree directories (3417)
3904 - allow default OST striping configuration per directory (1414)
3905 - fix compilation for qswnal for 2.6 kernels (3125)
3906 - increase maximum number of MDS request buffers for large systems
3907 - change liblustreapi to be useful for external progs like lfsck (3098)
3908 - increase local configuration timeout for slow disks (3353)
3909 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3911 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3914 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3915 - don't overwrite extent policy data in reply if lock was blocked (2901)
3916 - drop filter export grants atomically with removal from device (2663)
3917 - del obd_self_export from work_list in class_disconnect_exports (2908)
3918 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3919 - swab reply message in mdc_close, other PPC fixes (2464)
3920 - fix destroying of named logs (2325)
3921 - overwrite old logs when running lconf --write_conf (2264)
3922 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3923 - fix race in target_handle_connect (2898)
3924 - mds_reint_create() should take same inode create lock (2926)
3925 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3926 - don't close files for self_export to avoid uninitialized obd (2936)
3927 - allow MDS with the same name as client node (2939)
3928 - hold dentry reference for closed log files for unlink (2325)
3929 - reserve space for all logs during transactions (2059)
3930 - don't evict page beyond end of stripe extent (2925)
3931 - don't oops on a deleted current working directory (2399)
3932 - handle hard links to targets without a parent properly (2517)
3933 - don't dereference NULL lock when racing during eviction (2867)
3934 - don't grow lock extents when lots of conflicting locks (2919)
3936 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3939 - account for cache space usage on clients to avoid data loss (974)
3940 - lfsck support in lustre kernel code (2349)
3941 - reduce journal credits needed for BRW writes (2370)
3942 - orphan handling to avoid losing space on client/server crashes
3943 - ptlrpcd can be blocked, stopping ALL progress (2477)
3944 - use lock value blocks to assist in proper KMS, faster stat (1021)
3945 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3946 - recovery for initial connections (2355)
3947 - fixes for mds_cleanup_orphans (1934)
3948 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3949 - block all file creations until orphan recovery completes (1901)
3950 - client remove rq_connection from request struct (2423)
3951 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3952 - recovery timer race (2670)
3953 - mdc_close recovey bug (2532)
3954 - ptlrpc cleanup bug (2710)
3955 - mds timeout on local locks (2588)
3956 - namespace lock held during RPCs (2431)
3957 - handle interrupted sync write properly (2503)
3958 - don't try to handle a message that hasn't been replied to (2699)
3959 - client assert failure during cleanup after abort recovery (2701)
3960 - leak mdc device after failed mount (2712)
3961 - ptlrpc_check_set allows timedout requests to complete (2714)
3962 - wait for inflight reqs when ptlrpcd finishes (2710)
3963 - make sure unregistered services are removed from the srv_list
3964 - reset bulk XID's when resending them (caught by 1138 test)
3965 - unregister_bulk after timeout
3966 - fix lconf error (2694)
3967 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3968 - readahead locks pages, leaves pending causing memory pressure (2673)
3969 - increase OST request buffers to 4096 on large machines (2729)
3970 - fix up permission of existing directories in simple_mkdir (2661)
3971 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3972 - don't assign transno to errored transactions (2742)
3973 - don't delete objects on OST if given a bogus objid from MDS (2751)
3974 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3975 - if rq_no_resend, then timeout request after recovery (2432)
3976 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3977 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3978 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3979 - MDS recovery shouldn't delete objects that it has given out (2730)
3980 - if enqueue arrives after completion, don't clobber LVB (2819)
3981 - don't unlock pages twice when trigger_group_io returns error (2814)
3982 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3983 - don't write pages to disk if there was an error (1450)
3984 - don't ping imports that have recovery disabled (2676)
3985 - take buffered bytes into account when balancing socknal conn (2817)
3986 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3987 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3988 - remove little-endian swabbing of llog records (1987)
3989 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3990 - retry reposting request buffers if they fail (1191)
3991 - grow extent at grant time to avoid granting a revoked lock (2809)
3992 - lock revoke doesn't evict page if covered by a second lock (2765)
3993 - disable VM readahead to avoid reading outside lock extents (2805)
3995 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3996 - updated kernel patches for hp-2.4.20 kernel (2681)
3998 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4001 - fix truncated write corruption (2366)
4002 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4004 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4005 - ensure len(uuid) < 37 in lmc (1171)
4006 - fix ia64 OOPS in llog_test (2255)
4007 - zero end of page at obdfilter for partial page writes (2648)
4008 - don't leave stale dentries around after renames (bug 2428)
4009 - fix timeouts when evicting a client with a single lock held (2642)
4010 - set deadline for the initial HELLO message to drain (2634)
4011 - print out dotted-quad IP addresses in the socknal (2302)
4013 - additional debugging for MDS client eviction problem (2443)
4014 - fix mkfsoptions support for osts (2603, 2604)
4016 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4019 - add series for the vanilla 2.6.0 kernel
4020 - add series for the vanilla 2.4.24 kernel
4021 - add series for a cray x86/64 UL kernel drop
4022 - fix xattr patches for the vanilla 2.4.19 series
4024 - generate true UUIDs in lmc (1171)
4025 - have portals stack dumping break in UML (2466)
4026 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4027 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4028 - wake callback waiting threads on client eviction (2460)
4029 - Add --ptldebug and --subsystem to lmc (1719)
4030 - update assertion to allow safe interrupt allocation
4031 - set rq_no_resend for cancel requests (2432)
4032 - recalculate ptlrpcd timeout after resend (2494)
4033 - call vfs_rmdir when removing pending directories (2368)
4034 - fix renaming a file to itself (2429)
4035 - lmc creates a default one-stripe lov (2454)
4036 - expand procfs space to handle large clusters (2326)
4037 - increase UML stack to avoid overflow
4038 - update lconf's list of debug and subsystem masks
4039 - fix lfs find --obd (2510)
4040 - /proc tunable for disabling filter read caching (2591)
4041 - stop rpm packages from altering slapd.conf (2301)
4042 - disable nagle in the socknal under 0conf (2578)
4043 - choose mds inode size based on stripe count (2572)
4044 - fix kernel-source rpm problems (2516)
4046 - add --disable-doc to avoid pdf generation (2421)
4047 - update documentation, tests, type-os, comments
4048 - avoid format warnings on ia64
4049 - remove the TOE NAL
4050 - tiny code cleanups by removing unused fields
4052 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4055 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4056 - remove the most blatant lies from BUILDING file (2371)
4057 - change default debug level to reasonable production setting
4058 - reduce client side cache size to reduce cache flush time
4059 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4060 - make TCP zerocopy and pinger support enabled by default (2476)
4061 - sync writes completed after process exits caused crashes (2319)
4062 - maintain correct mount count on the MDS (2356)
4063 - backout 1557, because 2316 wasn't really fixed
4064 - better file I/O statistics gathering in /proc
4065 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4066 - another kernel patch to fix zero-copy TCP function export
4067 - don't take duplicate lock when processing re-sent getattr (2420)
4068 - lctl uses obd_self_export instead of creating new conn (2353)
4069 - MDS/OST recovery case which requires object creation asserted (2425)
4070 - move lfs from /usr/sbin to /usr/bin in packages
4071 - fix race between mds_client_add and mds_client_free (2417)
4072 - use kmalloc instead of slabs in portals (2430)
4073 - don't create duplicate records when a failover MDS is present (2442)
4074 - remove unnecessary mount age check (2332)
4075 - don't remove directory inodes from locks prematurely (2451)
4076 - don't break if MDS service name is the same as hostname (2103)
4077 - fix races in client write RPC generation when cache full (2482)
4079 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4082 - remove now-unused request->rq_obd (278)
4083 - if an allocation fails, print out how much memory we've used (1933)
4084 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4085 - add the "configurable stack size" patch to most series files (1256)
4086 - ability to write large log records, for 100+ OST configs (2306)
4087 - fix NULL deref when filter_prep fails (2314)
4088 - fix operator precedence error in filter_sync
4089 - dynamic allocation of socknal TX descriptors (2315)
4090 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4091 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4092 - ensure that CWARN messages in Portals always get to the syslog
4093 - __init/__exit are not for prototype decls (ldlm_init/exit)
4094 - x86-64 compile warning fixes
4095 - fix gateway LMC keyword conflict (2318)
4096 - fix MDS lock inversions in getattr/reint paths (1844)
4097 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4098 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4099 - DLM race condition prevented some lock evictions (2328)
4100 - ENOMEM detection and retry on socknal sends (2230)
4101 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4102 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4103 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4104 - better error messages when a client is rejected during recovery (1505)
4105 - avoid cancelling locks which were never granted, after failure (2330)
4106 - fix i_sem/journal inversion in mds_client_add (2333)
4107 - fix truncate/getattr lock cycle deadlock (2334)
4108 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4109 - fix two rare exit paths which could leak an l_lock() ref (2321)
4110 - fencepost error in MDS/OST orphan recovery (2226)
4111 - make log record alignment 8 bytes (1988)
4112 - lstripe now fails when requested offset > ost_count (2237)
4113 - ensure that all kernel series have a complete list.h (1607)
4114 - fix crashes in special-file operations (2316)
4115 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4117 - allow configurable automake binary, for testing new versions
4118 - small update to the lfs documentation
4120 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4122 * fix negative export reference count in fsfilt_sync (2312)
4124 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4125 * release candidate 0.9.1
4127 - orphans are moved into the PENDING directory for possible recovery
4128 - replayed opens now open by fid for orphan/rename safety (1042)
4129 - last close of an orphan inode generates a transno (683)
4130 - chdir() and mount() now pin the directory entry (1020)
4131 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4132 - discard very old requests without processing them (1502)
4133 - remove some common, well-understood CERRORs (1505)
4134 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4135 - clear "grant" flags in OST replies until OST grant code lands (1644)
4136 - fix read performance by not clobbering i_blksize on client (1598)
4137 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4138 - make LRU size a /proc tunable, clears locks when reduced (707)
4139 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4140 - server locks take references on exports now (1558)
4141 - build fixes for 2.4.20-rh trees (1663)
4142 - return an error from lov_create if all OSCs are inactive (1751)
4143 - fix import levels when a reconnect happens without a timeout (1597)
4144 - exit early from mds_open if we get a lookup error (1749)
4145 - partial page read at EOF wouldn't wait for disk before sending (1642)
4146 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4147 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4148 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4149 - write/truncate lock inversion (1639)
4150 - Don't auto-load obdclass, portals modules during cleanup (1495)
4151 - fix timestamps from jumping to "now" (1763)
4152 - extra journal assertions (1648)
4153 - add an extra multiunlink test (1771)
4154 - fix read_record/write_record API (1776)
4155 - fix leak of offset_extent, possible incorrect i_size later (1772)
4156 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4157 - Debugging for the kqswnal_get_idle_tx problems (1820)
4158 - Allow recovery to be attempted multiple times (1536)
4159 - Write out MDS last_rcvd file after it is first created (1600)
4160 - Fix tx_descriptor leak in failed transmit situations (1827)
4161 - ext3 journaling fixes for assertion failure after IO error (1871)
4162 - class_export_put() on freed export after completion AST error (1896)
4163 - Fix revalidate looping in VFS (1322)
4164 - Don't access a freed export during MDS_REINT timeout (1521)
4165 - Add open-unlink recovery support on the MDS (1673,1764)
4166 - Return an error if no MDS data was read from last_rcvd (1946)
4167 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4168 - Don't setup a disk device that doesn't match exported UUID (317)
4169 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4170 - avoid committing NULL handle in force close
4171 - local.sh is now a one-stripe LOV configuration
4172 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4173 - fix ext3 htree duplicate directory entry corruption (1516)
4174 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4175 - update to new LOV EA format (2097)
4176 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4178 2003-06-15 Phil Schwan <phil@clusterfs.com>
4181 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4182 - per-import/export recovery handling (958, 931, 959)
4183 - multiple last-rcvd slots, for serving multiple FSes (949)
4184 - connections are again shared between multiple imp/exports (963, 964)
4185 - "umount -f" would hang if any requests needed to be sent (393, 978)
4186 - avoid pinning large req buffer by copying for queued messages (989)
4187 - add "uuid" to "lctl device" command to help upcalls (991)
4188 - "open" RPCs with transnos would confuse recovery counters (1037)
4189 - do proper endian conversion of all wire messages (288, 340, 891)
4190 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4191 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4192 - fix ptlrpc_connection leak in target_handle_connect (1174)
4193 - fix import refcounting bug in OST and MDS cleanup (1134)
4194 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4195 - fix very unlikely obd_types race condition (501)
4196 - remove osc_open hack for echo_client (1187)
4197 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4198 - a failure in read_inode2 leads to deadlock (1139)
4199 - cancel ack-locks as soon as transaction is committed (1072)
4200 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4201 - make sure to commitrw after any preprw to avoid deadlock (1162)
4202 - failing to execute a file in a lustre FS would lock inode (1203)
4203 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4204 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4205 - obd_uuid2tgt would walk off the end of the list (1255)
4206 - on IA64 the osc would give portals incorrect bulk size (1258)
4207 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4208 - fix lock inversion caused by new llite matching code (1282)
4209 - limit the number of dirty pages on a client to 10MB (1286)
4210 - timed out locks were not being corrected cancelled (1289)
4211 - fix O_DIRECT above 4GB on IA-32 (1292)
4212 * major user-visible changes
4213 - fail out/fail over policy now controlled by the upcall (993)
4215 - add OBD_PING to check server availability and failure (954)
4216 - lustre messages are now sent in sending host order (288, 340, 891)
4217 - add eadatalen to MDS getattr reply (340)
4218 - OST read replies may contain second buffer, with per-page status (593)
4220 2003-03-11 Phil Schwan <phil@clusterfs.com>
4223 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4224 - failed open()s could cause deadlock; fixed (867, 869)
4225 - stop cancelling OST locks when files are closed (481)
4226 - overlapping XID spaces caused network corruption (851, 853)
4227 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4228 - setattr_raw added, to do single-RPC, server-side setattrs
4229 - lmc/lconf syntax change for OST UUIDs
4230 - fix crashy race condition between ptlrpc_free_req and osc_close
4231 - don't use request in mdc_enqueue if we hit a timeout (889)
4232 - don't set the inode i_size for regular files from the MDS (896)
4233 - handle out of order completion AST (842)
4234 - don't LBUG if a lock request times out after receiving AST (913)
4235 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4236 - if a bad lock AST arrives, send an error instead of dropping entirely
4237 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4238 - fix leak in bulk IO when only partially completed (899, 900, 926)
4239 - fix O_DIRECT for ia64 (55)
4240 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4241 - C-z after timeout could hang a process forever; fixed (977)
4243 - client-side I/O cache (678, 924, 929, 941, 970)
4245 - READPAGE and SETATTRs which don't take server-side locks get
4248 2003-02-11 Phil Schwan <phil@clusterfs.com>
4251 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4252 - Fix fsfilt_extN_readpage() to read a full page of directory
4253 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4254 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4255 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4256 - Most intent operations were removed, in favour of a new RPC mode
4257 that does a single RPC to the server and bypasses most of the VFS
4258 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4259 - Aggressively cancel local locks on DLM servers
4260 - mds_reint_unlink sends EA to the client if it's the last nlink.
4261 client uses that EA to unlink OST objects.
4262 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4263 - recursive symlinks were fixed (439)
4264 - fixed NULL deref in DEBUG_REQ
4265 - filter_update_lastobjid no longer calls sync, which annoyed extN
4266 - fixed multi-client small-writes to a single file problem (445)
4267 - fixed mtime updates during file writes (607)
4268 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4269 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4270 - fixed Linux OST size reporting problem (444, 656)
4271 - OST now updates object mtime with writes or setattr (607, 619)
4272 - client verifies file size before zeroing page past EOF (445)
4273 - OST now writes last allocated objid to disk with allocation (108)
4274 - LOV on echo now works (409)
4276 - mds_reint_unlink sends a new buffer, with the EA included. this
4277 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4278 set if a regular file was being unlinked, and it was the last link
4279 - use PtlGet from the target for bulk writes (315)
4280 - OST now updates object mtime with writes or setattr (607, 619)
4281 - LDLM now has a grant-time callback to revalidate locked items, if
4283 - Many MDS operations were reorganized to combat race conditions
4285 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4286 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4288 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4291 - Fully reactivate OST imports after reconnection (512, others)
4292 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4293 - More graceful error handling for truncating on dead OST (515)
4294 - Don't error out unless we're actually accessing dead stripes (474)
4295 - Fix garbage sizes when stripes are missing (410)
4296 - LRU counters were broken, causing constant lock purge (433, 432)
4297 - garbage on read from stripes with failed OSTs (441)
4298 - mark OSCs as active before reconnecting during recovery (438)
4299 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4300 - lfind did not preserve OST order in output (443)
4301 - symlinks cause hung clients, incorrect data (439)
4302 - stop dereferencing request after dropping refcount (457)
4303 - don't LASSERT(spin_is_locked) on non-SMP (455)
4304 - fixes for many rename() bugs
4305 - fstat didn't correctly synchronize attributes (399)
4306 - server must handle lock cancellation during blocking AST prep (487)
4307 - bulk descriptors were free()d too soon (511)
4308 - fix paths in lconf, which would load incorrect modules (451, 507)
4309 - fix confusing lconf 'host not found' error message (386)
4310 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4311 - fix race condition in mdc_blocking_ast() for inode access (526)
4312 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4313 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4314 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4315 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4316 - fix lstripe in several ways (kernel side) (527)
4317 - fix request leak in ldlm_cli_enqueue (262)
4318 - incorrect OSC was marked inactive after OST failure
4319 - call mds_fs_cleanup before unmounting filesystem (524)
4320 - fix races between taking ns_lock and ldlm_lock_change_resource
4321 - fix races updating LOV export open file list
4322 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4323 - fix recovery NULL deref in ldlm_cli_cancel_unused
4324 - fix some DLM races by using new hash table for lock handles (419)
4325 - permit the client to specify desired inodes, at replay
4326 - duplicate requests when we queue them for replay reintegration
4327 - fix last_rcvd offset calculation
4328 - sync after each recovered transaction, so we always make progress
4329 - never, not always, ERESTART requests without transnos
4330 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4332 - skip replay if the MDS says that the client is already connected
4333 - don't check for a recovery-enabled export to match lctl's UUID
4334 - don't INC_USE_COUNT for phantom exports
4335 - don't crash when cleaning up phantom exports (567)
4336 - don't double-finish or set replay data for errored mdc_open requests
4337 - abort requests when they time out, so we don't get old replies
4338 - send/receive replies for AST messages again
4339 - if the client says that it doesn't have the lock, cancel it on the
4341 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4342 mark it as destroyed, it will all work out when decref is called
4343 - fix module use counts (22, 581)
4345 - ASTs now expect a reply (server cancels lock on error reply)
4347 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4350 - fix many simultaneous client startup (392)
4351 - fix dentry->d_it clobbering
4352 - credentials weren't being shipped for readdir/getattr operations
4353 - remove invalid assertions triggered during some concurrent MD
4355 - proper Lustre versions added (336, 389)
4356 - fix memory leak for create error case (398)
4357 - fix LOV locking bug that would get cli/srv out of sync
4358 - fix echo client over LOV (409)
4359 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4360 - fix double-O_EXCL intent crash (424)
4361 - avoid sending multiple lock CANCELs (352)
4363 - MDS can do multi-client recovery (modulo bugs in new code)
4365 - many updates, edits, cleanups
4367 2002-11-18 Phil Schwan <phil@clusterfs.com>
4370 - fix null d_it dereference (346)
4371 - fix full OST/dbench hang (333)
4372 - fix permission problem with file removal (286)
4373 - fix removal of OSCs from LOV when they fail
4374 - fix NULL deref during bulk timeout (214)
4375 - fix problems related to multiple filesystems on one MDS (241)
4376 - fixed serious subtle metadata locking bugs
4377 - free locks on clients when inodes are removed due to memory
4379 - fix inode pointer in lock data (285)
4380 - partial support for multiple MDS on a single host (241)
4381 - data locks weren't cancelled at clear_inode time (290, 311)
4382 - intent locks could lead to unbounded lock growth (205)
4383 - added a maximum lock count, an LRU list, and a flusher
4384 - fix multiple rename (365)
4385 - properly abstracted the echo client
4386 - OSC locked 1 byte too many; fixed
4387 - rewrote brw callback code:
4388 - fixed recovery bugs related to LOVs (306)
4389 - fixed too-many-pages-in-one-write crash (191)
4390 - fixed (again) crash in sync_io_timeout (214)
4391 - probably fixed callback-related race (385)
4393 - Add capability to MDS protocol
4394 - LDLM cancellations and callbacks on different portals
4396 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4399 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4400 - defer open object destruction to close time (601981, 138)
4401 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4402 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4403 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4404 - don't call obd_setattr in ll_file_release for destroyed objects
4405 * protocol change to lustre_msg: move |version| and add |flags|
4406 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4407 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4408 * added replay of create, unlink, link and rename operations during
4409 MDS failover; recovery should be much more robust now
4410 * remove failed OSCs from LOVs (only lov_create uses this so far)
4411 * the lustre-HOWTO was brought (more) up to date (582544)
4413 2002-10-23 Phil Schwan <phil@clusterfs.com>
4416 - in-use dentries weren't being reused properly (617851)
4417 - prevent multiple LDLM setup (599178)
4418 - fix LOV size calculations for truncate (617853)
4419 - fix client handling of MDS intent errors (POSIX)
4420 - fix permission bug in lovstripe.c test (624321)
4421 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4422 - truncate past end of file could corrupt data
4423 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4424 - a race in recovery could return ETIMEDOUT to apps (623947)
4425 - building outside the source directory was fixed
4426 * the lustre-HOWTO was brought (more) up to date (582544)
4427 * major progress was made on recovery functionality
4429 2002-10-10 Phil Schwan <phil@clusterfs.com>
4432 - recovery deadlock fix
4433 - rm -rf causes LBUG fix (617817)
4434 - file open by multiple tasks fix (618962)
4435 - directory permissions bugs (602707 and 620007)
4436 - journal_stop fixed with locking (611313)
4437 - O_APPEND failures resolved (618273, perhaps 614459)
4438 - lconf PATH fix (619770)
4439 - IA64 build fix (621450)
4440 - RPC buffer sizes scale with amount of memory
4442 2002-10-01 Phil Schwan <phil@clusterfs.com>
4445 - locks would be cancelled without throwing away data pages,
4446 resulting in inconsistent data (605627)
4447 - inode attributes were not always being refreshed (605627, 612449)
4448 - lconf now continues to cleanup after lctl reports an error
4449 - MDS now enforces user permissions (602707)
4450 - lprocfs cleanup fixed, but not yet enabled (614157)
4451 - fixed infinite server hang, should a client not respond to an AST
4452 - avoid going into recovery if user calls readlink() with a buffer
4453 that's too small (613941)
4454 - AST RPCs no longer require replies (614867) -- this may be changed
4455 - don't crash server if client sends an IOV that's too big (611336)
4456 - fixed lock conversion deadlock (611892)
4457 - fixed the following of symlinks (614622)
4458 * recovery: the server can remove locks from a client that dies, other
4459 clients can make progress
4460 * more extN patch fixes
4461 * compile-time configurable ptlrpc buffer allocations
4463 - collaborative read cache document
4464 - Lustre Lite Performance CDR document-in-progress
4466 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4469 - fix typo in patch-2.4.18
4471 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4474 - clear ptlrpc request each time in handle_incoming_request()
4475 - unlink of files now destroys the object on the OST
4477 2002-09-19 Peter Braam <braam@clusterfs.com>
4479 * add hard link support
4480 * change obdfile creation method
4481 * kernel patch changed
4483 2002-09-19 Peter Braam <braam@clusterfs.com>
4486 - stack overflow bug in extN fixed
4488 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4490 * documentation updates
4491 - add man pages for config tools
4492 - update tests/README to describe testing with new config tools
4493 - finish metadata API descriptions
4494 * bug fixes and cleanups
4495 - statfs workaround for 16TB limit
4496 - LOV stripe allocation improved, can stripe on subset of OSTs
4497 - LOV file size/IO offset was wrong for files > 4GB in size
4498 - object EA data was being dropped, caused files to be unreadable
4499 - memory overflow with non-LOV OST caused memory corruption
4500 - fixed regression tests to work with new config tools, obdfilter
4501 - fixed bug when directory size became larger than 1 block
4502 - fixed bug (for single client case) when PWD was deleted
4503 - invalidate local directory pages when doing intent-based ops
4504 - avoid LDLM oops when lock callback contained bad data
4506 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4508 * documentation updates
4509 * bug fixes and cleanups
4510 - configuration tools
4513 - 64-bit compile warnings
4514 - 64-bit internal statfs data
4516 * test_brw on persistent OST devices
4518 * lprocfs (disabled)
4520 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4522 * documentation updates
4523 * bug fixes and cleanups
4524 * configuration tools
4526 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4529 * many small fixes to 0_5_4
4530 * io/network handling
4531 * thinkos in MDS operations
4533 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4536 * crucial basic fixes to 0.5.3
4537 * IOR, Iozone work over Elan
4540 2002-08-07 Phil Schwan <phil@clusterfs.com>
4541 * version 0_5_3, our first alpha
4542 * we use the new Portals iovs
4543 * documentation updates
4544 * bug fixes and cleanups
4545 * small changes in the DLM wire protocol
4547 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4548 * version 0_5_1 with some initial stability,
4549 * locking on MD and file I/O.
4550 * documentation updates
4551 * several bug fixes since 0.5.0
4552 * small changes in wire protocol
4554 2002-07-18 Phil Schwan <phil@clusterfs.com>
4556 * delivered as Lustre Light Alpha
4557 * fixed a crash after handling invalid MDS requests
4558 * fixed directory pages for architectures with non-4k pages sizes
4560 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4561 * release version v0_4_4
4562 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4564 * Fixes a number of interruption problems with OST operations.
4565 * Update documentation for portals header changes
4566 * Move all wire protocol structs/defines to lustre_idl.h
4567 * Fixes symlink length bug.
4568 * Add tcpdump to repository.
4570 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4571 * release version v0_4_3
4572 * Fixes statfs for inodes on extN.
4573 * Fixes bug in runtests which would delete /etc/hosts.
4574 * Use 64-bit object IDs wherever possible (not into VFS though)
4575 Remove ost_get_info, which is unused by lustre, and out of date.
4577 2002-07-03 Peter Braam <braam@clusterfs.com>
4578 * release version v0_4_2 Fixes a lookup error (type not passed)
4579 * move forward to head of Portals
4580 * move forward to latest Lustre kernel
4582 2002-06-25 Peter Braam <braam@clusterfs.com>
4583 * release version v0_4_1. Hopefully stable on single node use.