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 2.6.18-53.1.13.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).
16 Frequency : frequent on X2 node
18 Description: mdc_set_open_replay_data LBUG
19 Details : Set replay data for requests that are eligible for replay.
23 Description: lustre_mgs: operation 101 on unconnected MGS
24 Details : When MGC is disconnected from MGS long enough, MGS will evict the
25 MGC, and late on MGC cannot successfully connect to MGS and a lot
26 of the error messages complaining that MGS is not connected.
29 Frequency : on start mds
31 Description: Implement get_info(last_id) in obdfilter.
34 Frequency : occasional
36 Description: Correctly check stale fid, not start epoch if ost not support SOM
37 Details : open with flag O_CREATE need set old fid in op_fid3 because op_fid2
38 overwrited with new generated fid, but mds can anwer with one of these
39 two fids and both is not stale. setattr incorectly start epoch and
40 assume will be called done_writeting, but without SOM done_writing
44 Frequency : rare, depends on device drivers and load
46 Description: MDS or OSS nodes crash due to stack overflow
47 Details : Code changes in 1.8.0 increased the stack usage of some functions.
48 In some cases, in conjunction with device drivers that use a lot
49 of stack the MDS (or possibly OSS) service threads could overflow
50 the stack. One change which was identified to consume additional
51 stack has been reworked to avoid the extra stack usage.
54 Frequency : occasional
56 Description: Do not fail import if osc_interpret_create gets -EAGAIN
57 Details : If osc_interpret_create got -EAGAIN it immediately exits and
58 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
59 oscc_has_objects and see OSC has no objests and call
60 oscc_internal_create to resend create request.
62 Severity : enhancement
64 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
66 Severity : enhancement
68 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
73 Description: Fix replay if there is an un-replied request and open
74 Details : In some cases, older replay request will revert the
75 mcd->mcd_last_xid on MDS which is used to record the client's
78 Severity : enhancement
80 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
82 Severity : enhancement
84 Description: Add rhel5 support to HEAD.
86 Serverity : enhancement
88 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
93 Description: Don't allow skipping OSTs if index has been specified.
94 Details : Don't allow skipping OSTs if index has been specified, make locking
95 in internal create lots better.
99 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
100 Details : access to bitfield in structure is always rounded to long
101 and this produce problem with not atomic change any bit.
105 Description: Lustre make rpms failed.
106 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
107 builds Lustre rpms from tarball.
110 Frequency : rare on shutdown ost
112 Description: If llog cancel was not send before clean_exports phase, this can
113 produce deadlock in llog code.
114 Details : If llog thread has last reference to obd and call class_import_put
115 this produce deadlock because llog_cleanup_commit_master wait when
116 last llog_commit_thread exited, but this never success because was
117 called from llog_commit_thread.
121 Description: allow userland application know is lost one of stripes.
122 Details : fill lvb_blocks with error code on ost and return it to
123 application if error flag found.
127 Description: NULL lov_tgts causing MDS oops
128 Details : more safe checks for NULL lov_tgts for avoid oops.
130 Severity : enhancement
132 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
136 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
137 Details : If an OST is down the MDS will hang indefinitely in
138 obd_statfs_rqset() waiting for the statfs data. While for
139 MDS QOS usage of statfs, it should not stuck in waiting.
141 Severity : enhancement
143 Description: remote_acl support
144 Details : Support ACL-based permission check for remote user.
145 Support setfacl/getfacl for remote user with the utils
146 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
147 the system "{s,g}etfacl" utils.
149 Severity : enhancement
151 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
153 Severity : enhancement
155 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
159 Description: make mgs_setparam() handle fsname containing dash
160 Details : fsname containing a dash does not work with lctl conf_param
162 Severity : enhancement
164 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
166 Severity : enhancement
168 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
172 Frequency : rare, at shutdown
173 Description: access already free / zero obd_namespace.
174 Details : if client_disconnect_export was called without force flag set,
175 and exist connect request in flight, this can produce access to
176 NULL pointer (or already free pointer) when connect_interpret
177 store ocd flags in obd_namespace.
181 Frequency : only at startup
182 Description: not alloc memory with spinlock held.
183 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
184 if any spinlock held.
186 Severity : enhancement
188 Description: make lustre randomly fail allocating memory
189 Details : Make lustre randomly failed allocating memory for testing purpose.
191 Severity : enhancement
193 Description: lost problems with lov objid file
194 Details : Fixes some scability and access to not inited memory problems
195 in work with lov objdid file.
200 Description: lfs find does not continue on file error
201 Details : Continue other files processing when a file/dir is absent.
205 Description: Inconsistent usage of lustre_pack_reply()
206 Details : Standardize the usage of lustre_pack_reply() such that it
207 always generate a CERROR on failure.
210 Frequency : occasional
212 Description: MDS hang or stay in waiting lock
213 Details : If client receive lock with CBPENDING flag ldlm need send lock
214 cancel as separate rpc, to avoid situation when cancel request
215 can't processed due all i/o threads stay in wait lock.
219 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
223 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
227 Description: Update to SLES9 kernel-2.6.5-7.287.3.
231 Description: Update to RHEL4 latest kernel.
233 Severity : enhancement
235 Description: Build SLES10 patchless client fails
236 Details : The configure was broken by run ./configure with
237 --with-linux-obj=.... argument for patchless client. When the
238 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
239 can't search header adequately. Use absolute path such as
240 -I($LINUX)/include instead.
244 Description: interrupt oig_wait produce painc on resend.
245 Details : brw_redo_request can be used for resend requests from ptlrpcd and
246 private set, and this produce situation when rq_ptlrpcd_data not
247 copyed to new allocated request and triggered LBUG on assert
248 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
249 ptlrpcd set if request is changed and can be safety changed to use
254 Description: LASSERT_{REQ,REP}SWAB macros are buggy
255 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
256 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
257 these macros and replace them with their difinitions instead.
261 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
262 Details : Update which_patch & target file for SLES10 latest kernel.
265 Description: Remove client patches from SLES 10 kernel.
266 Details : This causes SLES 10 clients to behave as patchless clients
267 even on a Lustre-patched (server) kernel.
269 Severity : enhancement
271 Description: self-adjustable client's lru lists
272 Details : use adaptive algorithm for managing client cached locks lru
273 lists according to current server load, other client's work
274 pattern, memory activities, etc. Both, server and client
275 side namespaces provide number of proc tunables for controlling
278 Severity : enhancement
280 Description: light-weight GSS support
281 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
282 and reduce performance overhead.
284 Severity : enhancement
286 Description: Linux keyring support
287 Details : Support using service of Linux keyring for Lustre GSS internal
288 context refresh/cache mechanism.
292 Description: Fix errors in lfs documentation
293 Details : Fixes man pages
297 Description: don't use GFP_* in generic Lustre code.
298 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
302 Description: obdclass is limited by single OBD_ALLOC(idarray)
303 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
307 Description: warnings with build patchless client with vanila 2.6.19 and up
308 Details : change old ctl_table style and replace ctl_table/ctl_table_header
309 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
313 Frequency : Only for SLES
314 Description: sanity_quota fail test_1
315 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
316 file, which makes TSTID[2] inunique.
320 Frequency : Only for SLES9
321 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
322 Details : Update target/ChangeLog/which_patch .
326 Frequency : Only for test_10 in sanity-quota.sh
327 Description: a bug in quota test code
328 Details : checking a test flag in a wrong place
332 Frequency : Only for test_13 in sanity-quota.sh
333 Description: a bug in quota test script
334 Details : shouldn't check overfull allocated quota
338 Description: jbd statistics
339 Details : Port older jbd statistics patch for sles10
341 Severity : enhancement
343 Description: Add printing inode info into message about error in writepage.
348 Description: reply_lock_interpret crash due to race with it and lock cancel.
349 Details : Do not replay locks that are being cancelled. Do not reference
350 locks by their address during replay, just by their handle.
354 Frequency : When flocks are used.
355 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
356 Details : Flock locks might destroy just granted lock if it could be merged
357 with another existing flock, this is done in completion handler,
358 so teach ldlm_cli_enquque_fini that this is a valid case for
364 Description: Oops in read and write path when failing to allocate lock.
365 Details : Check if lock allocation failed and return error back.
369 Description: lstripe command fails for valid OST index
370 Details : The stripe offset is compared to lov->desc.ld_tgt_count
371 instead of lov->desc.ld_active_tgt_count.
375 Description: sanity.sh failed test 103
376 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
381 Description: df doesn't work properly if diskfs blocksize != 4K
382 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
386 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
387 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
391 Description: obd_health_check_timeout too short
392 Details : set obd_health_check_timeout as 1.5x of obd_timeout
396 Description: llapi_file_create() does not allow some changes
397 Details : add llapi_file_open() that allows specifying the mode and
398 open flags, and also returns an open file handle.
402 Description: merge and cleanup kernel patches.
403 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
407 Description: Add journal checksum support.(Kernel part)
408 Details : The journal checksum feature adds two new flags i.e
409 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
410 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
411 indicates that the commit block contains the checksum for
412 the blocks described by the descriptor blocks. Now commit
413 record can be sent to disk without waiting for descriptor
414 blocks to be written to disk. This behavior is controlled
415 using JBD2_FEATURE_ASYNC_COMMIT flag.
419 Description: OSS needs mutliple precreate threads
420 Details : Add ability to start more than one create thread per OSS.
424 Description: Sanity.sh test_65k failed: setstripe should have succeeded
425 Details : execute lfs setstripe on client
429 Description: mds_obd_create error creating tmp object
430 Details : When the user sets quota on root, llog will be affected and can't
431 create files and write files.
434 Frequency : Always on ia64 patchless client, and possibly others.
436 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
437 Details : This allows the patchless client to be loaded on architectures
442 Description: RedHat Update kernel for RHEL5
443 Details : Add the kernel config file for RHEL5.
447 Description: RedHat Update kernel for RHEL5
448 Details : Modify the kernel config file more closer RHEL5.
452 Description: Build failure against Centos5 (RHEL5)
453 Details : Define PAGE_SIZE when it isn't present.
457 Description: test_7 on sanity-quota.sh can't be used on separate machines
458 Details : add facet to handle it
462 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
463 Details : with NFS, the anon dentry's parent was set to itself in
464 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
465 corresponding dentry other than use rec->ur_name.
467 Severity : enhancement
469 Description: lfs setstripe enhancement
470 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
474 Description: enable data checksumming by default
475 Details : enable checksum by default, allow --disable-checksum
476 configure option and "-o nochecksum" mount option. Checksums
477 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
481 Description: lustre support for RHEL5
482 Details : Add support for RHEL5.
486 Description: Client eviction due to failover config
487 Details : after a connection loss, the lustre client should attempt to
488 reconnect to the last active server first before trying the
489 other potential connections.
493 Description: when mds and osts use different quota unit(32bit and 64bit),
494 quota will be released repeatly.
495 Details : void sending multiple quota reqs to mds, which will keep the
496 status between the reqs.
500 Description: osts not allocated evenly to files
501 Details : change the condition to increase offset_idx
504 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
506 Description: Data corruption for OSTs that are formatted larger than 2TB
508 Details : When generating the bio request for lustre file writes the
509 sector number would overflow a temporary variable before being
510 used for the IO. The data reads correctly from Lustre (which
511 will overflow in a similar manner) but other file data or
512 filesystem metadata may be corrupted in some cases.
516 Description: Tune the kernel for good SCSI performance.
517 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
518 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
523 Description: rewrite ext2-derived code in obdclass/uuid.c
524 Details : rewrite inherited code (uuid parsing code from ext2 utils)
525 from scratch preserving functionality.
529 Description: "lfs find -obd UUID" prints directories
530 Details : "lfs find -obd UUID" will return all directory names instead
531 of just file names. It is incorrect because the directories
532 do not reside on the OSTs.
536 Description: use i_size_read and i_size_write in 2.6 port
537 Details : replace inode->i_size access with i_size_read/write()
540 Frequency : when removing large files
542 Description: scheduling issue during removal of large Lustre files
543 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
544 It causes scheduling issues when removing large files (17TB in the
548 Frequency : only with liblustre clients on XT3
550 Description: evictions taking too long
551 Details : allow llrd to evict clients directly on OSTs
554 Frequency : only on ppc
556 Description: /proc/fs/lustre/devices broken on ppc
557 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
558 We should make sure the type of 'index' is loff_t and then cast
559 explicitly as needed below. Do not assign an explicitly cast
563 Frequency : only for rhel5
565 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
566 Details : Modify the target file & which_kernel.
568 Severity : enhancement
570 Description: omit set fsid for export NFS
571 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
572 as FS_REQUIRES_DEV for avoid problems with generate fsid.
576 Frequency : Always, for kernels after 2.6.16
577 Description: Fix warning idr_remove called for id=.. which is not allocated.
578 Details : Last kernels save old s_dev before kill super and not allow
579 to restore from callback - restore it before call kill_anon_super.
583 Description: buffer overruns could theoretically occur
584 Details : llapi_semantic_traverse() modifies the "path" argument by
585 appending values to the end of the origin string, and a
586 overrun may occur. Adding buffer overrun check in liblustreapi.
590 Description: Fix error on 'ls .' at the top of the Lustre mount.
591 Details : Don't revalidate dentry if it is a root dentry.
595 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
596 Details : Modify vm-tunables-rhel4.patch.
600 Description: Kernel config for 2.6.18-vanilla.
601 Details : Modify targets/2.6-vanilla.target.in.
602 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
603 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
604 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
605 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
609 Description: improve handling recoverable errors
610 Details : if request processig with error which can be recoverable on server
611 request should be resend, otherwise page released from cache and
616 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
617 Details : Modify target file & which_patch.
618 A flaw was found in the IA32 system call emulation provided
619 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
620 value could be stored in the %RAX register, which could trigger an
621 out-of-bounds system call table access. An untrusted local user
622 could exploit this flaw to run code in the kernel
623 (ie a root privilege escalation). (CVE-2007-4573).
627 Description: O_DIRECT bypasses client statistics.
628 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
629 read_ahead_stats were updated. copied the stats section from
630 osc_send_oap_rpc() into async_internal().
634 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
638 Description: change order of libsysio includes
639 Details : '#include sysio.h' should always come before '#include xtio.h'
643 Description: handle "serious error: objid * already exists" more gracefully
644 Details : If LAST_ID value on disk is smaller than the objects existing in
645 the O/0/d* directories, it indicates disk corruption and causes an
646 LBUG(). If the object is 0-length, then we should use the existing
647 object. This will help to avoid a full fsck in most cases.
649 Severity : enhancement
651 Description: adapt the lustre_config script to support the upgrade case
652 Details : Add "-u" option for lustre_config script to support upgrading 1.4
653 server targets to 1.6 in parallel.
657 Description: To avoid grant space > avaible space when the disk is almost
658 full. Without this patch you might see the error "grant XXXX >
659 available" or some LBUG about grant, when the disk is almost
661 Details : In filter_check_grant, for non_grant cache write, we should
662 check the left space by if (*left > ungranted + bytes), instead
663 of (*left > ungranted), because only we are sure the left space
664 is enough for another "bytes", then the ungrant space should be
665 increase. In client, we should update cl_avail_grant only there
666 is OBD_MD_FLGRANT in the reply.
670 Description: Update RHEL 4 kernel to fix local root privilege escalation.
671 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
672 described in CVE-2007-4573. This problem could allow untrusted
673 local users to gain root access.
676 Frequency : when using O_DIRECT and quotas
678 Description: Incorrect file ownership on O_DIRECT output files
679 Details : block usage reported by 'lfs quota' does not take into account
680 files that have been written with O_DIRECT.
685 Description: touch file failed when fs is not full
686 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
687 otherwise we can get ENOSP while fs is not full.
691 Description: parallel lock callbacks
692 Details : Instead of sending blocking and completion callbacks as separated
693 requests, adding them to a set and sending in parallel.
696 Frequency : only for Cray XT3
697 Bugzilla : 12829/13455
698 Description: Changing primary group doesn't change the group lustre assigns to
700 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
701 overridden with the primary group provided by the group upcall,
702 whereas the supplied fsgid can be trusted if it is in the list of
703 supplementary groups returned by the group upcall.
705 Severity : enhancement
707 Description: Allow masking D_WARNING, D_ERROR messages from console
708 Details : Console messages can now be disabled via lnet.printk.
712 Description: User code with malformed file open parameter crashes client node
713 Details : Before packing join_file req, all the related reference should be
714 checked carefully in case some malformed flags cause fake
715 join_file req on client.
719 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
720 Details : ldlm_completion_ast() assumes that a lock is granted when the req
721 mode is equal to the granted mode. However, it should also check
722 that LDLM_FL_CP_REQD is not set.
726 Description: Heavy nfs access might result in deadlocks
727 Details : After ELC code landed, it is now improper to enqueue any mds
728 locks under och_sem, because enqueue might want to decide to
729 cancel open locks for same inode we are holding och_sem for.
733 Description: Client eviction while running blogbench
734 Details : A lot of unlink operations with concurrent I/O can lead to a
735 deadlock causing evictions. To address the problem, the number of
736 oustanding OST_DESTROY requests is now throttled to
737 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
738 callbacks are processed in priority.
742 Description: enable ACLs on MDS by default
743 Details : ACLs must be enabled on MDS by default.
746 Frequency : PPC/PPC64 only
748 Description: conflicts between asm-ppc64/types.h and lustre_types.h
749 Details : fix duplicated definitions between asm-ppc64/types.h and
750 lustre_types.h on PPC.
753 Frequency : PPC/PPC64 only
755 Description: asm-ppc/segment.h does not exist
756 Details : fix compile issue on PPC.
760 Description: better handle error messages in extents code
763 Frequency : RHEL4 only
765 Description: mkfs is very slow on IA64/RHEL4
766 Details : A performance regression has been discovered in the MPT Fusion
767 driver between versions 3.02.73rh and 3.02.99.00rh. As a
768 consequence, we have downgraded the MPT Fusion driver in the RHEL4
769 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
771 Severity : enhancement
773 Description: SNMP support enhancement
774 Details : Adding total number of sampled request for an MDS node in snmp
777 Severity : enhancement
779 Description: Optimize ldlm waiting list processing for PR extent locks
780 Details : When processing waiting list for read extent lock and meeting read
781 lock that is same or wider to it that is not contended, skip
782 processing rest of the list and immediatelly return current
783 status of conflictness, since we are guaranteed there are no
784 conflicting locks in the rest of the list.
788 Description: Time out and refuse to reconnect
789 Details : When the failover node is the primary node, it is possible
790 to have two identical connections in imp_conn_list. We must
791 compare not conn's pointers but NIDs, otherwise we can defeat
792 connection throttling.
796 Description: port llog fixes from b1_6 into HEAD
797 Details : Port llog reference couting and some llog cleanups from b1_6
798 (bug 10800) into HEAD, for protect from panic and access to already
799 free llog structures.
803 Description: Detect stride IO mode in read-ahead
804 Details : When a client does stride read, read-ahead should detect that and
805 read-ahead pages according to the detected stride pattern.
809 Description: data checksumming impacts single node performance
810 Details : add support for several checksum algorithm. Currently, only CRC32
811 and Adler-32 are supported. The checksum type can be changed on
812 the fly via /proc/fs/lustre/osc/*/checksum_type.
816 Description: use adler32 for page checksums
817 Details : when available, use the Adler-32 algorithm instead of CRC32 for
820 --------------------------------------------------------------------------------
822 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
824 * Support for kernels:
825 2.6.5-7.283 (SLES 9),
826 2.6.9-55.EL (RHEL 4),
827 2.6.16.46-0.14 (SLES 10),
828 2.6.18.8 vanilla (kernel.org)
829 * Client support for unpatched kernels:
830 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
831 2.6.16 - 2.6.22 vanilla (kernel.org)
832 * Due to recently discovered recovery problems, we do not recommend
833 using patchless RHEL 4 clients with this or any earlier release.
834 * Recommended e2fsprogs version: 1.39.cfs8
835 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
836 * Starting with this release, the ldiskfs backing filesystem required
837 by Lustre is now in its own package, lustre-ldiskfs. This package
838 should be installed. It is versioned separately from Lustre and
839 may be released separately in future.
843 Description: block reactivating mgc import until all deactivates complete
844 Details : Fix race when failing back MDT/MGS to itself (testing)
847 Frequency : at statup only
849 Description: mds_lov_synchronize race leads to various problems
850 Details : simultaneous MDT->OST connections at startup can cause the
851 sync to abort, leaving the OSC in a bad state.
853 Severity : enhancement
855 Description: add optional extra BUILD_VERSION info
856 Details : add a new environment variable (namely LUSTRE_VERS) which allows
857 to override the lustre version.
860 Frequency : 2.6.18 servers only
862 Description: ll_kern_mount() doesn't release the module reference
863 Details : The ldiskfs module reference count never drops down to 0
864 because ll_kern_mount() doesn't release the module reference.
869 Description: server LBUG when using old ost_num_threads parameter
870 Details : Accept the old ost_num_threads parameter but warn that it
871 is deprecated, and fix an off-by-one error that caused an LBUG.
876 Description: Transient SCSI error results in persistent IO issue
877 Details : iobuf->dr_error is not reinitialized to 0 between two
881 Frequency : sometimes when underlying device returns I/O errors
883 Description: OSTs not going read-only during write failures
884 Details : OSTs are not remounted read-only when the journal commit threads
885 get I/O errors because fsfilt_ext3 calls journal_start/stop()
886 instead of the ext3 wrappers.
890 Description: poor connect scaling with increasing client count
891 Details : Don't run filter_grant_sanity_check for more than 100 exports
892 to improve scaling for large numbers of clients.
895 Frequency : SLES10 only
897 Description: sanity-quota.sh quotacheck failed: rc = -22
898 Details : Quotas cannot be enabled on SLES10.
901 Frequency : liblustre clients only
903 Description: getdirentries does not give error when run on compute nodes
904 Details : getdirentries does not fail when the size specified as an argument
905 is too small to contain at least one entry
907 Severity : enhancement
909 Description: Add LNET router traceability for debug purposes
910 Details : If a checksum failure occurs with a router as part of the
911 IO path, the NID of the last router that forwarded the bulk data
912 is printed so it can be identified.
917 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
918 Details : Due to a race condition, liblustre clients were occasionally
921 Severity : enhancement
923 Description: lfs setstripe use optional parameters instead of postional
926 Severity : enhancement
928 Description: Nanosecond timestamp support for ldiskfs
929 Details : The on-disk ldiskfs filesystem has added support for nanosecond
930 resolution timestamps. There is not yet support for this at
931 the Lustre filesystem level.
934 Frequency : during server recovery
936 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
937 Details : request with rq_no_resend flag not awake l_wait_event if they get a
941 Frequency : nfs export on patchless client
943 Description: connectathon hang when test nfs export over patchless client
944 Details : Disconnected dentry cannot be found with lookup, so we do not need
945 to unhash it or make it invalid
948 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
949 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
950 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
955 Description: open req refcounting wrong on reconnect
956 Details : If reconnect happened between getting open reply from server and
957 call to mdc_set_replay_data in ll_file_open, we will schedule
958 replay for unreferenced request that we are about to free.
959 Subsequent close will crash in variety of ways.
960 Check that request is still eligible for replay in
961 mdc_set_replay_data().
966 Description: disable writes to filesystem when reading health_check file
967 Details : the default for reading the health_check proc file has changed
968 to NOT do a journal transaction and write to disk, because this
969 can cause reads of the /proc file to hang and block HA state
970 checking on a healthy but otherwise heavily loaded system. It
971 is possible to return to the previous behaviour during configure
972 with --enable-health-write.
974 Severity : enhancement
976 Description: 64-bit inode version
977 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
978 to the inode. This will be required for version-based recovery.
983 Description: MDS fails to start if a duplicate client export is detected
984 Details : in some rare cases it was possible for a client to connect to
985 an MDS multiple times. Upon recovery the MDS would detect this
986 and fail during startup. Handle this more gracefully.
988 Severity : enhancement
990 Description: Add -o localflock option to simulate old noflock
992 Details : This will achieve local-only flock/fcntl locks
998 Description: log_commit_thread vs filter_destroy race leads to crash
999 Details : Take import reference before releasing llog record semaphore
1004 Description: Wrong request locking in request set processing
1005 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1006 imp_delayed_list, in this place should be used imp_lock.
1009 Frequency : when reconnection
1011 Description: Grant Leak when osc reconnect to OST
1012 Details : When osc reconnect ost, OST(filter) should check whether it
1013 should grant more space to client by comparing fed_grant and
1014 cl_avail_grant, and return the granted space to client instead
1015 of "new granted" space, because client will call osc_init_grant
1016 to update the client grant space info.
1019 Frequency : when client reconnect to OST
1021 Description: Grant Leak when osc do resend and replay bulk write
1022 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1023 bulk write request, because the grant info while be sync between
1024 OSC and OST when reconnect, and we should ignore the grant info
1025 these of resend/replay write req.
1030 Description: Grant space more than avaiable left space sometimes.
1031 Details : When then OST is about to be full, if two bulk writing from
1032 different clients came to OST. Accord the avaliable space of the
1033 OST, the first req should be permitted, and the second one
1034 should be denied by ENOSPC. But if the seconde arrived before
1035 the first one is commited. The OST might wrongly permit second
1036 writing, which will cause grant space > avaiable space.
1039 Frequency : when client is evicted
1041 Description: Grant might be wrongly erased when osc is evicted by OST
1042 Details : when the import is evicted by server, it will fork another
1043 thread ptlrpc_invalidate_import_thread to invalidate the
1044 import, where the grant will be set to 0. While the original
1045 thread will update the grant it got when connecting. So if
1046 the former happened latter, the grant will be wrongly errased
1047 because of this race.
1052 Description: Checking Stale with correct fid
1053 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1054 to check whether it is stale, because sometimes, we want the
1055 enqueue happened anyway, and op_data.fid2 will not be initialized.
1057 Severity : enhancement
1059 Description: update patchless client
1060 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1063 Frequency : only with 2.4 kernel
1065 Description: random memory corruption
1066 Details : size of struct ll_inode_info is to big for union inode.u and this
1067 can be cause of random memory corruption.
1072 Description: Memory leak in recovery
1073 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1074 It should also check obd_fail before fsfilt_start, otherwise if
1075 fsfilt_start return -EROFS,(failover mds during mds recovery).
1076 then the req will return with repmsg->transno = 0 and rc = EROFS.
1077 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1078 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1079 be freed no matter whether fsfilt_commit success or not.
1082 Frequency : only with huge count clients
1084 Description: Prevents from taking the superblock lock in llap_from_page for
1086 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1087 for llap_from_page call in ll_removepage prevents from taking the
1088 superblock lock for a soon died page.
1093 Description: Not check open intent error before release open handle
1094 Details : in some rare cases, the open intent error is not checked before
1095 release open handle, which may cause
1096 ASSERTION(open_req->rq_transno != 0), because it tries to release
1097 the failed open handle.
1102 Description: Set cat log bitmap only after create log success.
1103 Details : in some rare cases, the cat log bitmap is set too early. and it
1104 should be set only after create log success.
1108 Description: Accessing a block bevice can re-enable I/O when Lustre is
1109 tearing down a device.
1110 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1115 Description: service threads may hog cpus when there are a lot of requests
1116 Details : Insert cond_resched to give other threads a chance to use some CPU
1121 Description: the cat log was not initialized in recovery
1122 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1123 unlink log on mds will not be initialized until mds post
1124 recovery. And also in mds post recovery, the unlink log will
1125 initialization will be done asynchronausly, so there will be race
1126 between add unlink log and unlink log initialization.
1130 Description: brw_stats were being printed incorrectly
1131 Details : brw_stats were being printed as log2 but all of them were not
1132 recorded as log2. Also remove some code duplication arising from
1133 filter_tally_{read,write}.
1137 Frequency : rare, only in recovery.
1138 Description: ASSERTION(req->rq_type != LI_POISON) failed
1139 Details : imp_lock should be held while iterating over imp_sending_list for
1140 prevent destroy request after get timeout in ptlrpc_queue_wait.
1144 Description: replay-single.sh test 52 fails
1145 Details : A lock's skiplist need to be cleanup when it being unlinked
1146 from its resource list.
1150 Description: Short directio read returns full requested size rather than
1152 Details : Direct I/O operations should return actual amount of bytes
1153 transferred rather than requested size.
1155 Severity : enhancement
1157 Description: metadata RPC reduction (e.g. for rm performance)
1158 Details : decrease the amount of synchronous RPC between clients and servers
1159 by canceling conflicing lock before the operation on the client
1160 and packing thier handles into the main operation RPC to server.
1162 Severity : enhancement
1164 Description: Async OSC create to avoid the blocking unnecessarily.
1165 Details : If a OST has no remain object, system will block on the creating
1166 when need to create a new object on this OST. Now, ways use
1167 pre-created objects when available, instead of blocking on an
1168 empty osc while others are not empty. If we must block, we block
1169 for the shortest possible period of time.
1171 Severity : enhancement
1173 Description: refine locking for avoid write wrong info into lov_objid file
1174 Details : fix possible races with add new target and write/update data in
1177 --------------------------------------------------------------------------------
1179 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1184 Frequency : on some architectures
1186 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1187 Details : Uninitialized flags sometimes cause configuration commands to
1191 Frequency : patchless clients only
1193 Description: missing __iget() symbol export
1194 Details : The __iget() symbol export is missing. To avoid the need for
1195 this on patchless clients the deathrow inode reaper is turned
1196 off, and we depend on the VM to clean up old inodes. This
1197 dependency was during via the fix for bug 12181.
1199 --------------------------------------------------------------------------------
1201 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1203 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1204 INTEROPERATE with older versions automatically. In many cases a
1205 special upgrade step is needed. Please read the
1206 user documentation before upgrading any part of a live system.
1207 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
1208 version will not interoperate with 1.6 betas before beta5 (1.5.95).
1209 * WARNING: Lustre configuration and startup changes are required with
1210 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1216 Severity : enhancement
1218 Description: MountConf
1219 Details : Lustre configuration is now managed via mkfs and mount
1220 commands instead of lmc and lconf. New obd types (MGS, MGC)
1221 are added for dynamic configuration management. See
1222 https://mail.clusterfs.com/wikis/lustre/MountConf for
1225 Severity : enhancement
1227 Description: dynamic OST addition
1228 Details : OSTs can now be added to a live filesystem
1230 Severity : enhancement
1232 Description: startup order invariance
1233 Details : MDTs and OSTs can be started in any order. Clients only
1234 require the MDT to complete startup.
1236 Severity : enhancement
1238 Description: parallel, asynchronous orphan cleanup
1239 Details : orphan cleanup is now performed in separate threads for each
1240 OST, allowing parallel non-blocking operation.
1242 Severity : enhancement
1244 Description: optimized stripe assignment
1245 Details : stripe assignments are now made based on ost space available,
1246 ost previous usage, and OSS previous usage, in order to try
1247 to optimize storage space and networking resources.
1249 Severity : enhancement
1251 Description: Permanently set tunables
1252 Details : All writable /proc/fs/lustre tunables can now be permanently
1253 set on a per-server basis, at mkfs time or on a live system.
1255 Severity : enhancement
1257 Description: Lustre message v2
1258 Details : Add lustre message format v2.
1260 Severity : enhancement
1262 Description: client OST exclusion list
1263 Details : Clients can be started with a list of OSTs that should be
1264 declared "inactive" for known non-responsive OSTs.
1268 Description: SPEC SFS validation failure on NFS v2 over lustre.
1269 Details : Changes the blocksize for regular files to be 2x RPC size,
1270 and not depend on stripe size.
1272 Severity : enhancement
1274 Description: Multiple MD RPCs in flight.
1275 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1276 To avoid overly-overloading MDS, introduce a limit on number of
1277 MDS RPCs in flight for a single client and add /proc controls
1278 to adjust this limit.
1280 Severity : enhancement
1282 Description: client read/write statistics
1283 Details : Add client read/write call usage stats for performance
1284 analysis of user processes.
1285 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1286 file access. extents_stats shows chunk size distribution.
1287 extents_stats_per_process show chunk size distribution per
1290 Severity : enhancement
1292 Description: mds statistics
1293 Details : Add detailed mds operations statistics in
1294 /proc/fs/lustre/mds/*/stats.
1298 Description: Failure of copying files with lustre special EAs.
1299 Details : Client side always return success for setxattr call for lustre
1300 special xattr (currently only "trusted.lov").
1305 Description: Refcount LNET uuids
1306 Details : The global LNET uuid list grew linearly with every startup;
1307 refcount repeated list entries instead of always adding to
1310 Severity : enhancement
1312 Description: Dynamic service threads
1313 Details : Within a small range, start extra service threads
1314 automatically when the request queue builds up.
1317 Frequency : mixed-endian client/server environments
1319 Description: mixed-endian crashes
1320 Details : The new msg_v2 system had some failures in mixed-endian
1323 Severity : enhancement
1325 Description: Easy OST removal
1326 Details : OSTs can be permanently deactivated with e.g. 'lctl
1327 conf_param lustre-OST0001.osc.active=0'
1329 Severity : enhancement
1331 Description: MGS proc entries
1332 Details : Added basic proc entries for the MGS showing what filesystems
1335 Severity : enhancement
1337 Description: provide MGS failover
1338 Details : Added config lock reacquisition after MGS server failover.
1340 Severity : enhancement
1342 Description: add Linux 2.4 support
1343 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1347 Description: a large application tries to do I/O to the same resource and dies
1348 in the middle of it.
1349 Details : Check the req->rq_arrival time after the call to
1350 ost_brw_lock_get(), but before we do anything about
1351 processing it & sending the BULK transfer request. This
1352 should help move old stale pending locks off the queue as
1353 quickly as obd_timeout.
1356 Frequency : when an incorrect nid is specified during startup
1358 Description: ptlrpc connect to non-existant node causes kernel crash
1359 Details : LNET can't be re-entered from an event callback, which
1360 happened when we expire a message after the export has been
1361 cleaned up. Instead, hand the zombie cleanup off to another
1364 Severity : enhancement
1366 Description: plain/inodebits lock performance improvement
1367 Details : Grouping plain/inodebits in granted list by their request modes
1368 and bits policy, thus improving the performance of search through
1372 Frequency : only if OST filesystem is corrupted
1374 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1375 Details : for a short time RPCs with bulk IO are in the replay list,
1376 but replay of bulk IOs is unimplemented. If the OST filesystem
1377 is corrupted due to disk cache incoherency and then replay is
1378 started it is possible to trip an assertion. Avoid putting
1379 committed RPCs into the replay list at all to avoid this issue.
1382 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1384 Description: System hang on startup
1385 Details : This bug allowed the liblustre (e.g. catamount) client to
1386 return to the app before handling all startup RPCs. This
1387 could leave the node unresponsive to lustre network traffic
1388 and manifested as a server ptllnd timeout.
1390 Severity : enhancement
1392 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1393 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1394 boolean to control whether to print peer debug info when a
1395 client's RPC times out.
1398 Frequency : only for kernels with patches from Lustre below 1.4.3
1400 Description: Remove old rdonly API
1401 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1404 Frequency : only for devices with external journals
1406 Description: Set external device read-only also
1407 Details : During a commanded failover stop, we set the disk device
1408 read-only while the server shuts down. We now also set any
1409 external journal device read-only at the same time.
1412 Frequency : when upgrading from 1.4 while trying to change parameters
1414 Description: The wrong (new) MDC name was used when setting parameters for
1415 upgraded MDT's. Also allows changing of OSC (and MDC)
1416 parameters if --writeconf is specified at tunefs upgrade time.
1419 Frequency : when setting specific ost indicies
1421 Description: QOS code breaks on skipped indicies
1422 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1423 created with --index need not be sequential.
1425 Severity : enhancement
1427 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1428 Details : The uninit_groups feature works in conjunction with the kernel
1429 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1430 up the pass1 processing of e2fsck. This is a read-only feature
1431 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1432 mount filesystems that have had this feature enabled.
1434 Severity : enhancement
1436 Description: Improve multi-block allocation algorithm to avoid fragmentation
1437 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1438 allocation locality and avoid filesystem fragmentation.
1440 ------------------------------------------------------------------------------
1442 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1444 * Support for kernels:
1445 2.4.21-47.0.1.EL (RHEL 3)
1446 2.6.5-7.283 (SLES 9)
1447 2.6.9-42.0.10.EL (RHEL 4)
1448 2.6.12.6 vanilla (kernel.org)
1449 2.6.16.27-0.9 (SLES 10)
1450 * Recommended e2fsprogs version: 1.39.cfs5
1452 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1456 Frequency : occasional, depends on client load and configuration
1457 Bugzilla : 12181, 12203
1458 Description: data loss for recently-modified files
1460 Details : In some cases it is possible that recently written or created
1461 files may not be written to disk in a timely manner (this should
1462 normally be within 30s unless client IO load is very high).
1463 The problem appears as zero-length files or files that are a
1464 multiple of 1MB in size after a client crash or client eviction
1465 that are missing data at the end of the file.
1467 This problem is more likely to be hit on clients where files are
1468 repeatedly created and unlinked in the same directory, clients
1469 have a large amount of RAM, have many CPUs, the filesystem has
1470 many OSTs, the clients are rebooted frequently, and/or the files
1471 are not accessed by other nodes after being written.
1473 The presence of the problem can be detected by looking at
1474 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1475 smaller than the second (nr_unused) then dirty files will not
1476 be flushed automatically to disk. "sync; sleep 10" should be
1477 run several times on the node before unmounting it to update
1478 Lustre (this is also safe to run on nodes without this problem).
1480 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1481 that can cause this same problem, so customers using that kernel
1482 also need to update the kernel in addition to Lustre. In order
1483 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1485 It is normal that files written just before a client crash (less
1486 than 30s) may not yet have been flushed to disk, even for local
1490 Frequency : frequent on thin XT3 nodes
1492 Description: UUID collision on thin XT3 Linux nodes
1493 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1494 randomly, since we relied on an insufficiently-seeded PRNG.
1499 Description: OSS hangs after "All ost request buffers busy"
1500 Details : A deadlock between quota and journal operations caused OSS
1501 hangs after printing "All ost request buffers busy."
1504 Frequency : always on liblustre builds
1506 Description: Cleanup compiler warnings on liblustre
1509 Frequency : always on liblustre builds on XT3
1511 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1515 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1516 > 32000 subdirectories
1517 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1518 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1519 subdirectory count crosses 32000. This will aid e2fsck to
1520 correctly handle more than 32000 subdirectories.
1523 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1525 Description: System hang on startup
1526 Details : This bug allowed the liblustre (e.g. catamount) client to
1527 return to the app before handling all startup RPCs. This
1528 could leave the node unresponsive to lustre network traffic
1529 and manifested as a server ptllnd timeout.
1531 Severity : enhancement
1533 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1534 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1535 boolean to control whether to print peer debug info when a
1536 client's RPC times out.
1541 Description: make O_SYNC working on 2.6 kernels
1542 Details : 2.6 kernels use different method for mark pages for write,
1543 so need add a code to lustre for O_SYNC work.
1548 Description: Failure to close file and release space on NFS
1549 Details : Put inode details into lock acquired in ll_intent_file_open.
1550 Use mdc_intent_lock in ll_intent_open to properly
1551 detect all kind of errors unhandled by mdc_enqueue.
1556 Description: proc file read during shutdown sometimes raced obd removal,
1558 Details : Add lock to prevent obd access after proc file removal.
1561 Frequency : Only for files larger than 4GB on 32-bit clients.
1563 Description: improperly doing page alignment of locks
1564 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1565 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1570 Description: RPCs being resent when they shouldn't be
1571 Details : Some RPCs that should not be resent are being resent. This
1572 can cause inconsistencies in the RPC state machine. Do not
1573 resend such requests.
1576 Frequency : rare, only with NFS export
1578 Description: Crash on NFS re-export node
1579 Details : under very unusual load conditions an assertion is hit in
1580 ll_intent_file_open()
1583 Frequency : only if OST filesystem is corrupted
1585 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1586 Details : for a short time RPCs with bulk IO are in the replay list,
1587 but replay of bulk IOs is unimplemented. If the OST filesystem
1588 is corrupted due to disk cache incoherency and then replay is
1589 started it is possible to trip an assertion. Avoid putting
1590 committed RPCs into the replay list at all to avoid this issue.
1595 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1596 Details : Large single O_DIRECT read and write calls can fail to allocate
1597 a sufficiently large buffer to process the request. In case of
1598 allocation failure the allocation is retried with a smaller
1599 buffer and broken into smaller requests.
1601 Severity : enhancement
1603 Description: Add -o localflock option to simulate old noflock behaviour.
1604 Details : This will achieve local-only flock/fcntl locks coherentness.
1609 Description: versioning check is incomplete
1610 Details : Checking the version difference of client vs. server, report
1611 error if the gap is too big.
1616 Description: add support PG_writeback bit
1617 Details : add support for PG_writeback bit for Lustre, for more carefull
1618 work with page cache in 2.6 kernel. This also fix some deadlocks
1619 and remove hack for work O_SYNC with 2.6 kernel.
1621 Severity : enhancement
1623 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1624 Details : The uninit_groups feature works in conjunction with the kernel
1625 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1626 up the pass1 processing of e2fsck. This is a read-only feature
1627 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1628 mount filesystems that have had this feature enabled.
1630 Severity : enhancement
1632 Description: Improve multi-block allocation algorithm to avoid fragmentation
1633 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1634 allocation locality and avoid filesystem fragmentation.
1636 ------------------------------------------------------------------------------
1638 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1640 * Support for kernels:
1641 2.6.9-42.0.3.EL (RHEL 4)
1642 2.6.5-7.276 (SLES 9)
1643 2.4.21-47.0.1.EL (RHEL 3)
1644 2.6.12.6 vanilla (kernel.org)
1645 2.6.16.21-0.8 (SLES10)
1646 * Recommended e2fsprogs version: 1.39.cfs2-0
1648 * The backwards-compatible /proc/sys/portals symlink has been removed
1649 in this release. Before upgrading, please ensure that you change
1650 any configuration scripts or /etc/sysctl.conf files that access
1651 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1652 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1653 in advance of the upgrade on any system running Lustre 1.4.6 or
1654 newer, since /proc/sys/lnet was added in that version.
1655 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1659 Frequency : only when quota is used
1661 Description: avoid scanning export list for quota master
1662 Details : Change the algorithms to avoid scanning export list in order
1663 to improve the efficiency.
1666 Frequency : MDS failover only, very rarely
1668 Description: "went back in time" messages on mds failover
1669 Details : The greatest transno may be lost when the current operation
1670 finishes with an error (transno==0) and the client's last_rcvd
1671 record is over-written. Save the greatest transno in the
1672 mds_last_transno for this case.
1675 Frequency : always for specific kernels and striping counts
1677 Description: client may get "Matching packet too big" without ACL support
1678 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1679 when trying to access files in certain configurations. The
1680 clients should in fact be denied when mounting because they do
1681 not understand ACLs.
1684 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1686 Description: many clients connecting with IO in progress causes connect timeouts
1687 Details : Avoid synchronous journal commits to avoid delays caused by many
1688 clients connecting/disconnecting when bulk IO is in progress.
1689 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1690 OST_IO_PORTAL to avoid delays behind potentially many pending
1694 Frequency : occasionally with multiple writers to a single file
1696 Description: shared writes to file may result in wrong size reported by stat()
1697 Details : Allow growing of kms when extent lock is cancelled
1700 Frequency : always with random mmap IO to multi-striped file
1702 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1703 Details : Only if the hole is at the end of OST object so that kms is too
1704 small. Fix is to increase kms accordingly in ll_nopage.
1707 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1709 Description: writes to a missing object would leak memory on the OST
1710 Details : If there is an inconsistency between the MDS and OST filesystems,
1711 such that the MDS references an object that doesn't exist, writes
1712 to that object will leak memory due to incorrect cleanup in the
1713 error handling path, eventually running out of memory on the OST.
1718 Description: Creating too long symlink causes lustre errors
1719 Details : Check symlink and name lengths before sending requests to MDS.
1722 Frequency : only if flock is enabled (not on by default)
1724 Description: posix locks not released on fd closure on 2.6.9+
1725 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1726 kernels, this caused such locks not to be released on fd close and
1727 then assertions on fs unmount about still used locks.
1730 Frequency : MDS failover only, very rarely
1732 Description: clients may get ASSERTION(granted_lock != NULL)
1733 Details : When request was taking a long time, and a client was resending
1734 a getattr by name lock request. The were multiple lock requests
1735 with the same client lock handle and
1736 mds_getattr_name->fixup_handle_for_resent_request found one of the
1737 lock handles but later failed with ASSERTION(granted_lock != NULL).
1742 Description: handle->h_buffer_credits > 0, assertion failure
1743 Details : h_buffer_credits is zero after truncate, causing assertion
1744 failure. This patch extends the transaction or creates a new
1748 Frequency : NFS re-export or patchless client
1749 Bugzilla : 11179, 10796
1750 Description: Crash on NFS re-export node (__d_move)
1751 Details : We do not want to hash the dentry if we don't have a lock.
1752 But if this dentry is later used in d_move, we'd hit uninitialised
1753 list head d_hash, so we just do this to init d_hash field but
1754 leave dentry unhashed.
1757 Frequency : NFS re-export or patchless client
1759 Description: NFS exports has problem with symbolic link
1760 Details : lustre client didn't properly install dentry when re-exported
1761 to NFS or running patchless client.
1764 Frequency : NFS re-export or patchless client
1766 Description: Various nfs/patchless fixes.
1767 Details : fixes reuse disconected alias for lookup process - this fixes
1768 warning "find_exported_dentry: npd != pd",
1769 fix permission error with open files at nfs.
1770 fix apply umask when do revalidate.
1773 Frequency : occasional
1775 Description: Crash on NFS re-export node
1776 Details : calling clear_page() on the wrong pointer triggered oops in
1777 generic_mapping_read().
1780 Frequency : rarely, using O_DIRECT IO
1782 Description: unaligned directio crashes client with LASSERT
1783 Details : check for unaligned buffers before trying any requests.
1786 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1788 Description: stale data returned from RAID cache
1789 Details : If only a small amount of IO is done to the RAID device before
1790 reading it again it is possible to get stale data from the RAID
1791 cache instead of reading it from disk.
1794 Frequency : always for sles10 kernel
1796 Description: sles10 support
1797 Details : ll_follow_link: compile fixes and using of nd_set_link
1798 under newer kernels.
1801 Frequency : depends on arch, kernel and compiler version, always on sles10
1804 Description: recursive or deep enough symlinks cause stack overflow
1805 Details : getting rid of large stack-allocated variable in
1809 Frequency : depends on hardware
1811 Description: lustre write performance loss in the SLES10 kernel
1812 Details : the performance loss is caused by using of write barriers in the
1813 ext3 code. The SLES10 kernel turns barrier support on by
1814 default. The fix is to undo that change for ldiskfs.
1816 ------------------------------------------------------------------------------
1818 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1820 * Support for kernels:
1821 2.6.9-42.0.3EL (RHEL 4)
1822 2.6.5-7.276 (SLES 9)
1823 2.4.21-47.0.1.EL (RHEL 3)
1824 2.6.12.6 vanilla (kernel.org)
1828 Frequency : quota enabled and large files being deleted
1830 Description: releasing more than 4GB of quota at once hangs OST
1831 Details : If a user deletes more than 4GB of files on a single OST it
1832 will cause the OST to spin in an infinite loop. Release
1833 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1838 Description: statfs data retrieved from /proc may be stale or zero
1839 Details : When reading per-device statfs data from /proc, in the
1840 {kbytes,files}_{total,free,avail} files, it may appear
1841 as zero or be out of date.
1844 Frequency : systems with MD RAID1 external journal devices
1846 Description: lconf's call to blkid is confused by RAID1 journal devices
1847 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1848 of returning all block devices that match the journal UUID.
1851 Frequency : always, for aggregate stripe size over 4GB
1853 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1854 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1855 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1858 Frequency : always if "lfs find" used on a local file/directory
1860 Description: "lfs find" segfaults if used on a local file/directory
1861 Details : The case where a directory component was not specified wasn't
1862 handled correctly. Handle this properly.
1865 Frequency : always on ppc64
1867 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1868 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1869 perform endianness conversion.
1872 Frequency : rarely (truncate to non-zero file size after write under load)
1873 Bugzilla : 10730, 10687
1874 Description: Files padded with zeros to next 4K multiple
1875 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1876 it is possible that files that are truncated to a non-zero size
1877 immediately after being written are filled with zero bytes beyond
1878 the truncated size. No file data is lost.
1880 Severity : enhancement
1882 Description: Allow recovery/failover for liblustre clients.
1883 Details : liblustre clients were unaware of failover configurations until
1886 Severity : enhancement
1888 Description: user file locks should fail when not mounting with flock option
1889 Details : Set up an error-returning stub in ll_file_operations.lock field
1890 to prevent incorrect behaviour when client is mounted without
1891 flock option. Also, set up properly f_op->flock field for
1895 Frequency : always on ia64
1897 Description: "lfs df" loops on printing out MDS statfs information
1898 Details : The obd_ioctl_data was not initialized and in some systems
1899 this caused a failure during the ioctl that did not return
1900 an error. Initialize the struct and return an error on failure.
1903 Frequency : SLES 9 only
1905 Description: Error of copying files with lustre special EAs as root
1906 Details : Client side always return success for setxattr call for lustre
1907 special xattr (currently only "trusted.lov").
1910 Frequency : rarely on clusters with both ia64+i386 clients
1912 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1913 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1914 clients are doing concurrent writes to the same file it is
1915 possible that the ia64 clients may LASSERT because the OST
1916 extent locks are not PAGE_SIZE aligned. Ensure that grown
1917 locks are always aligned on the request boundary.
1920 Frequency : specific use, occasional
1922 Description: Overwriting in use executable truncates on-disk binary image
1923 Details : If one node attempts to overwrite an executable in use by
1924 another node, we now correctly return ETXTBSY instead of
1925 truncating the file.
1927 Severity : enhancement
1929 Description: Async OSC create to avoid the blocking unnecessarily.
1930 Details : If a OST has no remain object, system will block on the creating
1931 when need to create a new object on this OST. Now, ways use
1932 pre-created objects when available, instead of blocking on an
1933 empty osc while others are not empty. If we must block, we block
1934 for the shortest possible period of time.
1939 Description: chmod on Lustre root is propagated to other clients
1940 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1941 invalid by the follow_mount time.
1944 Frequency : liblustre clients only
1946 Description: Race in 'instant cancel' lock handling could lead to such locks
1947 never to be granted in case of SMP MDS
1948 Details : Do not destroy not yet granted but cbpending locks in
1952 Frequency : replay/resend of open
1954 Description: non null lock assetion failure in mds_intent_policy
1955 Details : Trying to replay/resend lockless open requests resulted in
1956 mds_open() returning 0 with no lock. Now it sets a flag if
1957 a lock is going to be returned.
1959 Severity : enhancement
1961 Description: Checksum enhancements
1962 Details : New checksum enhancements allow for resending RPCs that failed
1965 Severity : enhancement
1967 Description: Tunables on number of dirty pages in cacche
1968 Details : Allow to set limit on number of dirty pages cached.
1973 Description: client crash on unmount - lock still has references
1974 Details : In some error handling cases it was possible to leak a lock
1975 reference on a client while accessing a file. This was not
1976 harmful to the client during operation, but would cause the
1977 client to crash when the filesystem is unmounted.
1980 Frequency : specific case, rare
1982 Description: ETXTBSY on mds though file not in use by client
1983 Details : ETXTBSY is no longer incorrectly returned when attempting to
1984 chmod or chown a directory that the user previously tried to
1985 execute or a currently-executing binary.
1988 Frequency : extremely rare except on liblustre-based clients
1990 Description: Lustre space not freed when files are deleted
1991 Details : Clean up open-unlinked files after client eviction. Previously
1992 the unlink was skipped and the files remained as orphans.
1997 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1998 Details : In some cases it was possible to send a blocking callback to
1999 a client doing a glimpse, even though that client didn't get
2000 a lock granted. When the glimpse lock is cancelled on the OST
2001 the freed lock is left on the waiting list and corrupted the list.
2004 Frequency : all core dumps
2006 Description: Broke core dumps to lustre
2007 Details : Negative dentry may be unhashed if parent does not have UPDATE
2008 lock, but some callers, e.g. do_coredump, expect dentry to be
2009 hashed after successful create, hash it in ll_create_it.
2011 ------------------------------------------------------------------------------
2013 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2015 * Support for kernels:
2016 2.6.9-42.0.2.EL (RHEL 4)
2017 2.6.5-7.276 (SLES 9)
2018 2.4.21-40.EL (RHEL 3)
2019 2.6.12.6 vanilla (kernel.org)
2023 Frequency : always on RHEL 3
2025 Description: Number of open files grows over time
2026 Details : The number of open files grows over time, whether or not
2027 Lustre is started. This was due to a filp leak introduced
2028 by one of our kernel patches.
2030 ------------------------------------------------------------------------------
2032 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2034 * Support for kernels:
2035 2.6.9-42.EL (RHEL 4)
2036 2.6.5-7.276 (SLES 9)
2037 2.4.21-40.EL (RHEL 3)
2038 2.6.12.6 vanilla (kernel.org)
2043 Bugzilla : 5719, 9635, 9792, 9684
2044 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2045 Details : If a server is under heavy load and cannot reply to new
2046 connection requests before the client resends the (re)connect,
2047 the connection handling code can behave badly if two service
2048 threads are concurrently handing separate (re)connections from
2049 the same client. Add better locking to the connection handling
2050 code, and ensure that only a single connection will be processed
2051 for a given client UUID, even if the lock is dropped.
2053 Severity : enhancement
2055 Description: add TCP zero-copy support to kernel
2056 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2057 sends if the hardware supports scatter-gather and checksumming.
2058 This allows socklnd to do client-write and server-read more
2059 efficiently and reduce CPU utilization from skbuf copying.
2062 Frequency : only if NFS exporting from client
2064 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2065 Details : If setting attributes on a file created under NFS that had
2066 never been opened it would be possible to oops the client
2067 if the file had no objects.
2070 Frequency : always for liblustre
2072 Description: liblustre client does MDS+OSTs setattr RPC for each write
2073 Details : When doing a write from a liblustre client, the client
2074 incorrectly issued an RPC to the MDS and each OST the file was
2075 striped over in order to update the timestamps. When writing
2076 with small chunks and many clients this could overwhelm the MDS
2077 with RPCs. In all cases it would slow down the write because
2078 these RPCs are unnecessary.
2080 Severity : enhancement
2082 Description: allow number of MDS service threads to be changed at module load
2083 Details : It is now possible to change the number of MDS service threads
2084 running. Adding "options mds mds_num_threads={N}" to the MDS's
2085 /etc/modprobe.conf will set the number of threads for the next
2086 time Lustre is restarted (assuming the "mds" module is also
2087 reloaded at that time). The default number of threads will
2088 stay the same, 32 for most systems.
2093 Description: OST crash if filesystem is unformatted or corrupt
2094 Details : If an OST is started on a device that has never been formatted
2095 or if the filesystem is corrupt and cannot even mount then the
2096 error handling cleanup routines would dereference a NULL pointer.
2101 Description: NULL pointer deref in llap_from_page.
2102 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2103 incorrect) mapping if the page was truncated/reclaimed while it was
2104 searched for. Check for this condition and skip such pages when
2105 doing readahead. Introduce extra check to llap_from_page() to
2106 verify page->mapping->host is non-NULL (so page is not anonymous).
2109 Frequency : Sometimes when using sys_sendfile
2111 Description: "page not covered by a lock" warnings from ll_readpage
2112 Details : sendfile called ll_readpage without right page locks present.
2113 Now we introduced ll_file_sendfile that does necessary locking
2114 around call to generic_file_sendfile() much like we do in
2118 Frequency : with certain MDS communication failures at client mount time
2120 Description: NULL pointer deref after failed client mount
2121 Details : a client connection request may delayed by the network layer
2122 and not be sent until after the PTLRPC layer has timed out the
2123 request. If the client fails the mount immediately it will try
2124 to clean up before the network times out the request. Add a
2125 reference from the request import to the obd device and delay
2126 the cleanup until the network drops the request.
2129 Frequency : occasionally during client (re)connect
2131 Description: assertion failure during client (re)connect
2132 Details : processing a client connection request may be delayed by the
2133 client or server longer than the client connect timeout. This
2134 causes the client to resend the connection request. If the
2135 original connection request is replied in this interval, the
2136 client may trip an assertion failure in ptlrpc_connect_interpret()
2137 which thought it would be the only running connect process.
2140 Frequency : only with obd_echo servers and clients that are rebooted
2142 Description: kernel BUG accessing uninitialized data structure
2143 Details : When running an obd_echo server it did not start the ping_evictor
2144 thread, and when a client was evicted an uninitialized data
2145 structure was accessed. Start the ping_evictor in the RPC
2146 service startup instead of the OBD startup.
2148 Severity : enhancement
2149 Bugzilla : 10193 (patchless)
2150 Description: Remove dependency on various unexported kernel interfaces.
2151 Details : No longer need reparent_to_init, exit_mm, exit_files,
2152 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2155 Frequency : rare (only users of deprecated and unsupported LDAP config)
2157 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2158 Details : LDAP apparently contains 'lustreName' attributes instead of
2159 'name'. A simple remapping of the name is sufficient.
2162 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2164 Description: waiting_locks_callback trips kernel BUG if client is evicted
2165 Details : Running with the dump_on_timeout debug flag turned on makes
2166 it possible that the waiting_locks_callback() can try to dump
2167 the Lustre kernel debug logs from an interrupt handler. Defer
2168 this log dumping to the expired_lock_main() thread.
2170 Severity : enhancement
2172 Description: Support NFS exporting on 2.6 kernels.
2173 Details : Implement non-rawops metadata methods for NFS server to use without
2174 changing NFS server code.
2177 Frequency : very rare (synthetic metadata workload only)
2179 Description: two racing renames might cause an MDS thread to deadlock
2180 Details : Running the "racer" program may cause one MDS thread to rename
2181 a file from being the source of a rename to being the target of
2182 a rename at exactly the same time that another thread is doing
2183 so, and the second thread has already enqueued these locks after
2184 doing a lookup of the target and is trying to relock them in
2185 order. Ensure that we don't try to re-lock the same resource.
2188 Frequency : only very large systems with liblustre clients
2190 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2191 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2192 all OSTs in parallel. This allows the eviction of stale liblustre
2193 clients to proceed much faster than if they were done in series,
2194 and also offers similar improvements for other set_info RPCs.
2199 Description: excessive CPU usage during initial read phase on client
2200 Details : During the initial read phase on a client, it would agressively
2201 retry readahead on the file, consuming too much CPU and impacting
2202 performance (since 1.4.5.8). Improve the readahead algorithm
2203 to avoid this, and also improve some other common cases (read
2204 of small files in particular, where "small" is files smaller than
2205 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2210 Description: MDS crash when receiving packet with unknown intent.
2211 Details : Do not LBUG in unknown intent case, just return -EFAULT
2213 Severity : enhancement
2214 Bugzilla : 9293, 9385
2215 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2216 Details : Do not serialize getattr (non-intent version) and statfs.
2219 Frequency : occasional, when OST network is overloaded/intermittent
2221 Description: client evicted by OST after bulk IO timeout
2222 Details : If a client sends a bulk IO request (read or write) the OST
2223 may evict the client if it is unresposive to its data GET/PUT
2224 request. This is incorrect if the network is overloaded (takes
2225 too long to transfer the RPC data) or dropped the OST GET/PUT
2226 request. There is no need to evict the client at all, since
2227 the pinger and/or lock callbacks will handle this, and the
2228 client can restart the bulk request.
2231 Frequency : Always when mmapping file with no objects
2233 Description: client crashes when mmapping file with no objects
2234 Details : Check that we actually have objects in a file before doing any
2235 operations on objects in ll_vm_open, ll_vm_close and
2241 Description: Request leak when working with deleted CWD
2242 Details : Introduce advanced request refcount tracking for requests
2243 referenced from lustre intent.
2245 Severity : Enhancement
2247 Description: Cache open file handles on client.
2248 Details : MDS now will return special lock along with openhandle, if
2249 requested and client is allowed to hold openhandle, even if unused,
2250 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2251 closing files for every read/write openration.
2253 Severity : Enhancement
2255 Description: Cache open negative dentries on client when possible.
2256 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2257 negative dentries on lock revocation.
2262 Description: Remounting a client read-only wasn't possible with a zconf mount
2263 Details : It wasn't possible to remount a client read-only with llmount.
2265 Severity : enhancement
2266 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2267 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2268 MPICH is included to add Lustre-specific ADIO interfaces.
2269 This is based closely on the UFS ADIO layer and only differs
2270 in file creation, in order to allow the OST striping to be set.
2271 This is user-contributed code and not supported by CFS.
2276 Description: extended inode attributes (immutable, append-only) work improperly
2277 when 2.4 and 2.6 kernels are used on client/server or vice versa
2278 Details : Introduce kernel-independent values for these flags.
2280 Severity : enhancement
2283 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2284 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2285 granularity. Now, Fractional MB granularity is be supported,
2286 this is very useful for low memory system.
2288 Severity : enhancement
2290 Description: Getattr by fid
2291 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2292 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2293 have locks and avoids getattr rpc requests for it on every stat.
2296 Frequency : Always, for filesystems larger than 2TB
2298 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2299 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2300 filesystems larger than 2TB to be mounted with mballoc enabled.
2303 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2305 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2306 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2307 filesystems larger than 2TB is mounted then the kernel will
2308 silently corrupt the start of the filesystem. CONFIG_LBD is
2309 enabled for all CFS-supported kernels, but the possibility of
2310 this happening with a modified kernel config exists.
2312 Severity : enhancement
2314 Description: add client O_DIRECT support for 2.6 kernels
2315 Details : It is now possible to do O_DIRECT reads and writes to files
2316 in the Lustre client mountpoint on 2.6 kernel clients.
2318 Severity : enhancement
2320 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2321 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2322 parallel, not waiting for response from every OST before sending
2323 a rpc to the next OST.
2328 Description: setattr vs write race when updating file timestamps
2329 Details : Client processes that update a file timestamp into the past
2330 right after writing to the file (e.g. tar) it is possible that
2331 the updated file modification time can be reset to the current
2332 time due to a race between processing the setattr and write RPC.
2334 Severity : enhancement
2336 Description: Bring 'lfs find' closer in line with regular Linux find.
2337 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2338 -print0 options and obtains all the needed info through the lustre
2341 Severity : enhancement
2343 Description: support up to 1024 configured devices on one node
2344 Details : change obd_dev array from statically allocated to dynamically
2345 allocated structs as they are first used to reduce memory usage
2350 Description: Flush dirty partially truncated pages during truncate
2351 Details : Immediatelly flush partially truncated pages in filter_setattr,
2352 this way we completely avoid having any pages in page cache on OST
2353 and can retire ugly workarounds during writes to flush such pages.
2358 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2359 Details : protect inode from truncation within vfs_unlink() context
2360 just take a reference before calling vfs_unlink() and release it
2361 when parent's i_sem is free.
2366 Description: last_id value checked outside lock on OST caused LASSERT failure
2367 Details : If there were multiple MDS->OST object precreate requests in
2368 flight, it was possible that the OST's last object id was checked
2369 outside a lock and incorrectly tripped an assertion. Move checks
2370 inside locks, and discard old precreate requests.
2373 Frequency : always, if extents are used on OSTs
2375 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2376 Details : OSTs using the extents format would not zero the high 16 bits of
2377 the index physical block number. This is not a problem for any
2378 OST filesystems smaller than 16TB, and no kernels support ext3
2379 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2380 new/modified files) and can be fixed for existing filesystems
2381 with e2fsprogs-1.39-cfs1.
2386 Description: import connection selection may be incorrect if timer wraps
2387 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2388 import connections to be ignored if the 32-bit jiffies counter
2389 wraps. Use a 64-bit jiffies counter.
2392 Frequency : very large clusters immediately after boot
2394 Description: LNET request buffers exhausted under heavy short-term load
2395 Details : If a large number of client requests are generated on a service
2396 that has previously never seen so many requests it is possible
2397 that the request buffer growth cannot keep up with the spike in
2398 demand. Instead of dropping incoming requests, they are held in
2399 the LND until the RPC service can accept more requests.
2402 Frequency : Sometimes during replay
2404 Description: Assertion failure in ll_local_open after replay.
2405 Details : If replay happened on an open request reply before we were able
2406 to set replay handler, reply will become not swabbed tripping the
2407 assertion in ll_local_open. Now we set the handler right after
2408 recognising of open request
2411 Frequency : very rare
2413 Description: kernel reports "badness in vsnprintf"
2414 Details : Reading from the "recovery_status" /proc file in small chunks
2415 may cause a negative length in lprocfs_obd_rd_recovery_status()
2416 call to vsnprintf() (which is otherwise harmless). Exit early
2417 if there is no more space in the output buffer.
2419 Severity : enhancement
2421 Description: clear OBD RPC statistics by writing to them
2422 Details : It is now possible to clear the OBD RPC statistics by writing
2423 to the "stats" file.
2428 Description: Client mtime is not the same on different clients after utimes
2429 Details : In some cases, the client was using the utimes() syscall on
2430 a file cached on another node. The clients now validate the
2431 ctime from the MDS + OSTs to determine which one is right.
2436 Description: Inability to activate failout mode
2437 Details : lconf script incorrectly assumed that in pythong string's numeric
2438 value is used in comparisons.
2441 Frequency : always with multiple stripes per file
2443 Description: Inefficient object allocation for mutli-stripe files
2444 Details : When selecting which OSTs to stripe files over, for files with
2445 a stripe count that divides evenly into the number of OSTs,
2446 the MDS is always picking the same starting OST for each file.
2447 Return the OST selection heuristic to the original design.
2452 Description: mount failures may take full timeout to return an error
2453 Details : Under some heavy load conditions it is possible that a
2454 failed mount can wait for the full obd_timeout interval,
2455 possibly several minutes, before reporting an error.
2456 Instead return an error as soon as the status is known.
2458 Frequency : quota enabled and large files being deleted
2460 Description: releasing more than 4GB of quota at once hangs OST
2461 Details : If a user deletes more than 4GB of files on a single OST it
2462 will cause the OST to spin in an infinite loop. Release
2463 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2468 Description: statfs data retrieved from /proc may be stale or zero
2469 Details : When reading per-device statfs data from /proc, in the
2470 {kbytes,files}_{total,free,avail} files, it may appear
2471 as zero or be out of date.
2474 Frequency : systems with MD RAID1 external journal devices
2476 Description: lconf's call to blkid is confused by RAID1 journal devices
2477 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2478 of returning all block devices that match the journal UUID.
2481 Frequency : always, for aggregate stripe size over 4GB
2483 Description: assertion fails when trying to use 4GB stripe size
2484 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2485 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2488 Frequency : always on ppc64
2490 Description: the first write on an ext3 filesystem with mballoc got stuck
2491 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2492 perform endianness conversion.
2494 ------------------------------------------------------------------------------
2496 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2498 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2499 INTEROPERATE with older versions automatically. Please read the
2500 user documentation before upgrading any part of a live system.
2501 * WARNING: Lustre networking configuration changes are required with
2502 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2505 * Support for kernels:
2506 2.6.9-22.0.2.EL (RHEL 4)
2507 2.6.5-7.244 (SLES 9)
2508 2.6.12.6 vanilla (kernel.org)
2511 Severity : enhancement
2512 Bugzilla : 7981/8208
2513 Description: Introduced Lustre Networking (LNET)
2514 Details : LNET is new networking infrastructure for Lustre, it includes
2515 a reorganized network configuration mode (see the user
2516 documentation for full details) as well as support for routing
2517 between different network fabrics. Lustre Networking Devices
2518 (LNDs) for the supported network fabrics have also been
2519 created for this new infrastructure.
2521 Severity : enhancement
2522 Description: Introduced Access control lists
2523 Details : clients can set ACLs on files and directories in order to have
2524 more fine-grained permissions than the standard Unix UGO+RWX.
2525 The MDS must be started with the "-o acl" mount option.
2527 Severity : enhancement
2528 Description: Introduced filesystem quotas
2529 Details : Administrators may now establish per-user quotas on the
2532 Severity : enhancement
2534 Description: Configuration change for the XT3
2535 The PTLLND is now used to run Lustre over Portals on the XT3
2536 The configure option(s) --with-cray-portals are no longer used.
2537 Rather --with-portals=<path-to-portals-includes> is used to
2538 enable building on the XT3. In addition to enable XT3 specific
2539 features the option --enable-cray-xt3 must be used.
2544 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2545 Details : A race between allocating a new llcd and re-getting the llcd_lock
2546 allowed another thread to grab newly-allocated llcd.
2548 Severity : enhancement
2550 Description: 2.6 OST async journal commit and locking fix to improve performance
2551 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2552 2.6 kernels are now async as they already were in 2.4 kernels so
2553 that they can commit concurrently with the network bulk transfer.
2554 For block-allocated files the filter allocation semaphore is held
2555 to avoid filesystem fragmentation during allocation. BKL lock
2556 removed for 2.6 xattr operations where it is no longer needed.
2561 Description: lconf incorrectly determined whether two IP networks could talk
2562 Details : In some more complicated routing and multiple-network
2563 configurations, lconf will avoid trying to make a network
2564 connection to a disjoint part of the IP space. It was doing the
2565 math incorrectly for one set of cases.
2570 Description: Fix for potential infinite loop processing records in an llog.
2571 Details : If an llog record is corrupted/zeroed, it is possible to loop
2572 forever in llog_process(). Validate the llog record length
2573 and skip the remainder of the block on error.
2576 Frequency : occasional (liblustre only)
2578 Description: liblustre could not open files whose last component is a symlink
2579 Details : sysio_path_walk() would incorrectly pass the open intent to
2580 intermediate path components.
2583 Frequency : rare (liblustre only with non-standard tuning)
2584 Bugzilla : 7201 (7350)
2585 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2586 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2587 after it has been released, only do this for referenced locks
2589 Severity : enhancement
2591 Description: specifying an (invalid) directory default stripe_size of -1
2592 would reset the directory default striping
2593 Details : stripe_size -1 was used internally to signal directory stripe
2594 removal, now use "all default" to signal dir stripe removal
2595 as a directory striping of "all default" is not useful
2598 Frequency : common for large clusters running liblustre clients
2600 Description: doing an ls when liblustre clients are running is slow
2601 Details : sending a glimpse AST to a liblustre client waits for every AST
2602 to time out, as liblustre clients will not respond. Since they
2603 cannot cache data we refresh the OST lock LVB from disk instead.
2605 Severity : enhancement
2607 Description: doing an ls at the same time as file IO can be slow
2608 Details : enqueue and other "small" requests can be blocked behind many
2609 large IO requests. Create a new OST IO portal for non-IO
2610 requests so they can be processed faster.
2613 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2615 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2616 created file with mode 000
2617 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2618 allow an ftruncate() as a non-root user to a file with mode 000.
2619 Lustre now respects this flag to disable mode checks when
2620 truncating a file owned by the user
2623 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2626 Description: Revoking locks from clients that went dead or catatonic might take
2628 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2629 cancellation of such locks instant on servers without waiting for
2630 any reply from clients. Clients drops these locks when cancel
2631 notification from server is received without replying.
2634 Frequency : liblustre-only, when liblustre client dies or becomes busy
2636 Description: Doing ls on Linux clients can take a long time with active
2638 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2639 granting such locks to it in the first place if possible. Locks
2640 are taken by proxy on the OST during the read or write and
2641 dropped immediately afterward. Add connect flags handling, do
2642 not grant locks to liblustre clients for glimpse ASTs.
2644 Severity : enhancement
2646 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2647 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2648 sequential reads and behaves badly with applications doing random
2649 reads. Improve it by reading ahead at least read region, and
2650 avoiding excessive large RPC for small reads.
2652 Severity : enhancement
2654 Description: Creating more than 1000 files for a single job may cause a load
2655 imbalance on the OSTs if there are also a large number of OSTs.
2656 Details : qos_prep_create() uses an OST index reseed value that is an
2657 even multiple of the number of available OSTs so that if the
2658 reseed happens in the middle of the object allocation it will
2659 still utilize the OSTs as uniformly as possible.
2664 Description: OST or MDS may oops in ping_evictor_main()
2665 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2666 but doesn't restart at beginning of obd_exports_timed list
2669 Severity : enhancement
2671 Description: improve by-nid export eviction on the MDS and OST
2672 Details : allow multiple exports with the same NID to be evicted at one
2673 time without re-searching the exports list.
2676 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2678 Description: MDS may oops in groups_free()
2679 Details : in rare race conditions a newly allocated group_info struct is
2680 freed again, and this can be NULL. The 2.4 compatibility code
2681 for groups_free() checked for a NULL pointer, but 2.6 did not.
2684 Frequency : common for liblustre clients doing little filesystem IO
2685 Bugzilla : 9352, 7313
2686 Description: server may evict liblustre clients accessing contended locks
2687 Details : if a client is granted a lock or receives a completion AST
2688 with a blocking AST already set it would not reply to the AST
2689 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2692 Frequency : lfs setstripe, only systems with more than 160 OSTs
2694 Description: unable to set striping with a starting offset beyond OST 160
2695 Details : llapi_create_file() incorrectly limited the starting stripe
2696 index to the maximum single-file stripe count.
2699 Frequency : LDAP users only
2701 Description: lconf did not handle in-kernel recovery with LDAP properly
2702 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2704 Severity : enhancement
2706 Description: bind OST threads to NUMA nodes to improve performance
2707 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2708 node and do their allocations there to localize memory access
2710 Severity : enhancement
2712 Description: llmount can determine client NID directly from Myrinet (GM)
2713 Details : the client NID code from gmnalnid was moved directly into
2714 llmount, removing the need to use this or specifying the
2715 client NID explicitly when mounting GM clients with zeroconf
2718 Frequency : if client is started with down MDS
2720 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2721 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2722 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2723 ptlrpc_import_delayed_req() because we didn't check if the
2724 request was interrupted, and we also didn't break out of the
2725 event loop if there was no timeout
2730 Description: data loss during non-page-aligned writes to a single file from
2731 both multiple nodes and multiple threads on one node at same time
2732 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2733 inconsistency led to false short-reads, that were cached and later
2734 used by ->prepare_write() to fill in partially written page,
2735 leading to data loss.
2738 Frequency : always, if lconf --abort_recovery used
2740 Description: lconf --abort_recovery fails with 'Operation not supported'
2741 Details : lconf was attempting to abort recovery on the MDT device and not
2744 Severity : enhancement
2746 Description: remove cleanup logs
2747 Details : replace lconf-generated cleanup logs with lustre internal
2748 cleanup routines. Eliminates the need for client-cleanup and
2751 Severity : enhancement
2753 Description: add support for EAs (user and system) on lustre filesystems
2754 Details : it is now possible to store extended attributes in the Lustre
2755 client filesystem, and with the user_xattr mount option it
2756 is possible to allow users to store EAs on their files also
2758 Severity : enhancement
2760 Description: Add possibility (config option) to show minimal available OST free
2762 Details : When compiled with --enable-mindf configure option, statfs(2)
2763 (and so, df) will return least minimal free space available from
2764 all OSTs as amount of free space on FS, instead of summary of
2765 free spaces of all OSTs.
2767 Severity : enhancement
2769 Description: do not expand extent locks acquired on OST-side
2770 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2771 by server: they are not cached anyway.
2774 Frequency : when mmap is used/binaries executed from Lustre
2776 Description: Unmmap pages before throwing them away from read cache.
2777 Details : llap_shrink cache now attempts to unmap pages before discarding
2778 them (if unmapping failed - do not discard). SLES9 kernel has
2779 extra checks that trigger if this unmapping is not done first.
2784 Description: lconf didn't resolve symlinks before checking to see whether a
2785 given mountpoint was already in use
2788 Frequency : when migrating failover services
2789 Bugzilla : 6395, 9514
2790 Description: When migrating a subset of services from a node (e.g. failback
2791 from a failover service node) the remaining services would
2792 time out and evict clients.
2793 Details : lconf --force (implied by --failover) sets the global obd_timeout
2794 to 5 seconds in order to quickly disconnect, but this caused
2795 other RPCs to time out too quickly. Do not change the global
2796 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2798 Severity : enhancement
2799 Frequency : if MDS is started with down OST
2800 Bugzilla : 9439,5706
2801 Description: Allow startup/shutdown of an MDS without depending on the
2802 availability of the OSTs.
2803 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2804 Add appropriate locking and lov-osc refcounts for safe
2805 cleaning. Add osc abort_inflight calls in case the
2806 synchronize never started.
2809 Frequency : occasional (Cray XT3 only)
2811 Description: root not authorized to access files in CRAY_PORTALS environment
2812 Details : The client process capabilities were not honoured on the MDS in
2813 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2814 been accessed by an authorized user then root was able to access
2815 the file on the local client also. The root user capabilities
2816 are now allowed on the MDS, as this environment has secure UID.
2819 Frequency : occasional
2821 Description: ldiskfs "too long searching" message happens too often
2822 Details : A debugging message (otherwise harmless) prints too often on
2823 the OST console. This has been reduced to only happen when
2824 there are fragmentation problems on the filesystem.
2829 Description: Division by zero in statfs when all OSCs are inactive
2830 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2831 lov_statfs divides by value returned by lov_get_stripecnt().
2835 Bugzilla : 9489, 3273
2836 Description: First write from each client to each OST was only 4kB in size,
2837 to initialize client writeback cache, which caused sub-optimal
2838 RPCs and poor layout on disk for the first writen file.
2839 Details : Clients now request an initial cache grant at (re)connect time
2840 and so that they can start streaming writes to the cache right
2841 away and always do full-sized RPCs if there is enough data.
2842 If the OST is rebooted the client also re-establishes its grant
2843 so that client cached writes will be honoured under the grant.
2848 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2849 Details : Now I/O RPCs go to different portal number and (presumably) fast
2850 lock requests (and glimses) and other RPCs get their own service
2851 threads pool that should be able to service those RPCs
2854 Severity : enhancement
2856 Description: Ability to exchange lustre version between client and servers and
2857 issue warnings at client side if client is too old. Also for
2858 liblustre clients there is ability to refuse connection of too old
2860 Details : New 'version' field is added to connect data structure that is
2861 filled with version info. That info is later checked by server and
2865 Frequency : rare, liblustre only.
2866 Bugzilla : 9296, 9581
2867 Description: Two simultaneous writes from liblustre at offset within same page
2868 might proceed at the same time overwriting eachother with stale
2870 Details : I/O lock withing llu_file_prwv was released too early, before data
2871 actually was hitting the wire. Extended lock-holding time until
2872 server acknowledges receiving data.
2875 Frequency : extremely rare. Never observed in practice.
2877 Description: avoid generating lustre_handle cookie of 0.
2878 Details : class_handle_hash() generates handle cookies by incrementing
2879 global counter, and can hit 0 occasionaly (this is unlikely, but
2880 not impossible, because initial value of cookie counter is
2881 selected randonly). Value of 0 is used as a sentinel meaning
2882 "unassigned handle" --- avoid it. Also coalesce two critical
2883 sections in this function into one.
2885 Severity : enhancement
2887 Description: allow liblustre clients to delegate truncate locking to OST
2888 Details : To avoid overhead of locking, liblustre client instructs OST to
2889 take extent lock in ost_punch() on client's behalf. New connection
2890 flag is added to handle backward compatibility.
2892 Severity : enhancement
2893 Bugzilla : 4928, 7341, 9758
2894 Description: allow number of OST service threads to be specified
2895 Details : a module parameter allows the number of OST service threads
2896 to be specified via "options ost ost_num_threads={N}" in the
2897 OSS's /etc/modules.conf or /etc/modprobe.conf.
2901 Bugzilla : 6146, 9635, 9895
2902 Description: servers crash with bad pointer in target_handle_connect()
2903 Details : In rare cases when a client is reconnecting it was possible that
2904 the connection request was the last reference for that export.
2905 We would temporarily drop the export reference and get a new
2906 one, but this may have been the last reference and the export
2907 was just destroyed. Get new reference before dropping old one.
2909 Severity : enhancement
2910 Frequency : if client is started with failover MDS
2912 Description: Allow multiple MDS hostnames in the mount command
2913 Details : Try to read the configuration from all specified MDS
2914 hostnames during a client mount in case the "primary"
2917 Severity : enhancement
2919 Description: Stop sending data to evicted clients as soon as possible.
2920 Details : Check if the client we are about to send or are sending data to
2921 was evicted already. (Check is done every second of waiting,
2922 for which l_wait_event interface was extended to allow checking
2923 of exit condition at specified intervals).
2926 Frequency : rare, normally only when NFS exporting is done from client
2928 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2930 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2933 Frequency : always, if srand() is called before liblustre initialization
2935 Description: Liblustre uses system PRNG disturbing its usage by user application
2936 Details : Introduce internal to lustre fast and high-quality PRNG for
2937 lustre usage and make liblustre and some other places in generic
2938 lustre code to use it.
2940 Severity : enhancement
2941 Bugzilla : 9477, 9557, 9870
2942 Description: Verify that the MDS configuration logs are updated when xml is
2943 Details : Check if the .xml configuration logs are newer than the config
2944 logs stored on the MDS and report an error if this is the case.
2945 Request --write-conf, or allow starting with --old_conf.
2947 Severity : enhancement
2949 Description: Handle symlinks in the path when checking if Lustre is mounted.
2950 Details : Resolve intermediate symlinks when checking if a client has
2951 mounted a filesystem to avoid duplicate client mounts.
2956 Description: lconf can hit an error exception but still return success.
2957 Details : The lconf command catches the Command error exception at the top
2958 level script context and will exit with the associated exit
2959 status, but doesn't ensure that this exit status is non-zero.
2964 Description: failure of ptlrpc thread startup can cause oops
2965 Details : Starting a ptlrpc service thread can fail if there are a large
2966 number of threads or the server memory is very fragmented.
2967 Handle this without oopsing.
2970 Frequency : always, only if liblustre and non-default acceptor port was used
2972 Description: liblustre cannot connect to servers with non-default acceptor port
2973 Details : tcpnal_set_default_params() was not called and was therefore
2974 ignoring the environment varaible TCPNAL_PORT, as well as other
2975 TCPNAL_ environment variables
2980 Description: two objects could be created on the same OST for a single file
2981 Details : If an OST is down, in some cases it was possible to create two
2982 objects on a single OST for a single file. No problems other
2983 than potential performance impact and spurious error messages.
2987 Bugzilla : 5681, 9562
2988 Description: Client may oops in ll_unhash_aliases
2989 Details : Client dcache may become inconsistent in race condition.
2990 In some cases "getcwd" can fail if the current directory is
2996 Description: Inode refcounting problems in NFS export code
2997 Details : link_raw functions used to call d_instantiate without obtaining
2998 extra inode reference first.
3002 Bugzilla : 9942, 9903
3003 Description: Referencing freed requests leading to crash, memleaks with NFS.
3004 Details : We used to require that call to ll_revalidate_it was always
3005 followed by ll_lookup_it. Also with revalidate_special() it is
3006 possible to call ll_revalidate_it() twice for the same dentry
3007 even if first occurence returned success. This fix changes semantic
3008 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3009 reference on a request referred from the intent.
3010 ll_intent_release() then releases such a request.
3013 Frequency : rare, normally benchmark loads only
3015 Description: unlinked inodes were kept in memory on the client
3016 Details : If a client is repeatedly creating and unlinking files it
3017 can accumulate a lot of stale inodes in the inode slab cache.
3018 If there is no other client load running this can cause the
3019 client node to run out of memory. Instead flush old inodes
3020 from client cache that have the same inode number as a new inode.
3023 Frequency : SLES9 2.6.5 kernel and long filenames only
3024 Bugzilla : 9969, 10379
3025 Description: utime reports stale NFS file handle
3026 Details : SLES9 uses out-of-dentry names in some cases, which confused
3027 the lustre dentry revalidation. Change it to always use the
3031 Frequency : rare, unless heavy write-truncate concurrency is continuous
3032 Bugzilla : 4180, 6984, 7171, 9963, 9331
3033 Description: OST becomes very slow and/or deadlocked during object unlink
3034 Details : filter_destroy() was holding onto the parent directory lock
3035 while truncating+unlinking objects. For very large objects this
3036 may block other threads for a long time and slow overall OST
3037 responsiveness. It may also be possible to get a lock ordering
3038 deadlock in this case, or run out of journal credits because of
3039 the combined truncate+unlink. Solution is to do object truncate
3040 first in one transaction without parent lock, and then do the
3041 final unlink in a new transaction with the parent lock. This
3042 reduces the lock hold time dramatically.
3045 Frequency : rare, 2.4 kernels only
3047 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3048 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3049 before unmounting their filesystem in order to avoid deadlock.
3050 The kernel_locked() function in 2.4 kernels only checks whether
3051 the kernel lock is held, not whether it is this process that is
3052 holding it as 2.6 kernels do.
3057 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3058 Details : The client ptlrpc code may be trying to reconnect to a down
3059 server before a previous connection attempt has timed out.
3060 Increase the reconnect interval to be longer than the connection
3061 timeout interval to avoid sending duplicate connections to
3065 Frequency : echo_client brw_test command
3067 Description: fix echo_client to work with OST preallocated code
3068 Details : OST preallocation code (5137) didn't take echo_client IO path
3069 into account: echo_client calls filter methods outside of any
3070 OST thread and, hence, there is no per-thread preallocated
3071 pages and buffers to use. Solution: hijack pga pages for IO. As
3072 a byproduct, this avoids unnecessary data copying.
3076 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3077 Description: Client can oops in mdc_commit_close() after open replay
3078 Details : It was possible for the MDS to return an open request with no
3079 transaction number in mds_finish_transno() if the client was
3080 evicted, but without actually returning an error. Clients
3081 would later try to replay that open and may trip an assertion
3082 Simplify the client close codepath, and always return an error
3083 from the MDS in case the open is not successful.
3086 Frequency : rare, 2.6 OSTs only
3088 Description: OST may deadlock under high load on fragmented files
3089 Details : If there was a heavy load and highly-fragmented OST filesystems
3090 it was possible to have all the OST threads deadlock waiting on
3091 allocation of biovecs, because the biovecs were not released
3092 until the entire RPC IO was completed. Instead, release biovecs
3093 as soon as they are complete to ensure forward IO progress.
3095 Severity : enhancement
3097 Description: Support for specifying external journal device at mount
3098 Details : If an OST or MDS device is formatted with an external journal
3099 device, this device major/minor is stored in the ext3 superblock
3100 and may not be valid for failover. Allow detecting and
3101 specifying the external journal at mount time.
3106 Description: Mounting an MDS with pending unlinked files may cause oops
3107 Details : target_finish_recovery() calls mds_postrecov() which returned
3108 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3109 considers this an error and immediately begins cleaning up the
3110 lov, just after starting the mds_lov process
3112 Severity : enhancement
3114 Description: Implement 'lfs df' to report actual free space on per-OST basis
3115 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3116 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3117 usage of objects; '-h' to report in human readable format.
3119 ------------------------------------------------------------------------------
3121 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
3128 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3129 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3130 at startup during orphan unlinking. Instead, skip buddy update
3131 before setup, as it will be regenerated after recovery is complete.
3136 Description: If an OST is inactive, its locks might reference stale inodes.
3137 Details : lov_change_cbdata() must iterate over all namespaces, even if
3138 they are inactive to clear inode references from the lock.
3140 Severity : enhancement
3141 Frequency : occasional, if non-standard max_dirty_mb used
3143 Description: Client will block write RPCs if not enough grant
3144 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3145 then the client will block writes while waiting for another RPC
3146 to complete instead of consuming its dirty limit. With change
3147 we get improved performance when max_dirty_mb is small.
3149 Severity : enhancement
3150 Bugzilla : 3389, 6253
3151 Description: Add support for supplementary groups on the MDS.
3152 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3153 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3154 lookups for user supplementary groups into a cache.
3158 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3159 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3160 for liblustre O_EXCL is enforced by MDS.
3164 Description: readdir never returns NULL in liblustre.
3165 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3166 dentry in struct dirent.
3170 Frequency : liblustre only, depends on application IO pattern
3171 Description: liblustre clients evicted if not contacting servers
3172 Details : Don't put liblustre clients into the ping_evictor list, so
3173 they will not be evicted by the pinger ever.
3175 Severity : enhancement
3177 Description: Add ability to evict clients by NID from MDS.
3178 Details : By echoing "nid:$NID" string into
3179 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3180 $NID would be instantly evicted from this MDS and from all active
3181 OSTs connected to it.
3185 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3186 Details : lookup_it_finish() used to query file size from OSTs that was not
3191 Description: service threads change working directory to that of init
3192 Details : Starting lustre service threads may pin the working directory
3193 of the parent thread, making that filesystem busy. Threads
3194 now change to the working directory of init to avoid this.
3198 Frequency : during shutdown only
3199 Description: shutdown with a failed MDS or OST can cause unmount to hang
3200 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3205 Frequency : 2.6 only
3206 Description: chmod/chown may include an extra supplementary group
3207 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3208 supplementary group and if none is specified this is used.
3211 Bugzilla : 5479 (6816)
3213 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3214 Details : If lookup is in progress on a file that is unlinked we might try
3215 to revalidate the inode and fail in revalidate after lookup is
3216 complete and ll_file_open() enqueues the open again but
3217 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3220 Frequency : always, if lconf --abort_recovery used
3222 Description: lconf --abort_recovery fails with 'Operation not supported'
3223 Details : lconf was attempting to abort recovery on the MDT device and not
3226 ------------------------------------------------------------------------------
3228 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3233 Frequency : rare (only unsupported configurations with a node running as an
3235 Bugzilla : 6514, 5137
3236 Description: Mounting a Lustre file system on a node running as an OST could
3238 Details : OSTs now preallocates memory needed to write out data at
3239 startup, instead of when needed, to avoid having to
3240 allocate memory in possibly low memory situations.
3241 Specifically, if the file system is mounted on on OST,
3242 memory pressure could force it to try to write out data,
3243 which it needed to allocate memory to do. Due to the low
3244 memory, it would be unable to do so and the node would
3245 become unresponsive.
3247 Severity : enhancement
3249 Description: Addition of lconf --service command line option
3250 Details : lconf now accepts a '--service <arg>' option, which is
3251 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3253 Severity : enhancement
3255 Description: Failover mode is now the default for OSTs.
3256 Details : By default, OSTs will now run in failover mode. To return to
3257 the old behaviour, add '--failout' to the lmc line for OSTs.
3259 Severity : enhancement
3261 Description: Health checks are now provided for MDS and OSTs
3262 Details : Additional detailed health check information on MSD and OSTs
3263 is now provided through the procfs health_check value.
3266 Frequency : occasional, depends on IO load
3268 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3269 after numerous create/delete cycles
3270 Details : The ext3 inode allocation policy would not allocate new inodes
3271 very well on the OSTs because there are no new directories
3272 being created. Instead we look for groups with free space if
3273 the parent directories are nearly full.
3278 Description: Network or server problems during mount may cause partially
3279 mounted clients instead of returning an error.
3280 Details : The config llog parsing code may overwrite the error return
3281 code during mount error handling, returning success instead
3287 Description: MDS can fail to allocate large reply buffers
3288 Details : After long uptimes the MDS can fail to allocate large reply
3289 buffers (e.g. zconf client mount config records) due to memory
3290 fragmentation or consumption by the buffer cache. Preallocate
3291 some large reply buffers so that these replies can be sent even
3292 under memory pressure.
3296 Frequency : rare (liblustre)
3297 Description: fsx running with liblustre complained that using truncate() to
3298 extend the file doesn't work. This patch corrects that issue.
3299 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3300 ATTR_SIZE and lsm use in llu_setattr_raw.
3304 Frequency : rare, only 2.6 kernels
3305 Description: Unusual file access patterns on the MDS may result in inode
3306 data being lost in very rare circumstances.
3307 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3308 code in the 2.6 kernel caused the inode and/or EA data not to
3309 be read from disk, causing single-file corruption.
3313 Frequency : rare, only 2.6 filesystems using extents
3314 Description: Heavy concurrent write and delete load may cause data corruption.
3315 Details : It was possible under high-load situations to have an extent
3316 metadata block in the block device cache from a just-unlinked
3317 file overwrite a newly-allocated data block. We now unmap any
3318 metadata buffers that alias just-allocated data blocks.
3322 Frequency : filesystems with default stripe_count larger than 77
3323 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3324 Details : lconf specifies an inode size of 4096 bytes when the default
3325 stripe_count is larger than 77. This conflicts with the default
3326 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3327 this case to avoid pinning too much memory for large EAs.
3329 ------------------------------------------------------------------------------
3331 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3336 Frequency : rare (extremely heavy IO load with hundreds of clients)
3338 Description: Client is evicted, gets IO error writing to file
3339 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3340 caused clients to be evicted for AST timeouts. The fixes in
3341 bug 5192 mean we no longer need to have such short AST timeouts
3342 so ldlm_timeout has been increased.
3345 Frequency : occasional during --force or --failover shutdown under load
3346 Bugzilla : 5949, 4834
3347 Description: Server oops/LBUG if stopped with --force or --failover under load
3348 Details : a collection of import/export refcount and cleanup ordering
3349 issues fixed for safer force cleanup
3352 Frequency : only filesystems larger than 120 OSTs
3353 Bugzilla : 5990, 6223
3354 Description: lfs getstripe would oops on a very large filesystem
3355 Details : lov_getconfig used kfree on vmalloc'd memory
3358 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3359 Bugzilla : 6242, 6243
3360 Description: reading from files that had been truncated to a non-zero size
3361 but never opened returned no data
3362 Details : ll_file_read() reads zeros from no-object files to EOF
3367 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3368 Details : There's little harm in aborting MDS/OSS recovery and letting it
3369 try again, so I removed the LASSERT and return an error instead.
3371 Severity : enhancement
3373 Description: New debugging infrastructure for tracking down data corruption
3374 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3375 (b) cover more of the client-side code path, and (c) try to narrow
3376 down where problems occurred
3380 Bugzilla : 3819, 4364, 4397, 6313
3381 Description: Racing close and eviction MDS could cause assertion in mds_close
3382 Details : It was possible to get multiple mfd references during close and
3383 client eviction, leading to one thread referencing a freed mfd.
3385 Severity: : enhancement
3386 Bugzilla : 3262, 6359
3387 Description: Attempts to reconnect to servers are now more aggressive.
3388 Details : This builds on the enhanced upcall-less recovery that was added
3389 in 1.4.2. When trying to reconnect to servers, clients will
3390 now try each server in the failover group every 10 seconds. By
3391 default, clients would previously try one server every 25 seconds.
3396 Description: After recovery, certain operations trigger a failed
3397 assertion on a client.
3398 Details : Failing over an mds, using lconf -d --failover, while a
3399 client was doing a readdir() call would cause the client to
3400 LBUG after recovery completed and the readdir() was resent.
3402 Severity : enhancement
3404 Description: Default groups are now added by lconf
3405 Details : You can now run lconf --group <servicename> without having to
3406 manually add groups with lmc.
3409 Frequency : occasional
3411 Description: Nodes with an elan id of 0 trigger a failed assertion
3414 Frequency : always when accessing e.g. tty/console device nodes
3416 Description: tty and some other devices nodes cannot be used on lustre
3417 Details : file's private_data field is used by device data and lustre
3418 values in there got lost. New field was added to struct file to
3419 store fs-specific private data.
3422 Frequency : when exporting Lustre via NFS
3424 Description: NFSD failed occasionally when looking up a path component
3425 Details : NFSD is looking up ".." which was broken in ext3 directories
3426 that had grown large enough to become hashed.
3429 Frequency : Clusters with multiple interfaces not on the same subnet
3431 Description: Nodes will repeatedly try to reconnect to an interface which it
3432 cannot reach and report an error to the log.
3433 Details : Extra peer list entries will be created by lconf with some peers
3434 unreachable. lconf now validates the peer before adding it.
3437 Frequency : Only if a default stripe is set on the filesystem root.
3439 Description: Setting a default stripe on the filesystem root prevented the
3440 filesystem from being remounted.
3441 Details : The client was sending extra request flags in the root getattr
3442 request and did not allocate a reply buffer for the dir EA.
3445 Frequency : occasional, higher if lots of files are accessed by one client
3446 Bugzilla : 6159, 6097
3447 Description: Client trips assertion regarding lsm mismatch/magic
3448 Details : While revalidating inodes the VFS looks up inodes with ifind()
3449 and in rare cases can find an inode that is being freed.
3450 The ll_test_inode() code will free the lsm during ifind()
3451 when it finds an existing inode and then the VFS later attaches
3452 this free lsm to a new inode.
3456 Bugzilla : 6422, 7030
3457 Description: MDS deadlock between mkdir and client eviction
3458 Details : Creating a new file via mkdir or mknod (starting a transaction
3459 and getting the ns lock) can deadlock with client eviction
3460 (gets ns lock and trying to finish a synchronous transaction).
3463 Frequency : occasional
3464 Description: While starting a server, the fsfilt_ext3 module could not be
3466 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3467 kernels. Previously, lconf would still use the ext3 name
3468 when trying to load modules. Now, it will correctly use
3469 ext3 on 2.4 and ldiskfs on 2.6.
3471 Severity : enhancement
3472 Description: The default stripe count has been changed to 1
3473 Details : The interpretation of the default stripe count (0, to lfs
3474 or lmc) has been changed to mean striping across a single
3475 OST, rather than all available. For general usage we have
3476 found a stripe count of 1 or 2 works best.
3478 Severity : enhancement
3479 Description: Add support for compiling against Cray portals.
3480 Details : Conditional compiling for some areas that are different
3484 Frequency : occasional
3485 Bugzilla : 6409, 6834
3486 Description: Creating files with an explicit stripe count may lead to
3487 a failed assertion on the MDS
3488 Details : If some OSTs are full or unavailable, creating files may
3489 trigger a failed assertion on the MDS. Now, Lustre will
3490 try to use other servers or return an error to the
3494 Frequency : occasional
3496 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3497 caused lock timeout during readahead (since 1.4.2).
3498 Details : Processes doing readahead might match a lock that hasn't been
3499 granted yet if there are overlapping and conflicting lock
3500 requests. The readahead process waits on ungranted lock
3501 (original lock is CBPENDING), while OST waits for that process
3502 to cancel CBPENDING read lock and eventually evicts client.
3504 Severity : enhancement
3506 Description: Initial enabling of flock support for clients
3507 Details : Implements fcntl advisory locking and file status functions.
3508 This feature is provided as an optional mount flag (default
3509 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3510 locking are implemented yet, and those that are are not guaranteed
3511 to be completely correct in production environments.
3512 mount -t lustre -o [flock|noflock] ...
3515 Frequency : occasional
3517 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3518 assertion in the ext3 JBD (journaling) layer.
3519 Details : The b_committed_data struct is protected by the big kernel lock
3520 in 2.4 kernels, serializing journal_commit_transaction() and
3521 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3522 access to this struct. In 2.6 kernels there is finer grained
3523 locking to improve SMP performance of the JBD layer.
3527 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3532 Description: merge and cleanup kernel patches.
3535 -----------------------------------------------------------------------------
3537 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3539 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3540 versions of Lustre. Please update all servers and clients to
3541 version 1.4.2 or later at the same time. You must also run
3542 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3543 to update the configuration logs.
3545 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3546 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3547 - swab llog records of type '0' so we get proper header size/idx (5861)
3548 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3549 - fix rename of one directory over another leaking an inode (5953)
3550 - avoid SetPageDirty on 2.6 (5981)
3551 - don't re-add just-being-destroyed locks to the waiting list (5653)
3552 - when creating new directories, inherit the parent's custom
3553 striping settings if present parent (3048)
3554 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3555 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3556 - don't hold client locks on temporary worklist from l_lru (5666)
3557 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3558 - automatically evict dead clients (5921)
3559 - Update file size properly in create+truncate+fstat case (6196)
3560 - Do not unhash mountpoint dentries, do not allow removal of
3562 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3563 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3564 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3565 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3566 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3567 - Correctly handle reads of files with no objects (6243)
3568 - lctl recover will also mark a device active if deactivate used (5933)
3570 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3571 - allow --write-conf on an MDS with different nettype than client (5619)
3572 - don't write config llogs to MDS for mounts not from that MDS (5617)
3573 - lconf should create multiple TCP connections from a client (5201)
3574 - init scripts are now turned off by default; run chkconfig --on
3575 lustre and chkconfig --on lustrefs to use them
3576 - upcalls are no longer needed for clients to recover to failover
3578 - add --abort-recovery option to lconf to abort recovery on device
3580 - add support for an arbitrary number of OSTs (3026)
3581 - Quota support protocol changes.
3582 - forward compatibility changes to wire structs (6007)
3583 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3584 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3585 - improved "lustre status" script
3586 - initialize blocksize for non-regular files (6062)
3587 - added --disable-server and --disable-client configure options (5782)
3588 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3589 - Vanilla 2.4.29 support
3590 - increase maximum number of obd devices to 520 (6242)
3591 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3592 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3593 SLES 9 2.6.5 kernels
3594 - limit stripes per file to 160 (the maximum EA size) (6093)
3596 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3599 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3600 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3601 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3602 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3603 - small ext3 extents cleanups and fixes (5733)
3604 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3605 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3606 - avoid needless client->OST connect, fix handle mismatch (5317)
3607 - fix DLM error path that led to out-of-sync client, long delays (5779)
3608 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3609 - fix several locking issues related to i_size (5492,5624,5654,5672)
3610 - don't move pending lock onto export if it is already evicted (5683)
3611 - fix kernel oops when creating .foo in unlinked directory (5548)
3612 - fix deadlock in obdfilter statistics vs. object create (5811)
3613 - use time_{before,after} to avoid timer jiffies wrap (5882)
3614 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3615 - Do not leak request if server was not able to process it (5154)
3616 - If mds_open unable to find parent dir, make that negative lookup(5154)
3617 - don't create new directories with extent-mapping (5909, 5936)
3619 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3620 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3621 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3622 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3623 - support for RHEL4 kernel on i686 (5773)
3624 - provide error messages when incompatible logs are encountered (5898)
3626 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3627 * version 1.4.0.10 (1.4.1 release candidate 1)
3629 - don't keep a lock reference when lock is not granted (4238)
3630 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3631 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3632 - drop import inflight refcount on signal_completed_replay error (5255)
3633 - unlock page after async write error during send (3677)
3634 - handle missing objects in filter_preprw_read properly (5265)
3635 - no transno return for symlink open, don't save no-trasno open (3440)
3636 - don't try to complete elan receive that already failed (4012)
3637 - free RPC server reply state on error (5406)
3638 - clean up thread from ptlrpc_start_thread() on error (5160)
3639 - readahead could read extra page into cache that wasn't ejected (5388)
3640 - prevent races in class_attach/setup/cleanup/detach (5260)
3641 - don't dereference de->d_inode after l_dput of de (5458)
3642 - use "int" for stripe value returned from lock_to_stripe (5544)
3643 - mballoc allocation and error-checking fixes in 2.6 (5504)
3644 - block device patches to fix I/O request sizes in 2.6 (5482)
3645 - look up hostnames for IB nals (5602)
3646 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3647 - don't start multiple acceptors for the same port (5277)
3648 - fix incorrect LASSERT in mds_getattr_name (5635)
3649 - export a proc file for general "ping" checking (5628)
3650 - fix "lfs check" to not block when the MDS is down (5628)
3652 - service request history (4965)
3653 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3654 - create an "evict_client" /proc entry on OSTs, like the MDS has
3655 - fix mount usage message, return errors per mount(8) (5168)
3656 - change grep [] to grep "[]" in tests so they work in more UMLs
3657 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3658 - remove ancient LOV_MAGIC_V0 EA support (5047)
3659 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3660 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3661 - allow readahead allocations to fail when low on memory (5383)
3662 - mmap locking landed again, after considerable improvement (2828)
3663 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3665 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3668 - send OST transaction number in read/write reply to free req (4966)
3669 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3670 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3671 - the watchdog thread now runs as interruptible (5246)
3672 - flock/lockf fixes (but it's still disabled, pending 5135)
3673 - don't use EXT3 constants in llite code (5094)
3674 - memory shortage at startup could cause assertion (5176)
3676 - reorganization of lov code
3677 - single portals codebase
3679 - add extents/mballoc support (5025)
3680 - direct I/O reads in the obdfilter (4048)
3681 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3683 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3686 - send OST transaction number in read/write reply to free req (4966)
3687 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3688 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3689 - the watchdog thread now runs as interruptible (5246)
3690 - handle missing objects in filter_preprw_read properly (5265)
3691 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3692 - drop import inflight refcount on signal_completed_replay error (5255)
3693 - unlock page after async write error during send (3677)
3694 - return original error code on reconstructed replies (3761)
3695 - no transno return for symlink open, don't save no-trasno open (3440)
3697 - add pid to ldlm debugging output (4922)
3698 - bump the watchdog timeouts -- we can't handle 30sec yet
3699 - extra debugging for orphan dentry/inode bug (5259)
3701 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3704 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3705 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3706 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3707 - fix NULL dereference in /proc/sys/portals/routes (4827)
3708 - allow failed mdc_close() operations to be interrupted (4561)
3709 - stop precreate on OST before MDS would time out on it (4778)
3710 - don't send partial-page writes before EOF from client (4410)
3711 - discard client grant for sub-page writes on large-page clients (4520)
3712 - don't free dentries not owned by NFS code, check generation (4806)
3713 - fix lsm leak if mds_create_objects() fails (4801)
3714 - limit debug_daemon file size, always print CERROR messages (4789)
3715 - use transno after validating reply (3892)
3716 - process timed out requests if import state changes (3754)
3717 - update mtime on OST during writes, return in glimpse (4829)
3718 - add mkfsoptions to LDAP (4679)
3719 - use ->max_readahead method instead of zapping global ra (5039)
3720 - don't interrupt __l_wait_event() during strace
3722 - add software watchdogs to catch hung threads quickly (4941)
3723 - make lustrefs init script start after nfs is mounted
3724 - fix CWARN/ERROR duplication (4930)
3725 - return async write errors to application if possible (2248)
3726 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3727 - print NAL number in %x format (4645)
3728 - update barely-supported suse-2.4.21-171 series (4842)
3729 - support for sles 9 %post scripts
3730 - support for building 2.6 kernel-source packages
3731 - support for sles km_* packages
3733 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3736 - ignore -ENOENT errors in osc_destroy (3639)
3737 - notify osc create thread that OSC is being cleaned up (4600)
3738 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3739 - reconstruct ost_handle() like mds_handle() (4657)
3740 - create a new thread to do import eviction to avoid deadlock (3969)
3741 - let lconf resolve symlinked-to devices (4629)
3742 - don't unlink "objects" from directory with default EA (4554)
3743 - hold socknal file ref over connect in case target is down (4394)
3744 - allow more than 32000 subdirectories in a single directory (3244)
3745 - fix blocks count for O_DIRECT writes (3751)
3746 - OST returns ENOSPC from object create when no space left (4539)
3747 - don't send truncate RPC if file size isn't changing (4410)
3748 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3749 - bind to privileged port in socknal and tcpnal (3689)
3751 - rate limit CERROR/CWARN console message to avoid overload (4519)
3752 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3753 - basic mmap support (3918)
3754 - kernel patch series update from b1_4 (4711)
3756 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3759 - avoid crash during MDS cleanup with OST shut down (2775)
3760 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3761 - don't use bad inodes on the MDS (3744)
3762 - dynamic object preallocation to improve recovery speed (4236)
3763 - don't hold spinlock over lock dumping or change debug flags (4401)
3764 - don't zero obd_dev when it is force cleaned (3651)
3765 - print grants to console if they go negative (4431)
3766 - "lctl deactivate" will stop automatic recovery attempts (3406)
3767 - look for existing locks in ldlm_handle_enqueue() (3764)
3768 - don't resolve lock handle twice in recovery avoiding race (4401)
3769 - revalidate should check working dir is a directory (4134)
3771 - don't always mark "slow" obdfilter messages as errors (4418)
3773 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3776 - don't close LustreDB during write_conf until it is done (3860)
3777 - fix typo in lconf for_each_profile (3821)
3778 - allow dumping logs from multiple threads at one time (3820)
3779 - don't allow multiple threads in OSC recovery (3812)
3780 - fix debug_size parameters (3864)
3781 - fix mds_postrecov to initialize import for llog ctxt (3121)
3782 - replace config semaphore with spinlock (3306)
3783 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3784 - don't allow enqueue to complete on a destroyed export (3822)
3785 - down write_lock before checking llog header bitmap (3825)
3786 - recover from lock replay timeout (3764)
3787 - up llog sem before sending rpc (3652)
3788 - reduce ns lock hold times when setting kms (3267)
3789 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3790 - fix NULL deref and obd_dev leak on setup error (3312)
3791 - replace some LBUG about llog ops with error handling (3841)
3792 - don't match INVALID dentries from d_lookup and spin (3784)
3793 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3794 - fix invalid assertion in ptlrpc_set_wait (3880)
3796 - add libwrap support for the TCP acceptor (3996)
3797 - add /proc/sys/portals/routes for non-root route listing (3994)
3798 - allow setting MDS UUID in .xml (2580)
3799 - print the stack of a process that LBUGs (4228)
3801 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3804 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3805 - make sure to unset replay flag from failed open requests (3440)
3806 - if default stripe count is 0, use OST count for inode size (3636)
3807 - update parent mtime/ctime on client for create/unlink (2611)
3808 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3809 - free recovery state on server during a forced cleanup (3571)
3810 - unregister_reply for resent reqs (3063)
3811 - loop back devices mounting and status check on 2.6 (3563)
3812 - fix resource-creation race that can provoke i_size == 0 (3513)
3813 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3814 - more debugging for page-accounting assertion (3746)
3815 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3816 - avoid deadlock after precreation failure (3758)
3817 - fix race and lock order deadlock in orphan handling (3450, 3750)
3818 - add validity checks when grabbing inodes from l_ast_data (3599)
3820 - add /proc/.../recovery_status to obdfilter (3428)
3821 - lightweight CDEBUG infrastructure, debug daemon (3668)
3822 - change default OSC RPC parameters to be better on small clusters
3823 - turn off OST read cache for files smaller than 32MB
3824 - install man pages and include them in rpms (3100)
3825 - add new init script for (un)mounting lustre filesystems (2593)
3826 - run chkconfig in %post for init scripts (3701)
3827 - drop scimac NAL (unmaintained)
3829 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3832 - clean kiobufs before and after use (3485)
3833 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3834 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3835 - change init script to fail more clearly if not run as root (1528)
3836 - allow clients to reconnect during replay (1742)
3837 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3838 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3839 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3840 - fix cond_resched (3554)
3841 - only evict unfinished clients after recovery (3515)
3842 - allow bulk resend, prevent data loss (3570)
3843 - dynamic ptlrpc request buffer allocation (2102)
3844 - don't allow unlinking open directory if it isn't empty (2904)
3845 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3846 - remove extraneous obd dereference causing LASSERT failure (3334)
3847 - don't use get_cycles() when creating temp. files on the mds (3156)
3848 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3849 - handle EEXIST for set-stripe, set proper directory name (3336)
3851 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3852 - fix ksocknal_fmb_callback() error messages (2918)
3854 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3857 - don't copy lvb into (possibly NULL) reply on error (2983)
3858 - don't deref dentry after dput, don't free lvb on error (2922)
3859 - use the kms to determine writeback rpc length (2947)
3860 - increment oti_logcookies when osc is inactive (2948)
3861 - update client's i_blocks count via lvb messages (2543)
3862 - handle intent open/close of special files properly (1557)
3863 - mount MDS with errors=remount-ro, like obdfilter (2009)
3864 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3865 - don't use cancelling-locks' kms values (2947)
3866 - use highest lock extent for kms, not last one (2925)
3867 - don't dereference ERR_PTR() dentry in error handling path (3107)
3868 - fix thread race in portals_debug_dumplog() (3122)
3869 - create lprocfs device entries at setup instead of at attach (1519)
3870 - common AST error handler, don't evict client on completion race (3145)
3871 - zero nameidata in detach_mnt in 2.6 (3118)
3872 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3873 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3874 - initialize RPC timeout timer earlier for 2.6 (3219)
3875 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3876 - print nal/nid for unknown nid (3258)
3877 - additional checks for oscc recovery before doing precreate (3284)
3878 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3879 - don't crash in mdc_close for bad permissions on open (3285)
3880 - zero i_rdev for non-device files (3147)
3881 - clear page->private before handing to FS, better assertion (3119)
3882 - tune the read pipeline (3236)
3883 - fix incorrect decref of invalidated dentry (2350)
3884 - provide read-ahead stats and refine rpc in flight stats (3328)
3885 - don't hold journal transaction open across create RPC (3313)
3886 - update atime on MDS at close time (3265)
3887 - close LDAP connection when recovering to avoid server load (3315)
3888 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3889 - don't leak open file on MDS after open resend (3325)
3890 - serialize filter_precreate and filter_destroy_precreated (3329)
3891 - loop device shouldn't call sync_dev() for nul device (3092)
3892 - clear page cache after eviction (2766)
3893 - resynchronize MDS->OST in background (2824)
3894 - refuse to mount the same filesystem twice on same mountpoint (3394)
3895 - allow llmount to create routes for mounting behind routers (3320)
3896 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3897 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3898 - fix rare problem with rename on htree directories (3417)
3900 - allow default OST striping configuration per directory (1414)
3901 - fix compilation for qswnal for 2.6 kernels (3125)
3902 - increase maximum number of MDS request buffers for large systems
3903 - change liblustreapi to be useful for external progs like lfsck (3098)
3904 - increase local configuration timeout for slow disks (3353)
3905 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3907 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3910 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3911 - don't overwrite extent policy data in reply if lock was blocked (2901)
3912 - drop filter export grants atomically with removal from device (2663)
3913 - del obd_self_export from work_list in class_disconnect_exports (2908)
3914 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3915 - swab reply message in mdc_close, other PPC fixes (2464)
3916 - fix destroying of named logs (2325)
3917 - overwrite old logs when running lconf --write_conf (2264)
3918 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3919 - fix race in target_handle_connect (2898)
3920 - mds_reint_create() should take same inode create lock (2926)
3921 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3922 - don't close files for self_export to avoid uninitialized obd (2936)
3923 - allow MDS with the same name as client node (2939)
3924 - hold dentry reference for closed log files for unlink (2325)
3925 - reserve space for all logs during transactions (2059)
3926 - don't evict page beyond end of stripe extent (2925)
3927 - don't oops on a deleted current working directory (2399)
3928 - handle hard links to targets without a parent properly (2517)
3929 - don't dereference NULL lock when racing during eviction (2867)
3930 - don't grow lock extents when lots of conflicting locks (2919)
3932 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3935 - account for cache space usage on clients to avoid data loss (974)
3936 - lfsck support in lustre kernel code (2349)
3937 - reduce journal credits needed for BRW writes (2370)
3938 - orphan handling to avoid losing space on client/server crashes
3939 - ptlrpcd can be blocked, stopping ALL progress (2477)
3940 - use lock value blocks to assist in proper KMS, faster stat (1021)
3941 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3942 - recovery for initial connections (2355)
3943 - fixes for mds_cleanup_orphans (1934)
3944 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3945 - block all file creations until orphan recovery completes (1901)
3946 - client remove rq_connection from request struct (2423)
3947 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3948 - recovery timer race (2670)
3949 - mdc_close recovey bug (2532)
3950 - ptlrpc cleanup bug (2710)
3951 - mds timeout on local locks (2588)
3952 - namespace lock held during RPCs (2431)
3953 - handle interrupted sync write properly (2503)
3954 - don't try to handle a message that hasn't been replied to (2699)
3955 - client assert failure during cleanup after abort recovery (2701)
3956 - leak mdc device after failed mount (2712)
3957 - ptlrpc_check_set allows timedout requests to complete (2714)
3958 - wait for inflight reqs when ptlrpcd finishes (2710)
3959 - make sure unregistered services are removed from the srv_list
3960 - reset bulk XID's when resending them (caught by 1138 test)
3961 - unregister_bulk after timeout
3962 - fix lconf error (2694)
3963 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3964 - readahead locks pages, leaves pending causing memory pressure (2673)
3965 - increase OST request buffers to 4096 on large machines (2729)
3966 - fix up permission of existing directories in simple_mkdir (2661)
3967 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3968 - don't assign transno to errored transactions (2742)
3969 - don't delete objects on OST if given a bogus objid from MDS (2751)
3970 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3971 - if rq_no_resend, then timeout request after recovery (2432)
3972 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3973 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3974 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3975 - MDS recovery shouldn't delete objects that it has given out (2730)
3976 - if enqueue arrives after completion, don't clobber LVB (2819)
3977 - don't unlock pages twice when trigger_group_io returns error (2814)
3978 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3979 - don't write pages to disk if there was an error (1450)
3980 - don't ping imports that have recovery disabled (2676)
3981 - take buffered bytes into account when balancing socknal conn (2817)
3982 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3983 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3984 - remove little-endian swabbing of llog records (1987)
3985 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3986 - retry reposting request buffers if they fail (1191)
3987 - grow extent at grant time to avoid granting a revoked lock (2809)
3988 - lock revoke doesn't evict page if covered by a second lock (2765)
3989 - disable VM readahead to avoid reading outside lock extents (2805)
3991 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3992 - updated kernel patches for hp-2.4.20 kernel (2681)
3994 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3997 - fix truncated write corruption (2366)
3998 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4000 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4001 - ensure len(uuid) < 37 in lmc (1171)
4002 - fix ia64 OOPS in llog_test (2255)
4003 - zero end of page at obdfilter for partial page writes (2648)
4004 - don't leave stale dentries around after renames (bug 2428)
4005 - fix timeouts when evicting a client with a single lock held (2642)
4006 - set deadline for the initial HELLO message to drain (2634)
4007 - print out dotted-quad IP addresses in the socknal (2302)
4009 - additional debugging for MDS client eviction problem (2443)
4010 - fix mkfsoptions support for osts (2603, 2604)
4012 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4015 - add series for the vanilla 2.6.0 kernel
4016 - add series for the vanilla 2.4.24 kernel
4017 - add series for a cray x86/64 UL kernel drop
4018 - fix xattr patches for the vanilla 2.4.19 series
4020 - generate true UUIDs in lmc (1171)
4021 - have portals stack dumping break in UML (2466)
4022 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4023 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4024 - wake callback waiting threads on client eviction (2460)
4025 - Add --ptldebug and --subsystem to lmc (1719)
4026 - update assertion to allow safe interrupt allocation
4027 - set rq_no_resend for cancel requests (2432)
4028 - recalculate ptlrpcd timeout after resend (2494)
4029 - call vfs_rmdir when removing pending directories (2368)
4030 - fix renaming a file to itself (2429)
4031 - lmc creates a default one-stripe lov (2454)
4032 - expand procfs space to handle large clusters (2326)
4033 - increase UML stack to avoid overflow
4034 - update lconf's list of debug and subsystem masks
4035 - fix lfs find --obd (2510)
4036 - /proc tunable for disabling filter read caching (2591)
4037 - stop rpm packages from altering slapd.conf (2301)
4038 - disable nagle in the socknal under 0conf (2578)
4039 - choose mds inode size based on stripe count (2572)
4040 - fix kernel-source rpm problems (2516)
4042 - add --disable-doc to avoid pdf generation (2421)
4043 - update documentation, tests, type-os, comments
4044 - avoid format warnings on ia64
4045 - remove the TOE NAL
4046 - tiny code cleanups by removing unused fields
4048 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4051 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4052 - remove the most blatant lies from BUILDING file (2371)
4053 - change default debug level to reasonable production setting
4054 - reduce client side cache size to reduce cache flush time
4055 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4056 - make TCP zerocopy and pinger support enabled by default (2476)
4057 - sync writes completed after process exits caused crashes (2319)
4058 - maintain correct mount count on the MDS (2356)
4059 - backout 1557, because 2316 wasn't really fixed
4060 - better file I/O statistics gathering in /proc
4061 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4062 - another kernel patch to fix zero-copy TCP function export
4063 - don't take duplicate lock when processing re-sent getattr (2420)
4064 - lctl uses obd_self_export instead of creating new conn (2353)
4065 - MDS/OST recovery case which requires object creation asserted (2425)
4066 - move lfs from /usr/sbin to /usr/bin in packages
4067 - fix race between mds_client_add and mds_client_free (2417)
4068 - use kmalloc instead of slabs in portals (2430)
4069 - don't create duplicate records when a failover MDS is present (2442)
4070 - remove unnecessary mount age check (2332)
4071 - don't remove directory inodes from locks prematurely (2451)
4072 - don't break if MDS service name is the same as hostname (2103)
4073 - fix races in client write RPC generation when cache full (2482)
4075 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4078 - remove now-unused request->rq_obd (278)
4079 - if an allocation fails, print out how much memory we've used (1933)
4080 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4081 - add the "configurable stack size" patch to most series files (1256)
4082 - ability to write large log records, for 100+ OST configs (2306)
4083 - fix NULL deref when filter_prep fails (2314)
4084 - fix operator precedence error in filter_sync
4085 - dynamic allocation of socknal TX descriptors (2315)
4086 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4087 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4088 - ensure that CWARN messages in Portals always get to the syslog
4089 - __init/__exit are not for prototype decls (ldlm_init/exit)
4090 - x86-64 compile warning fixes
4091 - fix gateway LMC keyword conflict (2318)
4092 - fix MDS lock inversions in getattr/reint paths (1844)
4093 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4094 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4095 - DLM race condition prevented some lock evictions (2328)
4096 - ENOMEM detection and retry on socknal sends (2230)
4097 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4098 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4099 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4100 - better error messages when a client is rejected during recovery (1505)
4101 - avoid cancelling locks which were never granted, after failure (2330)
4102 - fix i_sem/journal inversion in mds_client_add (2333)
4103 - fix truncate/getattr lock cycle deadlock (2334)
4104 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4105 - fix two rare exit paths which could leak an l_lock() ref (2321)
4106 - fencepost error in MDS/OST orphan recovery (2226)
4107 - make log record alignment 8 bytes (1988)
4108 - lstripe now fails when requested offset > ost_count (2237)
4109 - ensure that all kernel series have a complete list.h (1607)
4110 - fix crashes in special-file operations (2316)
4111 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4113 - allow configurable automake binary, for testing new versions
4114 - small update to the lfs documentation
4116 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4118 * fix negative export reference count in fsfilt_sync (2312)
4120 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4121 * release candidate 0.9.1
4123 - orphans are moved into the PENDING directory for possible recovery
4124 - replayed opens now open by fid for orphan/rename safety (1042)
4125 - last close of an orphan inode generates a transno (683)
4126 - chdir() and mount() now pin the directory entry (1020)
4127 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4128 - discard very old requests without processing them (1502)
4129 - remove some common, well-understood CERRORs (1505)
4130 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4131 - clear "grant" flags in OST replies until OST grant code lands (1644)
4132 - fix read performance by not clobbering i_blksize on client (1598)
4133 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4134 - make LRU size a /proc tunable, clears locks when reduced (707)
4135 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4136 - server locks take references on exports now (1558)
4137 - build fixes for 2.4.20-rh trees (1663)
4138 - return an error from lov_create if all OSCs are inactive (1751)
4139 - fix import levels when a reconnect happens without a timeout (1597)
4140 - exit early from mds_open if we get a lookup error (1749)
4141 - partial page read at EOF wouldn't wait for disk before sending (1642)
4142 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4143 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4144 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4145 - write/truncate lock inversion (1639)
4146 - Don't auto-load obdclass, portals modules during cleanup (1495)
4147 - fix timestamps from jumping to "now" (1763)
4148 - extra journal assertions (1648)
4149 - add an extra multiunlink test (1771)
4150 - fix read_record/write_record API (1776)
4151 - fix leak of offset_extent, possible incorrect i_size later (1772)
4152 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4153 - Debugging for the kqswnal_get_idle_tx problems (1820)
4154 - Allow recovery to be attempted multiple times (1536)
4155 - Write out MDS last_rcvd file after it is first created (1600)
4156 - Fix tx_descriptor leak in failed transmit situations (1827)
4157 - ext3 journaling fixes for assertion failure after IO error (1871)
4158 - class_export_put() on freed export after completion AST error (1896)
4159 - Fix revalidate looping in VFS (1322)
4160 - Don't access a freed export during MDS_REINT timeout (1521)
4161 - Add open-unlink recovery support on the MDS (1673,1764)
4162 - Return an error if no MDS data was read from last_rcvd (1946)
4163 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4164 - Don't setup a disk device that doesn't match exported UUID (317)
4165 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4166 - avoid committing NULL handle in force close
4167 - local.sh is now a one-stripe LOV configuration
4168 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4169 - fix ext3 htree duplicate directory entry corruption (1516)
4170 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4171 - update to new LOV EA format (2097)
4172 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4174 2003-06-15 Phil Schwan <phil@clusterfs.com>
4177 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4178 - per-import/export recovery handling (958, 931, 959)
4179 - multiple last-rcvd slots, for serving multiple FSes (949)
4180 - connections are again shared between multiple imp/exports (963, 964)
4181 - "umount -f" would hang if any requests needed to be sent (393, 978)
4182 - avoid pinning large req buffer by copying for queued messages (989)
4183 - add "uuid" to "lctl device" command to help upcalls (991)
4184 - "open" RPCs with transnos would confuse recovery counters (1037)
4185 - do proper endian conversion of all wire messages (288, 340, 891)
4186 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4187 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4188 - fix ptlrpc_connection leak in target_handle_connect (1174)
4189 - fix import refcounting bug in OST and MDS cleanup (1134)
4190 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4191 - fix very unlikely obd_types race condition (501)
4192 - remove osc_open hack for echo_client (1187)
4193 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4194 - a failure in read_inode2 leads to deadlock (1139)
4195 - cancel ack-locks as soon as transaction is committed (1072)
4196 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4197 - make sure to commitrw after any preprw to avoid deadlock (1162)
4198 - failing to execute a file in a lustre FS would lock inode (1203)
4199 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4200 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4201 - obd_uuid2tgt would walk off the end of the list (1255)
4202 - on IA64 the osc would give portals incorrect bulk size (1258)
4203 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4204 - fix lock inversion caused by new llite matching code (1282)
4205 - limit the number of dirty pages on a client to 10MB (1286)
4206 - timed out locks were not being corrected cancelled (1289)
4207 - fix O_DIRECT above 4GB on IA-32 (1292)
4208 * major user-visible changes
4209 - fail out/fail over policy now controlled by the upcall (993)
4211 - add OBD_PING to check server availability and failure (954)
4212 - lustre messages are now sent in sending host order (288, 340, 891)
4213 - add eadatalen to MDS getattr reply (340)
4214 - OST read replies may contain second buffer, with per-page status (593)
4216 2003-03-11 Phil Schwan <phil@clusterfs.com>
4219 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4220 - failed open()s could cause deadlock; fixed (867, 869)
4221 - stop cancelling OST locks when files are closed (481)
4222 - overlapping XID spaces caused network corruption (851, 853)
4223 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4224 - setattr_raw added, to do single-RPC, server-side setattrs
4225 - lmc/lconf syntax change for OST UUIDs
4226 - fix crashy race condition between ptlrpc_free_req and osc_close
4227 - don't use request in mdc_enqueue if we hit a timeout (889)
4228 - don't set the inode i_size for regular files from the MDS (896)
4229 - handle out of order completion AST (842)
4230 - don't LBUG if a lock request times out after receiving AST (913)
4231 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4232 - if a bad lock AST arrives, send an error instead of dropping entirely
4233 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4234 - fix leak in bulk IO when only partially completed (899, 900, 926)
4235 - fix O_DIRECT for ia64 (55)
4236 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4237 - C-z after timeout could hang a process forever; fixed (977)
4239 - client-side I/O cache (678, 924, 929, 941, 970)
4241 - READPAGE and SETATTRs which don't take server-side locks get
4244 2003-02-11 Phil Schwan <phil@clusterfs.com>
4247 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4248 - Fix fsfilt_extN_readpage() to read a full page of directory
4249 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4250 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4251 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4252 - Most intent operations were removed, in favour of a new RPC mode
4253 that does a single RPC to the server and bypasses most of the VFS
4254 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4255 - Aggressively cancel local locks on DLM servers
4256 - mds_reint_unlink sends EA to the client if it's the last nlink.
4257 client uses that EA to unlink OST objects.
4258 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4259 - recursive symlinks were fixed (439)
4260 - fixed NULL deref in DEBUG_REQ
4261 - filter_update_lastobjid no longer calls sync, which annoyed extN
4262 - fixed multi-client small-writes to a single file problem (445)
4263 - fixed mtime updates during file writes (607)
4264 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4265 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4266 - fixed Linux OST size reporting problem (444, 656)
4267 - OST now updates object mtime with writes or setattr (607, 619)
4268 - client verifies file size before zeroing page past EOF (445)
4269 - OST now writes last allocated objid to disk with allocation (108)
4270 - LOV on echo now works (409)
4272 - mds_reint_unlink sends a new buffer, with the EA included. this
4273 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4274 set if a regular file was being unlinked, and it was the last link
4275 - use PtlGet from the target for bulk writes (315)
4276 - OST now updates object mtime with writes or setattr (607, 619)
4277 - LDLM now has a grant-time callback to revalidate locked items, if
4279 - Many MDS operations were reorganized to combat race conditions
4281 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4282 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4284 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4287 - Fully reactivate OST imports after reconnection (512, others)
4288 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4289 - More graceful error handling for truncating on dead OST (515)
4290 - Don't error out unless we're actually accessing dead stripes (474)
4291 - Fix garbage sizes when stripes are missing (410)
4292 - LRU counters were broken, causing constant lock purge (433, 432)
4293 - garbage on read from stripes with failed OSTs (441)
4294 - mark OSCs as active before reconnecting during recovery (438)
4295 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4296 - lfind did not preserve OST order in output (443)
4297 - symlinks cause hung clients, incorrect data (439)
4298 - stop dereferencing request after dropping refcount (457)
4299 - don't LASSERT(spin_is_locked) on non-SMP (455)
4300 - fixes for many rename() bugs
4301 - fstat didn't correctly synchronize attributes (399)
4302 - server must handle lock cancellation during blocking AST prep (487)
4303 - bulk descriptors were free()d too soon (511)
4304 - fix paths in lconf, which would load incorrect modules (451, 507)
4305 - fix confusing lconf 'host not found' error message (386)
4306 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4307 - fix race condition in mdc_blocking_ast() for inode access (526)
4308 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4309 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4310 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4311 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4312 - fix lstripe in several ways (kernel side) (527)
4313 - fix request leak in ldlm_cli_enqueue (262)
4314 - incorrect OSC was marked inactive after OST failure
4315 - call mds_fs_cleanup before unmounting filesystem (524)
4316 - fix races between taking ns_lock and ldlm_lock_change_resource
4317 - fix races updating LOV export open file list
4318 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4319 - fix recovery NULL deref in ldlm_cli_cancel_unused
4320 - fix some DLM races by using new hash table for lock handles (419)
4321 - permit the client to specify desired inodes, at replay
4322 - duplicate requests when we queue them for replay reintegration
4323 - fix last_rcvd offset calculation
4324 - sync after each recovered transaction, so we always make progress
4325 - never, not always, ERESTART requests without transnos
4326 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4328 - skip replay if the MDS says that the client is already connected
4329 - don't check for a recovery-enabled export to match lctl's UUID
4330 - don't INC_USE_COUNT for phantom exports
4331 - don't crash when cleaning up phantom exports (567)
4332 - don't double-finish or set replay data for errored mdc_open requests
4333 - abort requests when they time out, so we don't get old replies
4334 - send/receive replies for AST messages again
4335 - if the client says that it doesn't have the lock, cancel it on the
4337 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4338 mark it as destroyed, it will all work out when decref is called
4339 - fix module use counts (22, 581)
4341 - ASTs now expect a reply (server cancels lock on error reply)
4343 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4346 - fix many simultaneous client startup (392)
4347 - fix dentry->d_it clobbering
4348 - credentials weren't being shipped for readdir/getattr operations
4349 - remove invalid assertions triggered during some concurrent MD
4351 - proper Lustre versions added (336, 389)
4352 - fix memory leak for create error case (398)
4353 - fix LOV locking bug that would get cli/srv out of sync
4354 - fix echo client over LOV (409)
4355 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4356 - fix double-O_EXCL intent crash (424)
4357 - avoid sending multiple lock CANCELs (352)
4359 - MDS can do multi-client recovery (modulo bugs in new code)
4361 - many updates, edits, cleanups
4363 2002-11-18 Phil Schwan <phil@clusterfs.com>
4366 - fix null d_it dereference (346)
4367 - fix full OST/dbench hang (333)
4368 - fix permission problem with file removal (286)
4369 - fix removal of OSCs from LOV when they fail
4370 - fix NULL deref during bulk timeout (214)
4371 - fix problems related to multiple filesystems on one MDS (241)
4372 - fixed serious subtle metadata locking bugs
4373 - free locks on clients when inodes are removed due to memory
4375 - fix inode pointer in lock data (285)
4376 - partial support for multiple MDS on a single host (241)
4377 - data locks weren't cancelled at clear_inode time (290, 311)
4378 - intent locks could lead to unbounded lock growth (205)
4379 - added a maximum lock count, an LRU list, and a flusher
4380 - fix multiple rename (365)
4381 - properly abstracted the echo client
4382 - OSC locked 1 byte too many; fixed
4383 - rewrote brw callback code:
4384 - fixed recovery bugs related to LOVs (306)
4385 - fixed too-many-pages-in-one-write crash (191)
4386 - fixed (again) crash in sync_io_timeout (214)
4387 - probably fixed callback-related race (385)
4389 - Add capability to MDS protocol
4390 - LDLM cancellations and callbacks on different portals
4392 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4395 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4396 - defer open object destruction to close time (601981, 138)
4397 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4398 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4399 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4400 - don't call obd_setattr in ll_file_release for destroyed objects
4401 * protocol change to lustre_msg: move |version| and add |flags|
4402 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4403 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4404 * added replay of create, unlink, link and rename operations during
4405 MDS failover; recovery should be much more robust now
4406 * remove failed OSCs from LOVs (only lov_create uses this so far)
4407 * the lustre-HOWTO was brought (more) up to date (582544)
4409 2002-10-23 Phil Schwan <phil@clusterfs.com>
4412 - in-use dentries weren't being reused properly (617851)
4413 - prevent multiple LDLM setup (599178)
4414 - fix LOV size calculations for truncate (617853)
4415 - fix client handling of MDS intent errors (POSIX)
4416 - fix permission bug in lovstripe.c test (624321)
4417 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4418 - truncate past end of file could corrupt data
4419 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4420 - a race in recovery could return ETIMEDOUT to apps (623947)
4421 - building outside the source directory was fixed
4422 * the lustre-HOWTO was brought (more) up to date (582544)
4423 * major progress was made on recovery functionality
4425 2002-10-10 Phil Schwan <phil@clusterfs.com>
4428 - recovery deadlock fix
4429 - rm -rf causes LBUG fix (617817)
4430 - file open by multiple tasks fix (618962)
4431 - directory permissions bugs (602707 and 620007)
4432 - journal_stop fixed with locking (611313)
4433 - O_APPEND failures resolved (618273, perhaps 614459)
4434 - lconf PATH fix (619770)
4435 - IA64 build fix (621450)
4436 - RPC buffer sizes scale with amount of memory
4438 2002-10-01 Phil Schwan <phil@clusterfs.com>
4441 - locks would be cancelled without throwing away data pages,
4442 resulting in inconsistent data (605627)
4443 - inode attributes were not always being refreshed (605627, 612449)
4444 - lconf now continues to cleanup after lctl reports an error
4445 - MDS now enforces user permissions (602707)
4446 - lprocfs cleanup fixed, but not yet enabled (614157)
4447 - fixed infinite server hang, should a client not respond to an AST
4448 - avoid going into recovery if user calls readlink() with a buffer
4449 that's too small (613941)
4450 - AST RPCs no longer require replies (614867) -- this may be changed
4451 - don't crash server if client sends an IOV that's too big (611336)
4452 - fixed lock conversion deadlock (611892)
4453 - fixed the following of symlinks (614622)
4454 * recovery: the server can remove locks from a client that dies, other
4455 clients can make progress
4456 * more extN patch fixes
4457 * compile-time configurable ptlrpc buffer allocations
4459 - collaborative read cache document
4460 - Lustre Lite Performance CDR document-in-progress
4462 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4465 - fix typo in patch-2.4.18
4467 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4470 - clear ptlrpc request each time in handle_incoming_request()
4471 - unlink of files now destroys the object on the OST
4473 2002-09-19 Peter Braam <braam@clusterfs.com>
4475 * add hard link support
4476 * change obdfile creation method
4477 * kernel patch changed
4479 2002-09-19 Peter Braam <braam@clusterfs.com>
4482 - stack overflow bug in extN fixed
4484 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4486 * documentation updates
4487 - add man pages for config tools
4488 - update tests/README to describe testing with new config tools
4489 - finish metadata API descriptions
4490 * bug fixes and cleanups
4491 - statfs workaround for 16TB limit
4492 - LOV stripe allocation improved, can stripe on subset of OSTs
4493 - LOV file size/IO offset was wrong for files > 4GB in size
4494 - object EA data was being dropped, caused files to be unreadable
4495 - memory overflow with non-LOV OST caused memory corruption
4496 - fixed regression tests to work with new config tools, obdfilter
4497 - fixed bug when directory size became larger than 1 block
4498 - fixed bug (for single client case) when PWD was deleted
4499 - invalidate local directory pages when doing intent-based ops
4500 - avoid LDLM oops when lock callback contained bad data
4502 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4504 * documentation updates
4505 * bug fixes and cleanups
4506 - configuration tools
4509 - 64-bit compile warnings
4510 - 64-bit internal statfs data
4512 * test_brw on persistent OST devices
4514 * lprocfs (disabled)
4516 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4518 * documentation updates
4519 * bug fixes and cleanups
4520 * configuration tools
4522 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4525 * many small fixes to 0_5_4
4526 * io/network handling
4527 * thinkos in MDS operations
4529 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4532 * crucial basic fixes to 0.5.3
4533 * IOR, Iozone work over Elan
4536 2002-08-07 Phil Schwan <phil@clusterfs.com>
4537 * version 0_5_3, our first alpha
4538 * we use the new Portals iovs
4539 * documentation updates
4540 * bug fixes and cleanups
4541 * small changes in the DLM wire protocol
4543 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4544 * version 0_5_1 with some initial stability,
4545 * locking on MD and file I/O.
4546 * documentation updates
4547 * several bug fixes since 0.5.0
4548 * small changes in wire protocol
4550 2002-07-18 Phil Schwan <phil@clusterfs.com>
4552 * delivered as Lustre Light Alpha
4553 * fixed a crash after handling invalid MDS requests
4554 * fixed directory pages for architectures with non-4k pages sizes
4556 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4557 * release version v0_4_4
4558 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4560 * Fixes a number of interruption problems with OST operations.
4561 * Update documentation for portals header changes
4562 * Move all wire protocol structs/defines to lustre_idl.h
4563 * Fixes symlink length bug.
4564 * Add tcpdump to repository.
4566 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4567 * release version v0_4_3
4568 * Fixes statfs for inodes on extN.
4569 * Fixes bug in runtests which would delete /etc/hosts.
4570 * Use 64-bit object IDs wherever possible (not into VFS though)
4571 Remove ost_get_info, which is unused by lustre, and out of date.
4573 2002-07-03 Peter Braam <braam@clusterfs.com>
4574 * release version v0_4_2 Fixes a lookup error (type not passed)
4575 * move forward to head of Portals
4576 * move forward to latest Lustre kernel
4578 2002-06-25 Peter Braam <braam@clusterfs.com>
4579 * release version v0_4_1. Hopefully stable on single node use.