1 tbd Sun Microsystems, Inc.
4 2.6.9-67.0.4.EL (RHEL 4),
5 2.6.16.54-0.2.5 (SLES 10),
6 2.6.18-53.1.14.el5 (RHEL 5).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.40.4-cfs1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
17 Frequency : rare, on recovery
18 Description: read procfs can produce deadlock in some situation
19 Details : Holding lprocfs lock which send rpc can produce block for destroy
20 obd objects and this also block reconnect with -EALREADY. This isn't
21 fix all lprocfs bugs - but make it rare.
23 Severity : enhancement
25 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
28 Frequency : frequent on X2 node
30 Description: mdc_set_open_replay_data LBUG
31 Details : Set replay data for requests that are eligible for replay.
35 Description: lustre_mgs: operation 101 on unconnected MGS
36 Details : When MGC is disconnected from MGS long enough, MGS will evict the
37 MGC, and late on MGC cannot successfully connect to MGS and a lot
38 of the error messages complaining that MGS is not connected.
41 Frequency : on start mds
43 Description: Implement get_info(last_id) in obdfilter.
46 Frequency : occasional
48 Description: Correctly check stale fid, not start epoch if ost not support SOM
49 Details : open with flag O_CREATE need set old fid in op_fid3 because op_fid2
50 overwrited with new generated fid, but mds can anwer with one of these
51 two fids and both is not stale. setattr incorectly start epoch and
52 assume will be called done_writeting, but without SOM done_writing
56 Frequency : rare, depends on device drivers and load
58 Description: MDS or OSS nodes crash due to stack overflow
59 Details : Code changes in 1.8.0 increased the stack usage of some functions.
60 In some cases, in conjunction with device drivers that use a lot
61 of stack the MDS (or possibly OSS) service threads could overflow
62 the stack. One change which was identified to consume additional
63 stack has been reworked to avoid the extra stack usage.
66 Frequency : occasional
68 Description: Do not fail import if osc_interpret_create gets -EAGAIN
69 Details : If osc_interpret_create got -EAGAIN it immediately exits and
70 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
71 oscc_has_objects and see OSC has no objests and call
72 oscc_internal_create to resend create request.
74 Severity : enhancement
76 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
78 Severity : enhancement
80 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
85 Description: Fix replay if there is an un-replied request and open
86 Details : In some cases, older replay request will revert the
87 mcd->mcd_last_xid on MDS which is used to record the client's
90 Severity : enhancement
92 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
94 Severity : enhancement
96 Description: Add rhel5 support to HEAD.
98 Serverity : enhancement
100 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
105 Description: Don't allow skipping OSTs if index has been specified.
106 Details : Don't allow skipping OSTs if index has been specified, make locking
107 in internal create lots better.
111 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
112 Details : access to bitfield in structure is always rounded to long
113 and this produce problem with not atomic change any bit.
117 Description: Lustre make rpms failed.
118 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
119 builds Lustre rpms from tarball.
122 Frequency : rare on shutdown ost
124 Description: If llog cancel was not send before clean_exports phase, this can
125 produce deadlock in llog code.
126 Details : If llog thread has last reference to obd and call class_import_put
127 this produce deadlock because llog_cleanup_commit_master wait when
128 last llog_commit_thread exited, but this never success because was
129 called from llog_commit_thread.
133 Description: allow userland application know is lost one of stripes.
134 Details : fill lvb_blocks with error code on ost and return it to
135 application if error flag found.
139 Description: NULL lov_tgts causing MDS oops
140 Details : more safe checks for NULL lov_tgts for avoid oops.
142 Severity : enhancement
144 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
148 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
149 Details : If an OST is down the MDS will hang indefinitely in
150 obd_statfs_rqset() waiting for the statfs data. While for
151 MDS QOS usage of statfs, it should not stuck in waiting.
153 Severity : enhancement
155 Description: remote_acl support
156 Details : Support ACL-based permission check for remote user.
157 Support setfacl/getfacl for remote user with the utils
158 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
159 the system "{s,g}etfacl" utils.
161 Severity : enhancement
163 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
165 Severity : enhancement
167 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
171 Description: make mgs_setparam() handle fsname containing dash
172 Details : fsname containing a dash does not work with lctl conf_param
174 Severity : enhancement
176 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
178 Severity : enhancement
180 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
184 Frequency : rare, at shutdown
185 Description: access already free / zero obd_namespace.
186 Details : if client_disconnect_export was called without force flag set,
187 and exist connect request in flight, this can produce access to
188 NULL pointer (or already free pointer) when connect_interpret
189 store ocd flags in obd_namespace.
193 Frequency : only at startup
194 Description: not alloc memory with spinlock held.
195 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
196 if any spinlock held.
198 Severity : enhancement
200 Description: make lustre randomly fail allocating memory
201 Details : Make lustre randomly failed allocating memory for testing purpose.
203 Severity : enhancement
205 Description: lost problems with lov objid file
206 Details : Fixes some scability and access to not inited memory problems
207 in work with lov objdid file.
212 Description: lfs find does not continue on file error
213 Details : Continue other files processing when a file/dir is absent.
217 Description: Inconsistent usage of lustre_pack_reply()
218 Details : Standardize the usage of lustre_pack_reply() such that it
219 always generate a CERROR on failure.
222 Frequency : occasional
224 Description: MDS hang or stay in waiting lock
225 Details : If client receive lock with CBPENDING flag ldlm need send lock
226 cancel as separate rpc, to avoid situation when cancel request
227 can't processed due all i/o threads stay in wait lock.
231 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
235 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
239 Description: Update to SLES9 kernel-2.6.5-7.287.3.
243 Description: Update to RHEL4 latest kernel.
245 Severity : enhancement
247 Description: Build SLES10 patchless client fails
248 Details : The configure was broken by run ./configure with
249 --with-linux-obj=.... argument for patchless client. When the
250 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
251 can't search header adequately. Use absolute path such as
252 -I($LINUX)/include instead.
256 Description: interrupt oig_wait produce painc on resend.
257 Details : brw_redo_request can be used for resend requests from ptlrpcd and
258 private set, and this produce situation when rq_ptlrpcd_data not
259 copyed to new allocated request and triggered LBUG on assert
260 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
261 ptlrpcd set if request is changed and can be safety changed to use
266 Description: LASSERT_{REQ,REP}SWAB macros are buggy
267 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
268 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
269 these macros and replace them with their difinitions instead.
273 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
274 Details : Update which_patch & target file for SLES10 latest kernel.
277 Description: Remove client patches from SLES 10 kernel.
278 Details : This causes SLES 10 clients to behave as patchless clients
279 even on a Lustre-patched (server) kernel.
281 Severity : enhancement
283 Description: self-adjustable client's lru lists
284 Details : use adaptive algorithm for managing client cached locks lru
285 lists according to current server load, other client's work
286 pattern, memory activities, etc. Both, server and client
287 side namespaces provide number of proc tunables for controlling
290 Severity : enhancement
292 Description: light-weight GSS support
293 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
294 and reduce performance overhead.
296 Severity : enhancement
298 Description: Linux keyring support
299 Details : Support using service of Linux keyring for Lustre GSS internal
300 context refresh/cache mechanism.
304 Description: Fix errors in lfs documentation
305 Details : Fixes man pages
309 Description: don't use GFP_* in generic Lustre code.
310 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
314 Description: obdclass is limited by single OBD_ALLOC(idarray)
315 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
319 Description: warnings with build patchless client with vanila 2.6.19 and up
320 Details : change old ctl_table style and replace ctl_table/ctl_table_header
321 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
325 Frequency : Only for SLES
326 Description: sanity_quota fail test_1
327 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
328 file, which makes TSTID[2] inunique.
332 Frequency : Only for SLES9
333 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
334 Details : Update target/ChangeLog/which_patch .
338 Frequency : Only for test_10 in sanity-quota.sh
339 Description: a bug in quota test code
340 Details : checking a test flag in a wrong place
344 Frequency : Only for test_13 in sanity-quota.sh
345 Description: a bug in quota test script
346 Details : shouldn't check overfull allocated quota
350 Description: jbd statistics
351 Details : Port older jbd statistics patch for sles10
353 Severity : enhancement
355 Description: Add printing inode info into message about error in writepage.
360 Description: reply_lock_interpret crash due to race with it and lock cancel.
361 Details : Do not replay locks that are being cancelled. Do not reference
362 locks by their address during replay, just by their handle.
366 Frequency : When flocks are used.
367 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
368 Details : Flock locks might destroy just granted lock if it could be merged
369 with another existing flock, this is done in completion handler,
370 so teach ldlm_cli_enquque_fini that this is a valid case for
376 Description: Oops in read and write path when failing to allocate lock.
377 Details : Check if lock allocation failed and return error back.
381 Description: lstripe command fails for valid OST index
382 Details : The stripe offset is compared to lov->desc.ld_tgt_count
383 instead of lov->desc.ld_active_tgt_count.
387 Description: sanity.sh failed test 103
388 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
393 Description: df doesn't work properly if diskfs blocksize != 4K
394 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
398 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
399 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
403 Description: obd_health_check_timeout too short
404 Details : set obd_health_check_timeout as 1.5x of obd_timeout
408 Description: llapi_file_create() does not allow some changes
409 Details : add llapi_file_open() that allows specifying the mode and
410 open flags, and also returns an open file handle.
414 Description: merge and cleanup kernel patches.
415 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
419 Description: Add journal checksum support.(Kernel part)
420 Details : The journal checksum feature adds two new flags i.e
421 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
422 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
423 indicates that the commit block contains the checksum for
424 the blocks described by the descriptor blocks. Now commit
425 record can be sent to disk without waiting for descriptor
426 blocks to be written to disk. This behavior is controlled
427 using JBD2_FEATURE_ASYNC_COMMIT flag.
431 Description: OSS needs mutliple precreate threads
432 Details : Add ability to start more than one create thread per OSS.
436 Description: Sanity.sh test_65k failed: setstripe should have succeeded
437 Details : execute lfs setstripe on client
441 Description: mds_obd_create error creating tmp object
442 Details : When the user sets quota on root, llog will be affected and can't
443 create files and write files.
446 Frequency : Always on ia64 patchless client, and possibly others.
448 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
449 Details : This allows the patchless client to be loaded on architectures
454 Description: RedHat Update kernel for RHEL5
455 Details : Add the kernel config file for RHEL5.
459 Description: RedHat Update kernel for RHEL5
460 Details : Modify the kernel config file more closer RHEL5.
464 Description: Build failure against Centos5 (RHEL5)
465 Details : Define PAGE_SIZE when it isn't present.
469 Description: test_7 on sanity-quota.sh can't be used on separate machines
470 Details : add facet to handle it
474 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
475 Details : with NFS, the anon dentry's parent was set to itself in
476 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
477 corresponding dentry other than use rec->ur_name.
479 Severity : enhancement
481 Description: lfs setstripe enhancement
482 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
486 Description: enable data checksumming by default
487 Details : enable checksum by default, allow --disable-checksum
488 configure option and "-o nochecksum" mount option. Checksums
489 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
493 Description: lustre support for RHEL5
494 Details : Add support for RHEL5.
498 Description: Client eviction due to failover config
499 Details : after a connection loss, the lustre client should attempt to
500 reconnect to the last active server first before trying the
501 other potential connections.
505 Description: when mds and osts use different quota unit(32bit and 64bit),
506 quota will be released repeatly.
507 Details : void sending multiple quota reqs to mds, which will keep the
508 status between the reqs.
512 Description: osts not allocated evenly to files
513 Details : change the condition to increase offset_idx
516 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
518 Description: Data corruption for OSTs that are formatted larger than 2TB
520 Details : When generating the bio request for lustre file writes the
521 sector number would overflow a temporary variable before being
522 used for the IO. The data reads correctly from Lustre (which
523 will overflow in a similar manner) but other file data or
524 filesystem metadata may be corrupted in some cases.
528 Description: Tune the kernel for good SCSI performance.
529 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
530 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
535 Description: rewrite ext2-derived code in obdclass/uuid.c
536 Details : rewrite inherited code (uuid parsing code from ext2 utils)
537 from scratch preserving functionality.
541 Description: "lfs find -obd UUID" prints directories
542 Details : "lfs find -obd UUID" will return all directory names instead
543 of just file names. It is incorrect because the directories
544 do not reside on the OSTs.
548 Description: use i_size_read and i_size_write in 2.6 port
549 Details : replace inode->i_size access with i_size_read/write()
552 Frequency : when removing large files
554 Description: scheduling issue during removal of large Lustre files
555 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
556 It causes scheduling issues when removing large files (17TB in the
560 Frequency : only with liblustre clients on XT3
562 Description: evictions taking too long
563 Details : allow llrd to evict clients directly on OSTs
566 Frequency : only on ppc
568 Description: /proc/fs/lustre/devices broken on ppc
569 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
570 We should make sure the type of 'index' is loff_t and then cast
571 explicitly as needed below. Do not assign an explicitly cast
575 Frequency : only for rhel5
577 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
578 Details : Modify the target file & which_kernel.
580 Severity : enhancement
582 Description: omit set fsid for export NFS
583 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
584 as FS_REQUIRES_DEV for avoid problems with generate fsid.
588 Frequency : Always, for kernels after 2.6.16
589 Description: Fix warning idr_remove called for id=.. which is not allocated.
590 Details : Last kernels save old s_dev before kill super and not allow
591 to restore from callback - restore it before call kill_anon_super.
595 Description: buffer overruns could theoretically occur
596 Details : llapi_semantic_traverse() modifies the "path" argument by
597 appending values to the end of the origin string, and a
598 overrun may occur. Adding buffer overrun check in liblustreapi.
602 Description: Fix error on 'ls .' at the top of the Lustre mount.
603 Details : Don't revalidate dentry if it is a root dentry.
607 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
608 Details : Modify vm-tunables-rhel4.patch.
612 Description: Kernel config for 2.6.18-vanilla.
613 Details : Modify targets/2.6-vanilla.target.in.
614 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
615 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
616 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
617 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
621 Description: improve handling recoverable errors
622 Details : if request processig with error which can be recoverable on server
623 request should be resend, otherwise page released from cache and
628 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
629 Details : Modify target file & which_patch.
630 A flaw was found in the IA32 system call emulation provided
631 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
632 value could be stored in the %RAX register, which could trigger an
633 out-of-bounds system call table access. An untrusted local user
634 could exploit this flaw to run code in the kernel
635 (ie a root privilege escalation). (CVE-2007-4573).
639 Description: O_DIRECT bypasses client statistics.
640 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
641 read_ahead_stats were updated. copied the stats section from
642 osc_send_oap_rpc() into async_internal().
646 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
650 Description: change order of libsysio includes
651 Details : '#include sysio.h' should always come before '#include xtio.h'
655 Description: handle "serious error: objid * already exists" more gracefully
656 Details : If LAST_ID value on disk is smaller than the objects existing in
657 the O/0/d* directories, it indicates disk corruption and causes an
658 LBUG(). If the object is 0-length, then we should use the existing
659 object. This will help to avoid a full fsck in most cases.
661 Severity : enhancement
663 Description: adapt the lustre_config script to support the upgrade case
664 Details : Add "-u" option for lustre_config script to support upgrading 1.4
665 server targets to 1.6 in parallel.
669 Description: To avoid grant space > avaible space when the disk is almost
670 full. Without this patch you might see the error "grant XXXX >
671 available" or some LBUG about grant, when the disk is almost
673 Details : In filter_check_grant, for non_grant cache write, we should
674 check the left space by if (*left > ungranted + bytes), instead
675 of (*left > ungranted), because only we are sure the left space
676 is enough for another "bytes", then the ungrant space should be
677 increase. In client, we should update cl_avail_grant only there
678 is OBD_MD_FLGRANT in the reply.
682 Description: Update RHEL 4 kernel to fix local root privilege escalation.
683 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
684 described in CVE-2007-4573. This problem could allow untrusted
685 local users to gain root access.
688 Frequency : when using O_DIRECT and quotas
690 Description: Incorrect file ownership on O_DIRECT output files
691 Details : block usage reported by 'lfs quota' does not take into account
692 files that have been written with O_DIRECT.
697 Description: touch file failed when fs is not full
698 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
699 otherwise we can get ENOSP while fs is not full.
703 Description: parallel lock callbacks
704 Details : Instead of sending blocking and completion callbacks as separated
705 requests, adding them to a set and sending in parallel.
708 Frequency : only for Cray XT3
709 Bugzilla : 12829/13455
710 Description: Changing primary group doesn't change the group lustre assigns to
712 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
713 overridden with the primary group provided by the group upcall,
714 whereas the supplied fsgid can be trusted if it is in the list of
715 supplementary groups returned by the group upcall.
717 Severity : enhancement
719 Description: Allow masking D_WARNING, D_ERROR messages from console
720 Details : Console messages can now be disabled via lnet.printk.
724 Description: User code with malformed file open parameter crashes client node
725 Details : Before packing join_file req, all the related reference should be
726 checked carefully in case some malformed flags cause fake
727 join_file req on client.
731 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
732 Details : ldlm_completion_ast() assumes that a lock is granted when the req
733 mode is equal to the granted mode. However, it should also check
734 that LDLM_FL_CP_REQD is not set.
738 Description: Heavy nfs access might result in deadlocks
739 Details : After ELC code landed, it is now improper to enqueue any mds
740 locks under och_sem, because enqueue might want to decide to
741 cancel open locks for same inode we are holding och_sem for.
745 Description: Client eviction while running blogbench
746 Details : A lot of unlink operations with concurrent I/O can lead to a
747 deadlock causing evictions. To address the problem, the number of
748 oustanding OST_DESTROY requests is now throttled to
749 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
750 callbacks are processed in priority.
754 Description: enable ACLs on MDS by default
755 Details : ACLs must be enabled on MDS by default.
758 Frequency : PPC/PPC64 only
760 Description: conflicts between asm-ppc64/types.h and lustre_types.h
761 Details : fix duplicated definitions between asm-ppc64/types.h and
762 lustre_types.h on PPC.
765 Frequency : PPC/PPC64 only
767 Description: asm-ppc/segment.h does not exist
768 Details : fix compile issue on PPC.
772 Description: better handle error messages in extents code
775 Frequency : RHEL4 only
777 Description: mkfs is very slow on IA64/RHEL4
778 Details : A performance regression has been discovered in the MPT Fusion
779 driver between versions 3.02.73rh and 3.02.99.00rh. As a
780 consequence, we have downgraded the MPT Fusion driver in the RHEL4
781 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
783 Severity : enhancement
785 Description: SNMP support enhancement
786 Details : Adding total number of sampled request for an MDS node in snmp
789 Severity : enhancement
791 Description: Optimize ldlm waiting list processing for PR extent locks
792 Details : When processing waiting list for read extent lock and meeting read
793 lock that is same or wider to it that is not contended, skip
794 processing rest of the list and immediatelly return current
795 status of conflictness, since we are guaranteed there are no
796 conflicting locks in the rest of the list.
800 Description: Time out and refuse to reconnect
801 Details : When the failover node is the primary node, it is possible
802 to have two identical connections in imp_conn_list. We must
803 compare not conn's pointers but NIDs, otherwise we can defeat
804 connection throttling.
808 Description: port llog fixes from b1_6 into HEAD
809 Details : Port llog reference couting and some llog cleanups from b1_6
810 (bug 10800) into HEAD, for protect from panic and access to already
811 free llog structures.
815 Description: Detect stride IO mode in read-ahead
816 Details : When a client does stride read, read-ahead should detect that and
817 read-ahead pages according to the detected stride pattern.
821 Description: data checksumming impacts single node performance
822 Details : add support for several checksum algorithm. Currently, only CRC32
823 and Adler-32 are supported. The checksum type can be changed on
824 the fly via /proc/fs/lustre/osc/*/checksum_type.
828 Description: use adler32 for page checksums
829 Details : when available, use the Adler-32 algorithm instead of CRC32 for
832 --------------------------------------------------------------------------------
834 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
836 * Support for kernels:
837 2.6.5-7.283 (SLES 9),
838 2.6.9-55.EL (RHEL 4),
839 2.6.16.46-0.14 (SLES 10),
840 2.6.18.8 vanilla (kernel.org)
841 * Client support for unpatched kernels:
842 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
843 2.6.16 - 2.6.22 vanilla (kernel.org)
844 * Due to recently discovered recovery problems, we do not recommend
845 using patchless RHEL 4 clients with this or any earlier release.
846 * Recommended e2fsprogs version: 1.39.cfs8
847 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
848 * Starting with this release, the ldiskfs backing filesystem required
849 by Lustre is now in its own package, lustre-ldiskfs. This package
850 should be installed. It is versioned separately from Lustre and
851 may be released separately in future.
855 Description: block reactivating mgc import until all deactivates complete
856 Details : Fix race when failing back MDT/MGS to itself (testing)
859 Frequency : at statup only
861 Description: mds_lov_synchronize race leads to various problems
862 Details : simultaneous MDT->OST connections at startup can cause the
863 sync to abort, leaving the OSC in a bad state.
865 Severity : enhancement
867 Description: add optional extra BUILD_VERSION info
868 Details : add a new environment variable (namely LUSTRE_VERS) which allows
869 to override the lustre version.
872 Frequency : 2.6.18 servers only
874 Description: ll_kern_mount() doesn't release the module reference
875 Details : The ldiskfs module reference count never drops down to 0
876 because ll_kern_mount() doesn't release the module reference.
881 Description: server LBUG when using old ost_num_threads parameter
882 Details : Accept the old ost_num_threads parameter but warn that it
883 is deprecated, and fix an off-by-one error that caused an LBUG.
888 Description: Transient SCSI error results in persistent IO issue
889 Details : iobuf->dr_error is not reinitialized to 0 between two
893 Frequency : sometimes when underlying device returns I/O errors
895 Description: OSTs not going read-only during write failures
896 Details : OSTs are not remounted read-only when the journal commit threads
897 get I/O errors because fsfilt_ext3 calls journal_start/stop()
898 instead of the ext3 wrappers.
902 Description: poor connect scaling with increasing client count
903 Details : Don't run filter_grant_sanity_check for more than 100 exports
904 to improve scaling for large numbers of clients.
907 Frequency : SLES10 only
909 Description: sanity-quota.sh quotacheck failed: rc = -22
910 Details : Quotas cannot be enabled on SLES10.
913 Frequency : liblustre clients only
915 Description: getdirentries does not give error when run on compute nodes
916 Details : getdirentries does not fail when the size specified as an argument
917 is too small to contain at least one entry
919 Severity : enhancement
921 Description: Add LNET router traceability for debug purposes
922 Details : If a checksum failure occurs with a router as part of the
923 IO path, the NID of the last router that forwarded the bulk data
924 is printed so it can be identified.
929 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
930 Details : Due to a race condition, liblustre clients were occasionally
933 Severity : enhancement
935 Description: lfs setstripe use optional parameters instead of postional
938 Severity : enhancement
940 Description: Nanosecond timestamp support for ldiskfs
941 Details : The on-disk ldiskfs filesystem has added support for nanosecond
942 resolution timestamps. There is not yet support for this at
943 the Lustre filesystem level.
946 Frequency : during server recovery
948 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
949 Details : request with rq_no_resend flag not awake l_wait_event if they get a
953 Frequency : nfs export on patchless client
955 Description: connectathon hang when test nfs export over patchless client
956 Details : Disconnected dentry cannot be found with lookup, so we do not need
957 to unhash it or make it invalid
960 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
961 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
962 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
967 Description: open req refcounting wrong on reconnect
968 Details : If reconnect happened between getting open reply from server and
969 call to mdc_set_replay_data in ll_file_open, we will schedule
970 replay for unreferenced request that we are about to free.
971 Subsequent close will crash in variety of ways.
972 Check that request is still eligible for replay in
973 mdc_set_replay_data().
978 Description: disable writes to filesystem when reading health_check file
979 Details : the default for reading the health_check proc file has changed
980 to NOT do a journal transaction and write to disk, because this
981 can cause reads of the /proc file to hang and block HA state
982 checking on a healthy but otherwise heavily loaded system. It
983 is possible to return to the previous behaviour during configure
984 with --enable-health-write.
986 Severity : enhancement
988 Description: 64-bit inode version
989 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
990 to the inode. This will be required for version-based recovery.
995 Description: MDS fails to start if a duplicate client export is detected
996 Details : in some rare cases it was possible for a client to connect to
997 an MDS multiple times. Upon recovery the MDS would detect this
998 and fail during startup. Handle this more gracefully.
1000 Severity : enhancement
1002 Description: Add -o localflock option to simulate old noflock
1004 Details : This will achieve local-only flock/fcntl locks
1010 Description: log_commit_thread vs filter_destroy race leads to crash
1011 Details : Take import reference before releasing llog record semaphore
1016 Description: Wrong request locking in request set processing
1017 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1018 imp_delayed_list, in this place should be used imp_lock.
1021 Frequency : when reconnection
1023 Description: Grant Leak when osc reconnect to OST
1024 Details : When osc reconnect ost, OST(filter) should check whether it
1025 should grant more space to client by comparing fed_grant and
1026 cl_avail_grant, and return the granted space to client instead
1027 of "new granted" space, because client will call osc_init_grant
1028 to update the client grant space info.
1031 Frequency : when client reconnect to OST
1033 Description: Grant Leak when osc do resend and replay bulk write
1034 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1035 bulk write request, because the grant info while be sync between
1036 OSC and OST when reconnect, and we should ignore the grant info
1037 these of resend/replay write req.
1042 Description: Grant space more than avaiable left space sometimes.
1043 Details : When then OST is about to be full, if two bulk writing from
1044 different clients came to OST. Accord the avaliable space of the
1045 OST, the first req should be permitted, and the second one
1046 should be denied by ENOSPC. But if the seconde arrived before
1047 the first one is commited. The OST might wrongly permit second
1048 writing, which will cause grant space > avaiable space.
1051 Frequency : when client is evicted
1053 Description: Grant might be wrongly erased when osc is evicted by OST
1054 Details : when the import is evicted by server, it will fork another
1055 thread ptlrpc_invalidate_import_thread to invalidate the
1056 import, where the grant will be set to 0. While the original
1057 thread will update the grant it got when connecting. So if
1058 the former happened latter, the grant will be wrongly errased
1059 because of this race.
1064 Description: Checking Stale with correct fid
1065 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1066 to check whether it is stale, because sometimes, we want the
1067 enqueue happened anyway, and op_data.fid2 will not be initialized.
1069 Severity : enhancement
1071 Description: update patchless client
1072 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1075 Frequency : only with 2.4 kernel
1077 Description: random memory corruption
1078 Details : size of struct ll_inode_info is to big for union inode.u and this
1079 can be cause of random memory corruption.
1084 Description: Memory leak in recovery
1085 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1086 It should also check obd_fail before fsfilt_start, otherwise if
1087 fsfilt_start return -EROFS,(failover mds during mds recovery).
1088 then the req will return with repmsg->transno = 0 and rc = EROFS.
1089 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1090 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1091 be freed no matter whether fsfilt_commit success or not.
1094 Frequency : only with huge count clients
1096 Description: Prevents from taking the superblock lock in llap_from_page for
1098 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1099 for llap_from_page call in ll_removepage prevents from taking the
1100 superblock lock for a soon died page.
1105 Description: Not check open intent error before release open handle
1106 Details : in some rare cases, the open intent error is not checked before
1107 release open handle, which may cause
1108 ASSERTION(open_req->rq_transno != 0), because it tries to release
1109 the failed open handle.
1114 Description: Set cat log bitmap only after create log success.
1115 Details : in some rare cases, the cat log bitmap is set too early. and it
1116 should be set only after create log success.
1120 Description: Accessing a block bevice can re-enable I/O when Lustre is
1121 tearing down a device.
1122 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1127 Description: service threads may hog cpus when there are a lot of requests
1128 Details : Insert cond_resched to give other threads a chance to use some CPU
1133 Description: the cat log was not initialized in recovery
1134 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1135 unlink log on mds will not be initialized until mds post
1136 recovery. And also in mds post recovery, the unlink log will
1137 initialization will be done asynchronausly, so there will be race
1138 between add unlink log and unlink log initialization.
1142 Description: brw_stats were being printed incorrectly
1143 Details : brw_stats were being printed as log2 but all of them were not
1144 recorded as log2. Also remove some code duplication arising from
1145 filter_tally_{read,write}.
1149 Frequency : rare, only in recovery.
1150 Description: ASSERTION(req->rq_type != LI_POISON) failed
1151 Details : imp_lock should be held while iterating over imp_sending_list for
1152 prevent destroy request after get timeout in ptlrpc_queue_wait.
1156 Description: replay-single.sh test 52 fails
1157 Details : A lock's skiplist need to be cleanup when it being unlinked
1158 from its resource list.
1162 Description: Short directio read returns full requested size rather than
1164 Details : Direct I/O operations should return actual amount of bytes
1165 transferred rather than requested size.
1167 Severity : enhancement
1169 Description: metadata RPC reduction (e.g. for rm performance)
1170 Details : decrease the amount of synchronous RPC between clients and servers
1171 by canceling conflicing lock before the operation on the client
1172 and packing thier handles into the main operation RPC to server.
1174 Severity : enhancement
1176 Description: Async OSC create to avoid the blocking unnecessarily.
1177 Details : If a OST has no remain object, system will block on the creating
1178 when need to create a new object on this OST. Now, ways use
1179 pre-created objects when available, instead of blocking on an
1180 empty osc while others are not empty. If we must block, we block
1181 for the shortest possible period of time.
1183 Severity : enhancement
1185 Description: refine locking for avoid write wrong info into lov_objid file
1186 Details : fix possible races with add new target and write/update data in
1189 --------------------------------------------------------------------------------
1191 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1196 Frequency : on some architectures
1198 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1199 Details : Uninitialized flags sometimes cause configuration commands to
1203 Frequency : patchless clients only
1205 Description: missing __iget() symbol export
1206 Details : The __iget() symbol export is missing. To avoid the need for
1207 this on patchless clients the deathrow inode reaper is turned
1208 off, and we depend on the VM to clean up old inodes. This
1209 dependency was during via the fix for bug 12181.
1211 --------------------------------------------------------------------------------
1213 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1215 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1216 INTEROPERATE with older versions automatically. In many cases a
1217 special upgrade step is needed. Please read the
1218 user documentation before upgrading any part of a live system.
1219 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
1220 version will not interoperate with 1.6 betas before beta5 (1.5.95).
1221 * WARNING: Lustre configuration and startup changes are required with
1222 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1228 Severity : enhancement
1230 Description: MountConf
1231 Details : Lustre configuration is now managed via mkfs and mount
1232 commands instead of lmc and lconf. New obd types (MGS, MGC)
1233 are added for dynamic configuration management. See
1234 https://mail.clusterfs.com/wikis/lustre/MountConf for
1237 Severity : enhancement
1239 Description: dynamic OST addition
1240 Details : OSTs can now be added to a live filesystem
1242 Severity : enhancement
1244 Description: startup order invariance
1245 Details : MDTs and OSTs can be started in any order. Clients only
1246 require the MDT to complete startup.
1248 Severity : enhancement
1250 Description: parallel, asynchronous orphan cleanup
1251 Details : orphan cleanup is now performed in separate threads for each
1252 OST, allowing parallel non-blocking operation.
1254 Severity : enhancement
1256 Description: optimized stripe assignment
1257 Details : stripe assignments are now made based on ost space available,
1258 ost previous usage, and OSS previous usage, in order to try
1259 to optimize storage space and networking resources.
1261 Severity : enhancement
1263 Description: Permanently set tunables
1264 Details : All writable /proc/fs/lustre tunables can now be permanently
1265 set on a per-server basis, at mkfs time or on a live system.
1267 Severity : enhancement
1269 Description: Lustre message v2
1270 Details : Add lustre message format v2.
1272 Severity : enhancement
1274 Description: client OST exclusion list
1275 Details : Clients can be started with a list of OSTs that should be
1276 declared "inactive" for known non-responsive OSTs.
1280 Description: SPEC SFS validation failure on NFS v2 over lustre.
1281 Details : Changes the blocksize for regular files to be 2x RPC size,
1282 and not depend on stripe size.
1284 Severity : enhancement
1286 Description: Multiple MD RPCs in flight.
1287 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1288 To avoid overly-overloading MDS, introduce a limit on number of
1289 MDS RPCs in flight for a single client and add /proc controls
1290 to adjust this limit.
1292 Severity : enhancement
1294 Description: client read/write statistics
1295 Details : Add client read/write call usage stats for performance
1296 analysis of user processes.
1297 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1298 file access. extents_stats shows chunk size distribution.
1299 extents_stats_per_process show chunk size distribution per
1302 Severity : enhancement
1304 Description: mds statistics
1305 Details : Add detailed mds operations statistics in
1306 /proc/fs/lustre/mds/*/stats.
1310 Description: Failure of copying files with lustre special EAs.
1311 Details : Client side always return success for setxattr call for lustre
1312 special xattr (currently only "trusted.lov").
1317 Description: Refcount LNET uuids
1318 Details : The global LNET uuid list grew linearly with every startup;
1319 refcount repeated list entries instead of always adding to
1322 Severity : enhancement
1324 Description: Dynamic service threads
1325 Details : Within a small range, start extra service threads
1326 automatically when the request queue builds up.
1329 Frequency : mixed-endian client/server environments
1331 Description: mixed-endian crashes
1332 Details : The new msg_v2 system had some failures in mixed-endian
1335 Severity : enhancement
1337 Description: Easy OST removal
1338 Details : OSTs can be permanently deactivated with e.g. 'lctl
1339 conf_param lustre-OST0001.osc.active=0'
1341 Severity : enhancement
1343 Description: MGS proc entries
1344 Details : Added basic proc entries for the MGS showing what filesystems
1347 Severity : enhancement
1349 Description: provide MGS failover
1350 Details : Added config lock reacquisition after MGS server failover.
1352 Severity : enhancement
1354 Description: add Linux 2.4 support
1355 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1359 Description: a large application tries to do I/O to the same resource and dies
1360 in the middle of it.
1361 Details : Check the req->rq_arrival time after the call to
1362 ost_brw_lock_get(), but before we do anything about
1363 processing it & sending the BULK transfer request. This
1364 should help move old stale pending locks off the queue as
1365 quickly as obd_timeout.
1368 Frequency : when an incorrect nid is specified during startup
1370 Description: ptlrpc connect to non-existant node causes kernel crash
1371 Details : LNET can't be re-entered from an event callback, which
1372 happened when we expire a message after the export has been
1373 cleaned up. Instead, hand the zombie cleanup off to another
1376 Severity : enhancement
1378 Description: plain/inodebits lock performance improvement
1379 Details : Grouping plain/inodebits in granted list by their request modes
1380 and bits policy, thus improving the performance of search through
1384 Frequency : only if OST filesystem is corrupted
1386 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1387 Details : for a short time RPCs with bulk IO are in the replay list,
1388 but replay of bulk IOs is unimplemented. If the OST filesystem
1389 is corrupted due to disk cache incoherency and then replay is
1390 started it is possible to trip an assertion. Avoid putting
1391 committed RPCs into the replay list at all to avoid this issue.
1394 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1396 Description: System hang on startup
1397 Details : This bug allowed the liblustre (e.g. catamount) client to
1398 return to the app before handling all startup RPCs. This
1399 could leave the node unresponsive to lustre network traffic
1400 and manifested as a server ptllnd timeout.
1402 Severity : enhancement
1404 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1405 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1406 boolean to control whether to print peer debug info when a
1407 client's RPC times out.
1410 Frequency : only for kernels with patches from Lustre below 1.4.3
1412 Description: Remove old rdonly API
1413 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1416 Frequency : only for devices with external journals
1418 Description: Set external device read-only also
1419 Details : During a commanded failover stop, we set the disk device
1420 read-only while the server shuts down. We now also set any
1421 external journal device read-only at the same time.
1424 Frequency : when upgrading from 1.4 while trying to change parameters
1426 Description: The wrong (new) MDC name was used when setting parameters for
1427 upgraded MDT's. Also allows changing of OSC (and MDC)
1428 parameters if --writeconf is specified at tunefs upgrade time.
1431 Frequency : when setting specific ost indicies
1433 Description: QOS code breaks on skipped indicies
1434 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1435 created with --index need not be sequential.
1437 Severity : enhancement
1439 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1440 Details : The uninit_groups feature works in conjunction with the kernel
1441 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1442 up the pass1 processing of e2fsck. This is a read-only feature
1443 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1444 mount filesystems that have had this feature enabled.
1446 Severity : enhancement
1448 Description: Improve multi-block allocation algorithm to avoid fragmentation
1449 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1450 allocation locality and avoid filesystem fragmentation.
1452 ------------------------------------------------------------------------------
1454 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1456 * Support for kernels:
1457 2.4.21-47.0.1.EL (RHEL 3)
1458 2.6.5-7.283 (SLES 9)
1459 2.6.9-42.0.10.EL (RHEL 4)
1460 2.6.12.6 vanilla (kernel.org)
1461 2.6.16.27-0.9 (SLES 10)
1462 * Recommended e2fsprogs version: 1.39.cfs5
1464 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1468 Frequency : occasional, depends on client load and configuration
1469 Bugzilla : 12181, 12203
1470 Description: data loss for recently-modified files
1472 Details : In some cases it is possible that recently written or created
1473 files may not be written to disk in a timely manner (this should
1474 normally be within 30s unless client IO load is very high).
1475 The problem appears as zero-length files or files that are a
1476 multiple of 1MB in size after a client crash or client eviction
1477 that are missing data at the end of the file.
1479 This problem is more likely to be hit on clients where files are
1480 repeatedly created and unlinked in the same directory, clients
1481 have a large amount of RAM, have many CPUs, the filesystem has
1482 many OSTs, the clients are rebooted frequently, and/or the files
1483 are not accessed by other nodes after being written.
1485 The presence of the problem can be detected by looking at
1486 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1487 smaller than the second (nr_unused) then dirty files will not
1488 be flushed automatically to disk. "sync; sleep 10" should be
1489 run several times on the node before unmounting it to update
1490 Lustre (this is also safe to run on nodes without this problem).
1492 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1493 that can cause this same problem, so customers using that kernel
1494 also need to update the kernel in addition to Lustre. In order
1495 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1497 It is normal that files written just before a client crash (less
1498 than 30s) may not yet have been flushed to disk, even for local
1502 Frequency : frequent on thin XT3 nodes
1504 Description: UUID collision on thin XT3 Linux nodes
1505 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1506 randomly, since we relied on an insufficiently-seeded PRNG.
1511 Description: OSS hangs after "All ost request buffers busy"
1512 Details : A deadlock between quota and journal operations caused OSS
1513 hangs after printing "All ost request buffers busy."
1516 Frequency : always on liblustre builds
1518 Description: Cleanup compiler warnings on liblustre
1521 Frequency : always on liblustre builds on XT3
1523 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1527 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1528 > 32000 subdirectories
1529 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1530 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1531 subdirectory count crosses 32000. This will aid e2fsck to
1532 correctly handle more than 32000 subdirectories.
1535 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1537 Description: System hang on startup
1538 Details : This bug allowed the liblustre (e.g. catamount) client to
1539 return to the app before handling all startup RPCs. This
1540 could leave the node unresponsive to lustre network traffic
1541 and manifested as a server ptllnd timeout.
1543 Severity : enhancement
1545 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1546 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1547 boolean to control whether to print peer debug info when a
1548 client's RPC times out.
1553 Description: make O_SYNC working on 2.6 kernels
1554 Details : 2.6 kernels use different method for mark pages for write,
1555 so need add a code to lustre for O_SYNC work.
1560 Description: Failure to close file and release space on NFS
1561 Details : Put inode details into lock acquired in ll_intent_file_open.
1562 Use mdc_intent_lock in ll_intent_open to properly
1563 detect all kind of errors unhandled by mdc_enqueue.
1568 Description: proc file read during shutdown sometimes raced obd removal,
1570 Details : Add lock to prevent obd access after proc file removal.
1573 Frequency : Only for files larger than 4GB on 32-bit clients.
1575 Description: improperly doing page alignment of locks
1576 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1577 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1582 Description: RPCs being resent when they shouldn't be
1583 Details : Some RPCs that should not be resent are being resent. This
1584 can cause inconsistencies in the RPC state machine. Do not
1585 resend such requests.
1588 Frequency : rare, only with NFS export
1590 Description: Crash on NFS re-export node
1591 Details : under very unusual load conditions an assertion is hit in
1592 ll_intent_file_open()
1595 Frequency : only if OST filesystem is corrupted
1597 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1598 Details : for a short time RPCs with bulk IO are in the replay list,
1599 but replay of bulk IOs is unimplemented. If the OST filesystem
1600 is corrupted due to disk cache incoherency and then replay is
1601 started it is possible to trip an assertion. Avoid putting
1602 committed RPCs into the replay list at all to avoid this issue.
1607 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1608 Details : Large single O_DIRECT read and write calls can fail to allocate
1609 a sufficiently large buffer to process the request. In case of
1610 allocation failure the allocation is retried with a smaller
1611 buffer and broken into smaller requests.
1613 Severity : enhancement
1615 Description: Add -o localflock option to simulate old noflock behaviour.
1616 Details : This will achieve local-only flock/fcntl locks coherentness.
1621 Description: versioning check is incomplete
1622 Details : Checking the version difference of client vs. server, report
1623 error if the gap is too big.
1628 Description: add support PG_writeback bit
1629 Details : add support for PG_writeback bit for Lustre, for more carefull
1630 work with page cache in 2.6 kernel. This also fix some deadlocks
1631 and remove hack for work O_SYNC with 2.6 kernel.
1633 Severity : enhancement
1635 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1636 Details : The uninit_groups feature works in conjunction with the kernel
1637 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1638 up the pass1 processing of e2fsck. This is a read-only feature
1639 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1640 mount filesystems that have had this feature enabled.
1642 Severity : enhancement
1644 Description: Improve multi-block allocation algorithm to avoid fragmentation
1645 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1646 allocation locality and avoid filesystem fragmentation.
1648 ------------------------------------------------------------------------------
1650 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1652 * Support for kernels:
1653 2.6.9-42.0.3.EL (RHEL 4)
1654 2.6.5-7.276 (SLES 9)
1655 2.4.21-47.0.1.EL (RHEL 3)
1656 2.6.12.6 vanilla (kernel.org)
1657 2.6.16.21-0.8 (SLES10)
1658 * Recommended e2fsprogs version: 1.39.cfs2-0
1660 * The backwards-compatible /proc/sys/portals symlink has been removed
1661 in this release. Before upgrading, please ensure that you change
1662 any configuration scripts or /etc/sysctl.conf files that access
1663 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1664 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1665 in advance of the upgrade on any system running Lustre 1.4.6 or
1666 newer, since /proc/sys/lnet was added in that version.
1667 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1671 Frequency : only when quota is used
1673 Description: avoid scanning export list for quota master
1674 Details : Change the algorithms to avoid scanning export list in order
1675 to improve the efficiency.
1678 Frequency : MDS failover only, very rarely
1680 Description: "went back in time" messages on mds failover
1681 Details : The greatest transno may be lost when the current operation
1682 finishes with an error (transno==0) and the client's last_rcvd
1683 record is over-written. Save the greatest transno in the
1684 mds_last_transno for this case.
1687 Frequency : always for specific kernels and striping counts
1689 Description: client may get "Matching packet too big" without ACL support
1690 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1691 when trying to access files in certain configurations. The
1692 clients should in fact be denied when mounting because they do
1693 not understand ACLs.
1696 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1698 Description: many clients connecting with IO in progress causes connect timeouts
1699 Details : Avoid synchronous journal commits to avoid delays caused by many
1700 clients connecting/disconnecting when bulk IO is in progress.
1701 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1702 OST_IO_PORTAL to avoid delays behind potentially many pending
1706 Frequency : occasionally with multiple writers to a single file
1708 Description: shared writes to file may result in wrong size reported by stat()
1709 Details : Allow growing of kms when extent lock is cancelled
1712 Frequency : always with random mmap IO to multi-striped file
1714 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1715 Details : Only if the hole is at the end of OST object so that kms is too
1716 small. Fix is to increase kms accordingly in ll_nopage.
1719 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1721 Description: writes to a missing object would leak memory on the OST
1722 Details : If there is an inconsistency between the MDS and OST filesystems,
1723 such that the MDS references an object that doesn't exist, writes
1724 to that object will leak memory due to incorrect cleanup in the
1725 error handling path, eventually running out of memory on the OST.
1730 Description: Creating too long symlink causes lustre errors
1731 Details : Check symlink and name lengths before sending requests to MDS.
1734 Frequency : only if flock is enabled (not on by default)
1736 Description: posix locks not released on fd closure on 2.6.9+
1737 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1738 kernels, this caused such locks not to be released on fd close and
1739 then assertions on fs unmount about still used locks.
1742 Frequency : MDS failover only, very rarely
1744 Description: clients may get ASSERTION(granted_lock != NULL)
1745 Details : When request was taking a long time, and a client was resending
1746 a getattr by name lock request. The were multiple lock requests
1747 with the same client lock handle and
1748 mds_getattr_name->fixup_handle_for_resent_request found one of the
1749 lock handles but later failed with ASSERTION(granted_lock != NULL).
1754 Description: handle->h_buffer_credits > 0, assertion failure
1755 Details : h_buffer_credits is zero after truncate, causing assertion
1756 failure. This patch extends the transaction or creates a new
1760 Frequency : NFS re-export or patchless client
1761 Bugzilla : 11179, 10796
1762 Description: Crash on NFS re-export node (__d_move)
1763 Details : We do not want to hash the dentry if we don't have a lock.
1764 But if this dentry is later used in d_move, we'd hit uninitialised
1765 list head d_hash, so we just do this to init d_hash field but
1766 leave dentry unhashed.
1769 Frequency : NFS re-export or patchless client
1771 Description: NFS exports has problem with symbolic link
1772 Details : lustre client didn't properly install dentry when re-exported
1773 to NFS or running patchless client.
1776 Frequency : NFS re-export or patchless client
1778 Description: Various nfs/patchless fixes.
1779 Details : fixes reuse disconected alias for lookup process - this fixes
1780 warning "find_exported_dentry: npd != pd",
1781 fix permission error with open files at nfs.
1782 fix apply umask when do revalidate.
1785 Frequency : occasional
1787 Description: Crash on NFS re-export node
1788 Details : calling clear_page() on the wrong pointer triggered oops in
1789 generic_mapping_read().
1792 Frequency : rarely, using O_DIRECT IO
1794 Description: unaligned directio crashes client with LASSERT
1795 Details : check for unaligned buffers before trying any requests.
1798 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1800 Description: stale data returned from RAID cache
1801 Details : If only a small amount of IO is done to the RAID device before
1802 reading it again it is possible to get stale data from the RAID
1803 cache instead of reading it from disk.
1806 Frequency : always for sles10 kernel
1808 Description: sles10 support
1809 Details : ll_follow_link: compile fixes and using of nd_set_link
1810 under newer kernels.
1813 Frequency : depends on arch, kernel and compiler version, always on sles10
1816 Description: recursive or deep enough symlinks cause stack overflow
1817 Details : getting rid of large stack-allocated variable in
1821 Frequency : depends on hardware
1823 Description: lustre write performance loss in the SLES10 kernel
1824 Details : the performance loss is caused by using of write barriers in the
1825 ext3 code. The SLES10 kernel turns barrier support on by
1826 default. The fix is to undo that change for ldiskfs.
1828 ------------------------------------------------------------------------------
1830 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1832 * Support for kernels:
1833 2.6.9-42.0.3EL (RHEL 4)
1834 2.6.5-7.276 (SLES 9)
1835 2.4.21-47.0.1.EL (RHEL 3)
1836 2.6.12.6 vanilla (kernel.org)
1840 Frequency : quota enabled and large files being deleted
1842 Description: releasing more than 4GB of quota at once hangs OST
1843 Details : If a user deletes more than 4GB of files on a single OST it
1844 will cause the OST to spin in an infinite loop. Release
1845 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1850 Description: statfs data retrieved from /proc may be stale or zero
1851 Details : When reading per-device statfs data from /proc, in the
1852 {kbytes,files}_{total,free,avail} files, it may appear
1853 as zero or be out of date.
1856 Frequency : systems with MD RAID1 external journal devices
1858 Description: lconf's call to blkid is confused by RAID1 journal devices
1859 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1860 of returning all block devices that match the journal UUID.
1863 Frequency : always, for aggregate stripe size over 4GB
1865 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1866 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1867 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1870 Frequency : always if "lfs find" used on a local file/directory
1872 Description: "lfs find" segfaults if used on a local file/directory
1873 Details : The case where a directory component was not specified wasn't
1874 handled correctly. Handle this properly.
1877 Frequency : always on ppc64
1879 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1880 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1881 perform endianness conversion.
1884 Frequency : rarely (truncate to non-zero file size after write under load)
1885 Bugzilla : 10730, 10687
1886 Description: Files padded with zeros to next 4K multiple
1887 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1888 it is possible that files that are truncated to a non-zero size
1889 immediately after being written are filled with zero bytes beyond
1890 the truncated size. No file data is lost.
1892 Severity : enhancement
1894 Description: Allow recovery/failover for liblustre clients.
1895 Details : liblustre clients were unaware of failover configurations until
1898 Severity : enhancement
1900 Description: user file locks should fail when not mounting with flock option
1901 Details : Set up an error-returning stub in ll_file_operations.lock field
1902 to prevent incorrect behaviour when client is mounted without
1903 flock option. Also, set up properly f_op->flock field for
1907 Frequency : always on ia64
1909 Description: "lfs df" loops on printing out MDS statfs information
1910 Details : The obd_ioctl_data was not initialized and in some systems
1911 this caused a failure during the ioctl that did not return
1912 an error. Initialize the struct and return an error on failure.
1915 Frequency : SLES 9 only
1917 Description: Error of copying files with lustre special EAs as root
1918 Details : Client side always return success for setxattr call for lustre
1919 special xattr (currently only "trusted.lov").
1922 Frequency : rarely on clusters with both ia64+i386 clients
1924 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1925 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1926 clients are doing concurrent writes to the same file it is
1927 possible that the ia64 clients may LASSERT because the OST
1928 extent locks are not PAGE_SIZE aligned. Ensure that grown
1929 locks are always aligned on the request boundary.
1932 Frequency : specific use, occasional
1934 Description: Overwriting in use executable truncates on-disk binary image
1935 Details : If one node attempts to overwrite an executable in use by
1936 another node, we now correctly return ETXTBSY instead of
1937 truncating the file.
1939 Severity : enhancement
1941 Description: Async OSC create to avoid the blocking unnecessarily.
1942 Details : If a OST has no remain object, system will block on the creating
1943 when need to create a new object on this OST. Now, ways use
1944 pre-created objects when available, instead of blocking on an
1945 empty osc while others are not empty. If we must block, we block
1946 for the shortest possible period of time.
1951 Description: chmod on Lustre root is propagated to other clients
1952 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1953 invalid by the follow_mount time.
1956 Frequency : liblustre clients only
1958 Description: Race in 'instant cancel' lock handling could lead to such locks
1959 never to be granted in case of SMP MDS
1960 Details : Do not destroy not yet granted but cbpending locks in
1964 Frequency : replay/resend of open
1966 Description: non null lock assetion failure in mds_intent_policy
1967 Details : Trying to replay/resend lockless open requests resulted in
1968 mds_open() returning 0 with no lock. Now it sets a flag if
1969 a lock is going to be returned.
1971 Severity : enhancement
1973 Description: Checksum enhancements
1974 Details : New checksum enhancements allow for resending RPCs that failed
1977 Severity : enhancement
1979 Description: Tunables on number of dirty pages in cacche
1980 Details : Allow to set limit on number of dirty pages cached.
1985 Description: client crash on unmount - lock still has references
1986 Details : In some error handling cases it was possible to leak a lock
1987 reference on a client while accessing a file. This was not
1988 harmful to the client during operation, but would cause the
1989 client to crash when the filesystem is unmounted.
1992 Frequency : specific case, rare
1994 Description: ETXTBSY on mds though file not in use by client
1995 Details : ETXTBSY is no longer incorrectly returned when attempting to
1996 chmod or chown a directory that the user previously tried to
1997 execute or a currently-executing binary.
2000 Frequency : extremely rare except on liblustre-based clients
2002 Description: Lustre space not freed when files are deleted
2003 Details : Clean up open-unlinked files after client eviction. Previously
2004 the unlink was skipped and the files remained as orphans.
2009 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2010 Details : In some cases it was possible to send a blocking callback to
2011 a client doing a glimpse, even though that client didn't get
2012 a lock granted. When the glimpse lock is cancelled on the OST
2013 the freed lock is left on the waiting list and corrupted the list.
2016 Frequency : all core dumps
2018 Description: Broke core dumps to lustre
2019 Details : Negative dentry may be unhashed if parent does not have UPDATE
2020 lock, but some callers, e.g. do_coredump, expect dentry to be
2021 hashed after successful create, hash it in ll_create_it.
2023 ------------------------------------------------------------------------------
2025 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2027 * Support for kernels:
2028 2.6.9-42.0.2.EL (RHEL 4)
2029 2.6.5-7.276 (SLES 9)
2030 2.4.21-40.EL (RHEL 3)
2031 2.6.12.6 vanilla (kernel.org)
2035 Frequency : always on RHEL 3
2037 Description: Number of open files grows over time
2038 Details : The number of open files grows over time, whether or not
2039 Lustre is started. This was due to a filp leak introduced
2040 by one of our kernel patches.
2042 ------------------------------------------------------------------------------
2044 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2046 * Support for kernels:
2047 2.6.9-42.EL (RHEL 4)
2048 2.6.5-7.276 (SLES 9)
2049 2.4.21-40.EL (RHEL 3)
2050 2.6.12.6 vanilla (kernel.org)
2055 Bugzilla : 5719, 9635, 9792, 9684
2056 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2057 Details : If a server is under heavy load and cannot reply to new
2058 connection requests before the client resends the (re)connect,
2059 the connection handling code can behave badly if two service
2060 threads are concurrently handing separate (re)connections from
2061 the same client. Add better locking to the connection handling
2062 code, and ensure that only a single connection will be processed
2063 for a given client UUID, even if the lock is dropped.
2065 Severity : enhancement
2067 Description: add TCP zero-copy support to kernel
2068 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2069 sends if the hardware supports scatter-gather and checksumming.
2070 This allows socklnd to do client-write and server-read more
2071 efficiently and reduce CPU utilization from skbuf copying.
2074 Frequency : only if NFS exporting from client
2076 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2077 Details : If setting attributes on a file created under NFS that had
2078 never been opened it would be possible to oops the client
2079 if the file had no objects.
2082 Frequency : always for liblustre
2084 Description: liblustre client does MDS+OSTs setattr RPC for each write
2085 Details : When doing a write from a liblustre client, the client
2086 incorrectly issued an RPC to the MDS and each OST the file was
2087 striped over in order to update the timestamps. When writing
2088 with small chunks and many clients this could overwhelm the MDS
2089 with RPCs. In all cases it would slow down the write because
2090 these RPCs are unnecessary.
2092 Severity : enhancement
2094 Description: allow number of MDS service threads to be changed at module load
2095 Details : It is now possible to change the number of MDS service threads
2096 running. Adding "options mds mds_num_threads={N}" to the MDS's
2097 /etc/modprobe.conf will set the number of threads for the next
2098 time Lustre is restarted (assuming the "mds" module is also
2099 reloaded at that time). The default number of threads will
2100 stay the same, 32 for most systems.
2105 Description: OST crash if filesystem is unformatted or corrupt
2106 Details : If an OST is started on a device that has never been formatted
2107 or if the filesystem is corrupt and cannot even mount then the
2108 error handling cleanup routines would dereference a NULL pointer.
2113 Description: NULL pointer deref in llap_from_page.
2114 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2115 incorrect) mapping if the page was truncated/reclaimed while it was
2116 searched for. Check for this condition and skip such pages when
2117 doing readahead. Introduce extra check to llap_from_page() to
2118 verify page->mapping->host is non-NULL (so page is not anonymous).
2121 Frequency : Sometimes when using sys_sendfile
2123 Description: "page not covered by a lock" warnings from ll_readpage
2124 Details : sendfile called ll_readpage without right page locks present.
2125 Now we introduced ll_file_sendfile that does necessary locking
2126 around call to generic_file_sendfile() much like we do in
2130 Frequency : with certain MDS communication failures at client mount time
2132 Description: NULL pointer deref after failed client mount
2133 Details : a client connection request may delayed by the network layer
2134 and not be sent until after the PTLRPC layer has timed out the
2135 request. If the client fails the mount immediately it will try
2136 to clean up before the network times out the request. Add a
2137 reference from the request import to the obd device and delay
2138 the cleanup until the network drops the request.
2141 Frequency : occasionally during client (re)connect
2143 Description: assertion failure during client (re)connect
2144 Details : processing a client connection request may be delayed by the
2145 client or server longer than the client connect timeout. This
2146 causes the client to resend the connection request. If the
2147 original connection request is replied in this interval, the
2148 client may trip an assertion failure in ptlrpc_connect_interpret()
2149 which thought it would be the only running connect process.
2152 Frequency : only with obd_echo servers and clients that are rebooted
2154 Description: kernel BUG accessing uninitialized data structure
2155 Details : When running an obd_echo server it did not start the ping_evictor
2156 thread, and when a client was evicted an uninitialized data
2157 structure was accessed. Start the ping_evictor in the RPC
2158 service startup instead of the OBD startup.
2160 Severity : enhancement
2161 Bugzilla : 10193 (patchless)
2162 Description: Remove dependency on various unexported kernel interfaces.
2163 Details : No longer need reparent_to_init, exit_mm, exit_files,
2164 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2167 Frequency : rare (only users of deprecated and unsupported LDAP config)
2169 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2170 Details : LDAP apparently contains 'lustreName' attributes instead of
2171 'name'. A simple remapping of the name is sufficient.
2174 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2176 Description: waiting_locks_callback trips kernel BUG if client is evicted
2177 Details : Running with the dump_on_timeout debug flag turned on makes
2178 it possible that the waiting_locks_callback() can try to dump
2179 the Lustre kernel debug logs from an interrupt handler. Defer
2180 this log dumping to the expired_lock_main() thread.
2182 Severity : enhancement
2184 Description: Support NFS exporting on 2.6 kernels.
2185 Details : Implement non-rawops metadata methods for NFS server to use without
2186 changing NFS server code.
2189 Frequency : very rare (synthetic metadata workload only)
2191 Description: two racing renames might cause an MDS thread to deadlock
2192 Details : Running the "racer" program may cause one MDS thread to rename
2193 a file from being the source of a rename to being the target of
2194 a rename at exactly the same time that another thread is doing
2195 so, and the second thread has already enqueued these locks after
2196 doing a lookup of the target and is trying to relock them in
2197 order. Ensure that we don't try to re-lock the same resource.
2200 Frequency : only very large systems with liblustre clients
2202 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2203 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2204 all OSTs in parallel. This allows the eviction of stale liblustre
2205 clients to proceed much faster than if they were done in series,
2206 and also offers similar improvements for other set_info RPCs.
2211 Description: excessive CPU usage during initial read phase on client
2212 Details : During the initial read phase on a client, it would agressively
2213 retry readahead on the file, consuming too much CPU and impacting
2214 performance (since 1.4.5.8). Improve the readahead algorithm
2215 to avoid this, and also improve some other common cases (read
2216 of small files in particular, where "small" is files smaller than
2217 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2222 Description: MDS crash when receiving packet with unknown intent.
2223 Details : Do not LBUG in unknown intent case, just return -EFAULT
2225 Severity : enhancement
2226 Bugzilla : 9293, 9385
2227 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2228 Details : Do not serialize getattr (non-intent version) and statfs.
2231 Frequency : occasional, when OST network is overloaded/intermittent
2233 Description: client evicted by OST after bulk IO timeout
2234 Details : If a client sends a bulk IO request (read or write) the OST
2235 may evict the client if it is unresposive to its data GET/PUT
2236 request. This is incorrect if the network is overloaded (takes
2237 too long to transfer the RPC data) or dropped the OST GET/PUT
2238 request. There is no need to evict the client at all, since
2239 the pinger and/or lock callbacks will handle this, and the
2240 client can restart the bulk request.
2243 Frequency : Always when mmapping file with no objects
2245 Description: client crashes when mmapping file with no objects
2246 Details : Check that we actually have objects in a file before doing any
2247 operations on objects in ll_vm_open, ll_vm_close and
2253 Description: Request leak when working with deleted CWD
2254 Details : Introduce advanced request refcount tracking for requests
2255 referenced from lustre intent.
2257 Severity : Enhancement
2259 Description: Cache open file handles on client.
2260 Details : MDS now will return special lock along with openhandle, if
2261 requested and client is allowed to hold openhandle, even if unused,
2262 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2263 closing files for every read/write openration.
2265 Severity : Enhancement
2267 Description: Cache open negative dentries on client when possible.
2268 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2269 negative dentries on lock revocation.
2274 Description: Remounting a client read-only wasn't possible with a zconf mount
2275 Details : It wasn't possible to remount a client read-only with llmount.
2277 Severity : enhancement
2278 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2279 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2280 MPICH is included to add Lustre-specific ADIO interfaces.
2281 This is based closely on the UFS ADIO layer and only differs
2282 in file creation, in order to allow the OST striping to be set.
2283 This is user-contributed code and not supported by CFS.
2288 Description: extended inode attributes (immutable, append-only) work improperly
2289 when 2.4 and 2.6 kernels are used on client/server or vice versa
2290 Details : Introduce kernel-independent values for these flags.
2292 Severity : enhancement
2295 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2296 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2297 granularity. Now, Fractional MB granularity is be supported,
2298 this is very useful for low memory system.
2300 Severity : enhancement
2302 Description: Getattr by fid
2303 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2304 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2305 have locks and avoids getattr rpc requests for it on every stat.
2308 Frequency : Always, for filesystems larger than 2TB
2310 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2311 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2312 filesystems larger than 2TB to be mounted with mballoc enabled.
2315 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2317 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2318 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2319 filesystems larger than 2TB is mounted then the kernel will
2320 silently corrupt the start of the filesystem. CONFIG_LBD is
2321 enabled for all CFS-supported kernels, but the possibility of
2322 this happening with a modified kernel config exists.
2324 Severity : enhancement
2326 Description: add client O_DIRECT support for 2.6 kernels
2327 Details : It is now possible to do O_DIRECT reads and writes to files
2328 in the Lustre client mountpoint on 2.6 kernel clients.
2330 Severity : enhancement
2332 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2333 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2334 parallel, not waiting for response from every OST before sending
2335 a rpc to the next OST.
2340 Description: setattr vs write race when updating file timestamps
2341 Details : Client processes that update a file timestamp into the past
2342 right after writing to the file (e.g. tar) it is possible that
2343 the updated file modification time can be reset to the current
2344 time due to a race between processing the setattr and write RPC.
2346 Severity : enhancement
2348 Description: Bring 'lfs find' closer in line with regular Linux find.
2349 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2350 -print0 options and obtains all the needed info through the lustre
2353 Severity : enhancement
2355 Description: support up to 1024 configured devices on one node
2356 Details : change obd_dev array from statically allocated to dynamically
2357 allocated structs as they are first used to reduce memory usage
2362 Description: Flush dirty partially truncated pages during truncate
2363 Details : Immediatelly flush partially truncated pages in filter_setattr,
2364 this way we completely avoid having any pages in page cache on OST
2365 and can retire ugly workarounds during writes to flush such pages.
2370 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2371 Details : protect inode from truncation within vfs_unlink() context
2372 just take a reference before calling vfs_unlink() and release it
2373 when parent's i_sem is free.
2378 Description: last_id value checked outside lock on OST caused LASSERT failure
2379 Details : If there were multiple MDS->OST object precreate requests in
2380 flight, it was possible that the OST's last object id was checked
2381 outside a lock and incorrectly tripped an assertion. Move checks
2382 inside locks, and discard old precreate requests.
2385 Frequency : always, if extents are used on OSTs
2387 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2388 Details : OSTs using the extents format would not zero the high 16 bits of
2389 the index physical block number. This is not a problem for any
2390 OST filesystems smaller than 16TB, and no kernels support ext3
2391 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2392 new/modified files) and can be fixed for existing filesystems
2393 with e2fsprogs-1.39-cfs1.
2398 Description: import connection selection may be incorrect if timer wraps
2399 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2400 import connections to be ignored if the 32-bit jiffies counter
2401 wraps. Use a 64-bit jiffies counter.
2404 Frequency : very large clusters immediately after boot
2406 Description: LNET request buffers exhausted under heavy short-term load
2407 Details : If a large number of client requests are generated on a service
2408 that has previously never seen so many requests it is possible
2409 that the request buffer growth cannot keep up with the spike in
2410 demand. Instead of dropping incoming requests, they are held in
2411 the LND until the RPC service can accept more requests.
2414 Frequency : Sometimes during replay
2416 Description: Assertion failure in ll_local_open after replay.
2417 Details : If replay happened on an open request reply before we were able
2418 to set replay handler, reply will become not swabbed tripping the
2419 assertion in ll_local_open. Now we set the handler right after
2420 recognising of open request
2423 Frequency : very rare
2425 Description: kernel reports "badness in vsnprintf"
2426 Details : Reading from the "recovery_status" /proc file in small chunks
2427 may cause a negative length in lprocfs_obd_rd_recovery_status()
2428 call to vsnprintf() (which is otherwise harmless). Exit early
2429 if there is no more space in the output buffer.
2431 Severity : enhancement
2433 Description: clear OBD RPC statistics by writing to them
2434 Details : It is now possible to clear the OBD RPC statistics by writing
2435 to the "stats" file.
2440 Description: Client mtime is not the same on different clients after utimes
2441 Details : In some cases, the client was using the utimes() syscall on
2442 a file cached on another node. The clients now validate the
2443 ctime from the MDS + OSTs to determine which one is right.
2448 Description: Inability to activate failout mode
2449 Details : lconf script incorrectly assumed that in pythong string's numeric
2450 value is used in comparisons.
2453 Frequency : always with multiple stripes per file
2455 Description: Inefficient object allocation for mutli-stripe files
2456 Details : When selecting which OSTs to stripe files over, for files with
2457 a stripe count that divides evenly into the number of OSTs,
2458 the MDS is always picking the same starting OST for each file.
2459 Return the OST selection heuristic to the original design.
2464 Description: mount failures may take full timeout to return an error
2465 Details : Under some heavy load conditions it is possible that a
2466 failed mount can wait for the full obd_timeout interval,
2467 possibly several minutes, before reporting an error.
2468 Instead return an error as soon as the status is known.
2470 Frequency : quota enabled and large files being deleted
2472 Description: releasing more than 4GB of quota at once hangs OST
2473 Details : If a user deletes more than 4GB of files on a single OST it
2474 will cause the OST to spin in an infinite loop. Release
2475 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2480 Description: statfs data retrieved from /proc may be stale or zero
2481 Details : When reading per-device statfs data from /proc, in the
2482 {kbytes,files}_{total,free,avail} files, it may appear
2483 as zero or be out of date.
2486 Frequency : systems with MD RAID1 external journal devices
2488 Description: lconf's call to blkid is confused by RAID1 journal devices
2489 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2490 of returning all block devices that match the journal UUID.
2493 Frequency : always, for aggregate stripe size over 4GB
2495 Description: assertion fails when trying to use 4GB stripe size
2496 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2497 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2500 Frequency : always on ppc64
2502 Description: the first write on an ext3 filesystem with mballoc got stuck
2503 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2504 perform endianness conversion.
2506 ------------------------------------------------------------------------------
2508 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2510 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2511 INTEROPERATE with older versions automatically. Please read the
2512 user documentation before upgrading any part of a live system.
2513 * WARNING: Lustre networking configuration changes are required with
2514 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2517 * Support for kernels:
2518 2.6.9-22.0.2.EL (RHEL 4)
2519 2.6.5-7.244 (SLES 9)
2520 2.6.12.6 vanilla (kernel.org)
2523 Severity : enhancement
2524 Bugzilla : 7981/8208
2525 Description: Introduced Lustre Networking (LNET)
2526 Details : LNET is new networking infrastructure for Lustre, it includes
2527 a reorganized network configuration mode (see the user
2528 documentation for full details) as well as support for routing
2529 between different network fabrics. Lustre Networking Devices
2530 (LNDs) for the supported network fabrics have also been
2531 created for this new infrastructure.
2533 Severity : enhancement
2534 Description: Introduced Access control lists
2535 Details : clients can set ACLs on files and directories in order to have
2536 more fine-grained permissions than the standard Unix UGO+RWX.
2537 The MDS must be started with the "-o acl" mount option.
2539 Severity : enhancement
2540 Description: Introduced filesystem quotas
2541 Details : Administrators may now establish per-user quotas on the
2544 Severity : enhancement
2546 Description: Configuration change for the XT3
2547 The PTLLND is now used to run Lustre over Portals on the XT3
2548 The configure option(s) --with-cray-portals are no longer used.
2549 Rather --with-portals=<path-to-portals-includes> is used to
2550 enable building on the XT3. In addition to enable XT3 specific
2551 features the option --enable-cray-xt3 must be used.
2556 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2557 Details : A race between allocating a new llcd and re-getting the llcd_lock
2558 allowed another thread to grab newly-allocated llcd.
2560 Severity : enhancement
2562 Description: 2.6 OST async journal commit and locking fix to improve performance
2563 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2564 2.6 kernels are now async as they already were in 2.4 kernels so
2565 that they can commit concurrently with the network bulk transfer.
2566 For block-allocated files the filter allocation semaphore is held
2567 to avoid filesystem fragmentation during allocation. BKL lock
2568 removed for 2.6 xattr operations where it is no longer needed.
2573 Description: lconf incorrectly determined whether two IP networks could talk
2574 Details : In some more complicated routing and multiple-network
2575 configurations, lconf will avoid trying to make a network
2576 connection to a disjoint part of the IP space. It was doing the
2577 math incorrectly for one set of cases.
2582 Description: Fix for potential infinite loop processing records in an llog.
2583 Details : If an llog record is corrupted/zeroed, it is possible to loop
2584 forever in llog_process(). Validate the llog record length
2585 and skip the remainder of the block on error.
2588 Frequency : occasional (liblustre only)
2590 Description: liblustre could not open files whose last component is a symlink
2591 Details : sysio_path_walk() would incorrectly pass the open intent to
2592 intermediate path components.
2595 Frequency : rare (liblustre only with non-standard tuning)
2596 Bugzilla : 7201 (7350)
2597 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2598 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2599 after it has been released, only do this for referenced locks
2601 Severity : enhancement
2603 Description: specifying an (invalid) directory default stripe_size of -1
2604 would reset the directory default striping
2605 Details : stripe_size -1 was used internally to signal directory stripe
2606 removal, now use "all default" to signal dir stripe removal
2607 as a directory striping of "all default" is not useful
2610 Frequency : common for large clusters running liblustre clients
2612 Description: doing an ls when liblustre clients are running is slow
2613 Details : sending a glimpse AST to a liblustre client waits for every AST
2614 to time out, as liblustre clients will not respond. Since they
2615 cannot cache data we refresh the OST lock LVB from disk instead.
2617 Severity : enhancement
2619 Description: doing an ls at the same time as file IO can be slow
2620 Details : enqueue and other "small" requests can be blocked behind many
2621 large IO requests. Create a new OST IO portal for non-IO
2622 requests so they can be processed faster.
2625 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2627 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2628 created file with mode 000
2629 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2630 allow an ftruncate() as a non-root user to a file with mode 000.
2631 Lustre now respects this flag to disable mode checks when
2632 truncating a file owned by the user
2635 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2638 Description: Revoking locks from clients that went dead or catatonic might take
2640 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2641 cancellation of such locks instant on servers without waiting for
2642 any reply from clients. Clients drops these locks when cancel
2643 notification from server is received without replying.
2646 Frequency : liblustre-only, when liblustre client dies or becomes busy
2648 Description: Doing ls on Linux clients can take a long time with active
2650 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2651 granting such locks to it in the first place if possible. Locks
2652 are taken by proxy on the OST during the read or write and
2653 dropped immediately afterward. Add connect flags handling, do
2654 not grant locks to liblustre clients for glimpse ASTs.
2656 Severity : enhancement
2658 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2659 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2660 sequential reads and behaves badly with applications doing random
2661 reads. Improve it by reading ahead at least read region, and
2662 avoiding excessive large RPC for small reads.
2664 Severity : enhancement
2666 Description: Creating more than 1000 files for a single job may cause a load
2667 imbalance on the OSTs if there are also a large number of OSTs.
2668 Details : qos_prep_create() uses an OST index reseed value that is an
2669 even multiple of the number of available OSTs so that if the
2670 reseed happens in the middle of the object allocation it will
2671 still utilize the OSTs as uniformly as possible.
2676 Description: OST or MDS may oops in ping_evictor_main()
2677 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2678 but doesn't restart at beginning of obd_exports_timed list
2681 Severity : enhancement
2683 Description: improve by-nid export eviction on the MDS and OST
2684 Details : allow multiple exports with the same NID to be evicted at one
2685 time without re-searching the exports list.
2688 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2690 Description: MDS may oops in groups_free()
2691 Details : in rare race conditions a newly allocated group_info struct is
2692 freed again, and this can be NULL. The 2.4 compatibility code
2693 for groups_free() checked for a NULL pointer, but 2.6 did not.
2696 Frequency : common for liblustre clients doing little filesystem IO
2697 Bugzilla : 9352, 7313
2698 Description: server may evict liblustre clients accessing contended locks
2699 Details : if a client is granted a lock or receives a completion AST
2700 with a blocking AST already set it would not reply to the AST
2701 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2704 Frequency : lfs setstripe, only systems with more than 160 OSTs
2706 Description: unable to set striping with a starting offset beyond OST 160
2707 Details : llapi_create_file() incorrectly limited the starting stripe
2708 index to the maximum single-file stripe count.
2711 Frequency : LDAP users only
2713 Description: lconf did not handle in-kernel recovery with LDAP properly
2714 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2716 Severity : enhancement
2718 Description: bind OST threads to NUMA nodes to improve performance
2719 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2720 node and do their allocations there to localize memory access
2722 Severity : enhancement
2724 Description: llmount can determine client NID directly from Myrinet (GM)
2725 Details : the client NID code from gmnalnid was moved directly into
2726 llmount, removing the need to use this or specifying the
2727 client NID explicitly when mounting GM clients with zeroconf
2730 Frequency : if client is started with down MDS
2732 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2733 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2734 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2735 ptlrpc_import_delayed_req() because we didn't check if the
2736 request was interrupted, and we also didn't break out of the
2737 event loop if there was no timeout
2742 Description: data loss during non-page-aligned writes to a single file from
2743 both multiple nodes and multiple threads on one node at same time
2744 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2745 inconsistency led to false short-reads, that were cached and later
2746 used by ->prepare_write() to fill in partially written page,
2747 leading to data loss.
2750 Frequency : always, if lconf --abort_recovery used
2752 Description: lconf --abort_recovery fails with 'Operation not supported'
2753 Details : lconf was attempting to abort recovery on the MDT device and not
2756 Severity : enhancement
2758 Description: remove cleanup logs
2759 Details : replace lconf-generated cleanup logs with lustre internal
2760 cleanup routines. Eliminates the need for client-cleanup and
2763 Severity : enhancement
2765 Description: add support for EAs (user and system) on lustre filesystems
2766 Details : it is now possible to store extended attributes in the Lustre
2767 client filesystem, and with the user_xattr mount option it
2768 is possible to allow users to store EAs on their files also
2770 Severity : enhancement
2772 Description: Add possibility (config option) to show minimal available OST free
2774 Details : When compiled with --enable-mindf configure option, statfs(2)
2775 (and so, df) will return least minimal free space available from
2776 all OSTs as amount of free space on FS, instead of summary of
2777 free spaces of all OSTs.
2779 Severity : enhancement
2781 Description: do not expand extent locks acquired on OST-side
2782 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2783 by server: they are not cached anyway.
2786 Frequency : when mmap is used/binaries executed from Lustre
2788 Description: Unmmap pages before throwing them away from read cache.
2789 Details : llap_shrink cache now attempts to unmap pages before discarding
2790 them (if unmapping failed - do not discard). SLES9 kernel has
2791 extra checks that trigger if this unmapping is not done first.
2796 Description: lconf didn't resolve symlinks before checking to see whether a
2797 given mountpoint was already in use
2800 Frequency : when migrating failover services
2801 Bugzilla : 6395, 9514
2802 Description: When migrating a subset of services from a node (e.g. failback
2803 from a failover service node) the remaining services would
2804 time out and evict clients.
2805 Details : lconf --force (implied by --failover) sets the global obd_timeout
2806 to 5 seconds in order to quickly disconnect, but this caused
2807 other RPCs to time out too quickly. Do not change the global
2808 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2810 Severity : enhancement
2811 Frequency : if MDS is started with down OST
2812 Bugzilla : 9439,5706
2813 Description: Allow startup/shutdown of an MDS without depending on the
2814 availability of the OSTs.
2815 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2816 Add appropriate locking and lov-osc refcounts for safe
2817 cleaning. Add osc abort_inflight calls in case the
2818 synchronize never started.
2821 Frequency : occasional (Cray XT3 only)
2823 Description: root not authorized to access files in CRAY_PORTALS environment
2824 Details : The client process capabilities were not honoured on the MDS in
2825 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2826 been accessed by an authorized user then root was able to access
2827 the file on the local client also. The root user capabilities
2828 are now allowed on the MDS, as this environment has secure UID.
2831 Frequency : occasional
2833 Description: ldiskfs "too long searching" message happens too often
2834 Details : A debugging message (otherwise harmless) prints too often on
2835 the OST console. This has been reduced to only happen when
2836 there are fragmentation problems on the filesystem.
2841 Description: Division by zero in statfs when all OSCs are inactive
2842 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2843 lov_statfs divides by value returned by lov_get_stripecnt().
2847 Bugzilla : 9489, 3273
2848 Description: First write from each client to each OST was only 4kB in size,
2849 to initialize client writeback cache, which caused sub-optimal
2850 RPCs and poor layout on disk for the first writen file.
2851 Details : Clients now request an initial cache grant at (re)connect time
2852 and so that they can start streaming writes to the cache right
2853 away and always do full-sized RPCs if there is enough data.
2854 If the OST is rebooted the client also re-establishes its grant
2855 so that client cached writes will be honoured under the grant.
2860 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2861 Details : Now I/O RPCs go to different portal number and (presumably) fast
2862 lock requests (and glimses) and other RPCs get their own service
2863 threads pool that should be able to service those RPCs
2866 Severity : enhancement
2868 Description: Ability to exchange lustre version between client and servers and
2869 issue warnings at client side if client is too old. Also for
2870 liblustre clients there is ability to refuse connection of too old
2872 Details : New 'version' field is added to connect data structure that is
2873 filled with version info. That info is later checked by server and
2877 Frequency : rare, liblustre only.
2878 Bugzilla : 9296, 9581
2879 Description: Two simultaneous writes from liblustre at offset within same page
2880 might proceed at the same time overwriting eachother with stale
2882 Details : I/O lock withing llu_file_prwv was released too early, before data
2883 actually was hitting the wire. Extended lock-holding time until
2884 server acknowledges receiving data.
2887 Frequency : extremely rare. Never observed in practice.
2889 Description: avoid generating lustre_handle cookie of 0.
2890 Details : class_handle_hash() generates handle cookies by incrementing
2891 global counter, and can hit 0 occasionaly (this is unlikely, but
2892 not impossible, because initial value of cookie counter is
2893 selected randonly). Value of 0 is used as a sentinel meaning
2894 "unassigned handle" --- avoid it. Also coalesce two critical
2895 sections in this function into one.
2897 Severity : enhancement
2899 Description: allow liblustre clients to delegate truncate locking to OST
2900 Details : To avoid overhead of locking, liblustre client instructs OST to
2901 take extent lock in ost_punch() on client's behalf. New connection
2902 flag is added to handle backward compatibility.
2904 Severity : enhancement
2905 Bugzilla : 4928, 7341, 9758
2906 Description: allow number of OST service threads to be specified
2907 Details : a module parameter allows the number of OST service threads
2908 to be specified via "options ost ost_num_threads={N}" in the
2909 OSS's /etc/modules.conf or /etc/modprobe.conf.
2913 Bugzilla : 6146, 9635, 9895
2914 Description: servers crash with bad pointer in target_handle_connect()
2915 Details : In rare cases when a client is reconnecting it was possible that
2916 the connection request was the last reference for that export.
2917 We would temporarily drop the export reference and get a new
2918 one, but this may have been the last reference and the export
2919 was just destroyed. Get new reference before dropping old one.
2921 Severity : enhancement
2922 Frequency : if client is started with failover MDS
2924 Description: Allow multiple MDS hostnames in the mount command
2925 Details : Try to read the configuration from all specified MDS
2926 hostnames during a client mount in case the "primary"
2929 Severity : enhancement
2931 Description: Stop sending data to evicted clients as soon as possible.
2932 Details : Check if the client we are about to send or are sending data to
2933 was evicted already. (Check is done every second of waiting,
2934 for which l_wait_event interface was extended to allow checking
2935 of exit condition at specified intervals).
2938 Frequency : rare, normally only when NFS exporting is done from client
2940 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2942 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2945 Frequency : always, if srand() is called before liblustre initialization
2947 Description: Liblustre uses system PRNG disturbing its usage by user application
2948 Details : Introduce internal to lustre fast and high-quality PRNG for
2949 lustre usage and make liblustre and some other places in generic
2950 lustre code to use it.
2952 Severity : enhancement
2953 Bugzilla : 9477, 9557, 9870
2954 Description: Verify that the MDS configuration logs are updated when xml is
2955 Details : Check if the .xml configuration logs are newer than the config
2956 logs stored on the MDS and report an error if this is the case.
2957 Request --write-conf, or allow starting with --old_conf.
2959 Severity : enhancement
2961 Description: Handle symlinks in the path when checking if Lustre is mounted.
2962 Details : Resolve intermediate symlinks when checking if a client has
2963 mounted a filesystem to avoid duplicate client mounts.
2968 Description: lconf can hit an error exception but still return success.
2969 Details : The lconf command catches the Command error exception at the top
2970 level script context and will exit with the associated exit
2971 status, but doesn't ensure that this exit status is non-zero.
2976 Description: failure of ptlrpc thread startup can cause oops
2977 Details : Starting a ptlrpc service thread can fail if there are a large
2978 number of threads or the server memory is very fragmented.
2979 Handle this without oopsing.
2982 Frequency : always, only if liblustre and non-default acceptor port was used
2984 Description: liblustre cannot connect to servers with non-default acceptor port
2985 Details : tcpnal_set_default_params() was not called and was therefore
2986 ignoring the environment varaible TCPNAL_PORT, as well as other
2987 TCPNAL_ environment variables
2992 Description: two objects could be created on the same OST for a single file
2993 Details : If an OST is down, in some cases it was possible to create two
2994 objects on a single OST for a single file. No problems other
2995 than potential performance impact and spurious error messages.
2999 Bugzilla : 5681, 9562
3000 Description: Client may oops in ll_unhash_aliases
3001 Details : Client dcache may become inconsistent in race condition.
3002 In some cases "getcwd" can fail if the current directory is
3008 Description: Inode refcounting problems in NFS export code
3009 Details : link_raw functions used to call d_instantiate without obtaining
3010 extra inode reference first.
3014 Bugzilla : 9942, 9903
3015 Description: Referencing freed requests leading to crash, memleaks with NFS.
3016 Details : We used to require that call to ll_revalidate_it was always
3017 followed by ll_lookup_it. Also with revalidate_special() it is
3018 possible to call ll_revalidate_it() twice for the same dentry
3019 even if first occurence returned success. This fix changes semantic
3020 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3021 reference on a request referred from the intent.
3022 ll_intent_release() then releases such a request.
3025 Frequency : rare, normally benchmark loads only
3027 Description: unlinked inodes were kept in memory on the client
3028 Details : If a client is repeatedly creating and unlinking files it
3029 can accumulate a lot of stale inodes in the inode slab cache.
3030 If there is no other client load running this can cause the
3031 client node to run out of memory. Instead flush old inodes
3032 from client cache that have the same inode number as a new inode.
3035 Frequency : SLES9 2.6.5 kernel and long filenames only
3036 Bugzilla : 9969, 10379
3037 Description: utime reports stale NFS file handle
3038 Details : SLES9 uses out-of-dentry names in some cases, which confused
3039 the lustre dentry revalidation. Change it to always use the
3043 Frequency : rare, unless heavy write-truncate concurrency is continuous
3044 Bugzilla : 4180, 6984, 7171, 9963, 9331
3045 Description: OST becomes very slow and/or deadlocked during object unlink
3046 Details : filter_destroy() was holding onto the parent directory lock
3047 while truncating+unlinking objects. For very large objects this
3048 may block other threads for a long time and slow overall OST
3049 responsiveness. It may also be possible to get a lock ordering
3050 deadlock in this case, or run out of journal credits because of
3051 the combined truncate+unlink. Solution is to do object truncate
3052 first in one transaction without parent lock, and then do the
3053 final unlink in a new transaction with the parent lock. This
3054 reduces the lock hold time dramatically.
3057 Frequency : rare, 2.4 kernels only
3059 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3060 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3061 before unmounting their filesystem in order to avoid deadlock.
3062 The kernel_locked() function in 2.4 kernels only checks whether
3063 the kernel lock is held, not whether it is this process that is
3064 holding it as 2.6 kernels do.
3069 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3070 Details : The client ptlrpc code may be trying to reconnect to a down
3071 server before a previous connection attempt has timed out.
3072 Increase the reconnect interval to be longer than the connection
3073 timeout interval to avoid sending duplicate connections to
3077 Frequency : echo_client brw_test command
3079 Description: fix echo_client to work with OST preallocated code
3080 Details : OST preallocation code (5137) didn't take echo_client IO path
3081 into account: echo_client calls filter methods outside of any
3082 OST thread and, hence, there is no per-thread preallocated
3083 pages and buffers to use. Solution: hijack pga pages for IO. As
3084 a byproduct, this avoids unnecessary data copying.
3088 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3089 Description: Client can oops in mdc_commit_close() after open replay
3090 Details : It was possible for the MDS to return an open request with no
3091 transaction number in mds_finish_transno() if the client was
3092 evicted, but without actually returning an error. Clients
3093 would later try to replay that open and may trip an assertion
3094 Simplify the client close codepath, and always return an error
3095 from the MDS in case the open is not successful.
3098 Frequency : rare, 2.6 OSTs only
3100 Description: OST may deadlock under high load on fragmented files
3101 Details : If there was a heavy load and highly-fragmented OST filesystems
3102 it was possible to have all the OST threads deadlock waiting on
3103 allocation of biovecs, because the biovecs were not released
3104 until the entire RPC IO was completed. Instead, release biovecs
3105 as soon as they are complete to ensure forward IO progress.
3107 Severity : enhancement
3109 Description: Support for specifying external journal device at mount
3110 Details : If an OST or MDS device is formatted with an external journal
3111 device, this device major/minor is stored in the ext3 superblock
3112 and may not be valid for failover. Allow detecting and
3113 specifying the external journal at mount time.
3118 Description: Mounting an MDS with pending unlinked files may cause oops
3119 Details : target_finish_recovery() calls mds_postrecov() which returned
3120 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3121 considers this an error and immediately begins cleaning up the
3122 lov, just after starting the mds_lov process
3124 Severity : enhancement
3126 Description: Implement 'lfs df' to report actual free space on per-OST basis
3127 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3128 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3129 usage of objects; '-h' to report in human readable format.
3131 ------------------------------------------------------------------------------
3133 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
3140 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3141 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3142 at startup during orphan unlinking. Instead, skip buddy update
3143 before setup, as it will be regenerated after recovery is complete.
3148 Description: If an OST is inactive, its locks might reference stale inodes.
3149 Details : lov_change_cbdata() must iterate over all namespaces, even if
3150 they are inactive to clear inode references from the lock.
3152 Severity : enhancement
3153 Frequency : occasional, if non-standard max_dirty_mb used
3155 Description: Client will block write RPCs if not enough grant
3156 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3157 then the client will block writes while waiting for another RPC
3158 to complete instead of consuming its dirty limit. With change
3159 we get improved performance when max_dirty_mb is small.
3161 Severity : enhancement
3162 Bugzilla : 3389, 6253
3163 Description: Add support for supplementary groups on the MDS.
3164 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3165 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3166 lookups for user supplementary groups into a cache.
3170 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3171 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3172 for liblustre O_EXCL is enforced by MDS.
3176 Description: readdir never returns NULL in liblustre.
3177 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3178 dentry in struct dirent.
3182 Frequency : liblustre only, depends on application IO pattern
3183 Description: liblustre clients evicted if not contacting servers
3184 Details : Don't put liblustre clients into the ping_evictor list, so
3185 they will not be evicted by the pinger ever.
3187 Severity : enhancement
3189 Description: Add ability to evict clients by NID from MDS.
3190 Details : By echoing "nid:$NID" string into
3191 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3192 $NID would be instantly evicted from this MDS and from all active
3193 OSTs connected to it.
3197 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3198 Details : lookup_it_finish() used to query file size from OSTs that was not
3203 Description: service threads change working directory to that of init
3204 Details : Starting lustre service threads may pin the working directory
3205 of the parent thread, making that filesystem busy. Threads
3206 now change to the working directory of init to avoid this.
3210 Frequency : during shutdown only
3211 Description: shutdown with a failed MDS or OST can cause unmount to hang
3212 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3217 Frequency : 2.6 only
3218 Description: chmod/chown may include an extra supplementary group
3219 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3220 supplementary group and if none is specified this is used.
3223 Bugzilla : 5479 (6816)
3225 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3226 Details : If lookup is in progress on a file that is unlinked we might try
3227 to revalidate the inode and fail in revalidate after lookup is
3228 complete and ll_file_open() enqueues the open again but
3229 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3232 Frequency : always, if lconf --abort_recovery used
3234 Description: lconf --abort_recovery fails with 'Operation not supported'
3235 Details : lconf was attempting to abort recovery on the MDT device and not
3238 ------------------------------------------------------------------------------
3240 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3245 Frequency : rare (only unsupported configurations with a node running as an
3247 Bugzilla : 6514, 5137
3248 Description: Mounting a Lustre file system on a node running as an OST could
3250 Details : OSTs now preallocates memory needed to write out data at
3251 startup, instead of when needed, to avoid having to
3252 allocate memory in possibly low memory situations.
3253 Specifically, if the file system is mounted on on OST,
3254 memory pressure could force it to try to write out data,
3255 which it needed to allocate memory to do. Due to the low
3256 memory, it would be unable to do so and the node would
3257 become unresponsive.
3259 Severity : enhancement
3261 Description: Addition of lconf --service command line option
3262 Details : lconf now accepts a '--service <arg>' option, which is
3263 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3265 Severity : enhancement
3267 Description: Failover mode is now the default for OSTs.
3268 Details : By default, OSTs will now run in failover mode. To return to
3269 the old behaviour, add '--failout' to the lmc line for OSTs.
3271 Severity : enhancement
3273 Description: Health checks are now provided for MDS and OSTs
3274 Details : Additional detailed health check information on MSD and OSTs
3275 is now provided through the procfs health_check value.
3278 Frequency : occasional, depends on IO load
3280 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3281 after numerous create/delete cycles
3282 Details : The ext3 inode allocation policy would not allocate new inodes
3283 very well on the OSTs because there are no new directories
3284 being created. Instead we look for groups with free space if
3285 the parent directories are nearly full.
3290 Description: Network or server problems during mount may cause partially
3291 mounted clients instead of returning an error.
3292 Details : The config llog parsing code may overwrite the error return
3293 code during mount error handling, returning success instead
3299 Description: MDS can fail to allocate large reply buffers
3300 Details : After long uptimes the MDS can fail to allocate large reply
3301 buffers (e.g. zconf client mount config records) due to memory
3302 fragmentation or consumption by the buffer cache. Preallocate
3303 some large reply buffers so that these replies can be sent even
3304 under memory pressure.
3308 Frequency : rare (liblustre)
3309 Description: fsx running with liblustre complained that using truncate() to
3310 extend the file doesn't work. This patch corrects that issue.
3311 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3312 ATTR_SIZE and lsm use in llu_setattr_raw.
3316 Frequency : rare, only 2.6 kernels
3317 Description: Unusual file access patterns on the MDS may result in inode
3318 data being lost in very rare circumstances.
3319 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3320 code in the 2.6 kernel caused the inode and/or EA data not to
3321 be read from disk, causing single-file corruption.
3325 Frequency : rare, only 2.6 filesystems using extents
3326 Description: Heavy concurrent write and delete load may cause data corruption.
3327 Details : It was possible under high-load situations to have an extent
3328 metadata block in the block device cache from a just-unlinked
3329 file overwrite a newly-allocated data block. We now unmap any
3330 metadata buffers that alias just-allocated data blocks.
3334 Frequency : filesystems with default stripe_count larger than 77
3335 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3336 Details : lconf specifies an inode size of 4096 bytes when the default
3337 stripe_count is larger than 77. This conflicts with the default
3338 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3339 this case to avoid pinning too much memory for large EAs.
3341 ------------------------------------------------------------------------------
3343 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3348 Frequency : rare (extremely heavy IO load with hundreds of clients)
3350 Description: Client is evicted, gets IO error writing to file
3351 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3352 caused clients to be evicted for AST timeouts. The fixes in
3353 bug 5192 mean we no longer need to have such short AST timeouts
3354 so ldlm_timeout has been increased.
3357 Frequency : occasional during --force or --failover shutdown under load
3358 Bugzilla : 5949, 4834
3359 Description: Server oops/LBUG if stopped with --force or --failover under load
3360 Details : a collection of import/export refcount and cleanup ordering
3361 issues fixed for safer force cleanup
3364 Frequency : only filesystems larger than 120 OSTs
3365 Bugzilla : 5990, 6223
3366 Description: lfs getstripe would oops on a very large filesystem
3367 Details : lov_getconfig used kfree on vmalloc'd memory
3370 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3371 Bugzilla : 6242, 6243
3372 Description: reading from files that had been truncated to a non-zero size
3373 but never opened returned no data
3374 Details : ll_file_read() reads zeros from no-object files to EOF
3379 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3380 Details : There's little harm in aborting MDS/OSS recovery and letting it
3381 try again, so I removed the LASSERT and return an error instead.
3383 Severity : enhancement
3385 Description: New debugging infrastructure for tracking down data corruption
3386 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3387 (b) cover more of the client-side code path, and (c) try to narrow
3388 down where problems occurred
3392 Bugzilla : 3819, 4364, 4397, 6313
3393 Description: Racing close and eviction MDS could cause assertion in mds_close
3394 Details : It was possible to get multiple mfd references during close and
3395 client eviction, leading to one thread referencing a freed mfd.
3397 Severity: : enhancement
3398 Bugzilla : 3262, 6359
3399 Description: Attempts to reconnect to servers are now more aggressive.
3400 Details : This builds on the enhanced upcall-less recovery that was added
3401 in 1.4.2. When trying to reconnect to servers, clients will
3402 now try each server in the failover group every 10 seconds. By
3403 default, clients would previously try one server every 25 seconds.
3408 Description: After recovery, certain operations trigger a failed
3409 assertion on a client.
3410 Details : Failing over an mds, using lconf -d --failover, while a
3411 client was doing a readdir() call would cause the client to
3412 LBUG after recovery completed and the readdir() was resent.
3414 Severity : enhancement
3416 Description: Default groups are now added by lconf
3417 Details : You can now run lconf --group <servicename> without having to
3418 manually add groups with lmc.
3421 Frequency : occasional
3423 Description: Nodes with an elan id of 0 trigger a failed assertion
3426 Frequency : always when accessing e.g. tty/console device nodes
3428 Description: tty and some other devices nodes cannot be used on lustre
3429 Details : file's private_data field is used by device data and lustre
3430 values in there got lost. New field was added to struct file to
3431 store fs-specific private data.
3434 Frequency : when exporting Lustre via NFS
3436 Description: NFSD failed occasionally when looking up a path component
3437 Details : NFSD is looking up ".." which was broken in ext3 directories
3438 that had grown large enough to become hashed.
3441 Frequency : Clusters with multiple interfaces not on the same subnet
3443 Description: Nodes will repeatedly try to reconnect to an interface which it
3444 cannot reach and report an error to the log.
3445 Details : Extra peer list entries will be created by lconf with some peers
3446 unreachable. lconf now validates the peer before adding it.
3449 Frequency : Only if a default stripe is set on the filesystem root.
3451 Description: Setting a default stripe on the filesystem root prevented the
3452 filesystem from being remounted.
3453 Details : The client was sending extra request flags in the root getattr
3454 request and did not allocate a reply buffer for the dir EA.
3457 Frequency : occasional, higher if lots of files are accessed by one client
3458 Bugzilla : 6159, 6097
3459 Description: Client trips assertion regarding lsm mismatch/magic
3460 Details : While revalidating inodes the VFS looks up inodes with ifind()
3461 and in rare cases can find an inode that is being freed.
3462 The ll_test_inode() code will free the lsm during ifind()
3463 when it finds an existing inode and then the VFS later attaches
3464 this free lsm to a new inode.
3468 Bugzilla : 6422, 7030
3469 Description: MDS deadlock between mkdir and client eviction
3470 Details : Creating a new file via mkdir or mknod (starting a transaction
3471 and getting the ns lock) can deadlock with client eviction
3472 (gets ns lock and trying to finish a synchronous transaction).
3475 Frequency : occasional
3476 Description: While starting a server, the fsfilt_ext3 module could not be
3478 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3479 kernels. Previously, lconf would still use the ext3 name
3480 when trying to load modules. Now, it will correctly use
3481 ext3 on 2.4 and ldiskfs on 2.6.
3483 Severity : enhancement
3484 Description: The default stripe count has been changed to 1
3485 Details : The interpretation of the default stripe count (0, to lfs
3486 or lmc) has been changed to mean striping across a single
3487 OST, rather than all available. For general usage we have
3488 found a stripe count of 1 or 2 works best.
3490 Severity : enhancement
3491 Description: Add support for compiling against Cray portals.
3492 Details : Conditional compiling for some areas that are different
3496 Frequency : occasional
3497 Bugzilla : 6409, 6834
3498 Description: Creating files with an explicit stripe count may lead to
3499 a failed assertion on the MDS
3500 Details : If some OSTs are full or unavailable, creating files may
3501 trigger a failed assertion on the MDS. Now, Lustre will
3502 try to use other servers or return an error to the
3506 Frequency : occasional
3508 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3509 caused lock timeout during readahead (since 1.4.2).
3510 Details : Processes doing readahead might match a lock that hasn't been
3511 granted yet if there are overlapping and conflicting lock
3512 requests. The readahead process waits on ungranted lock
3513 (original lock is CBPENDING), while OST waits for that process
3514 to cancel CBPENDING read lock and eventually evicts client.
3516 Severity : enhancement
3518 Description: Initial enabling of flock support for clients
3519 Details : Implements fcntl advisory locking and file status functions.
3520 This feature is provided as an optional mount flag (default
3521 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3522 locking are implemented yet, and those that are are not guaranteed
3523 to be completely correct in production environments.
3524 mount -t lustre -o [flock|noflock] ...
3527 Frequency : occasional
3529 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3530 assertion in the ext3 JBD (journaling) layer.
3531 Details : The b_committed_data struct is protected by the big kernel lock
3532 in 2.4 kernels, serializing journal_commit_transaction() and
3533 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3534 access to this struct. In 2.6 kernels there is finer grained
3535 locking to improve SMP performance of the JBD layer.
3539 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3544 Description: merge and cleanup kernel patches.
3547 -----------------------------------------------------------------------------
3549 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3551 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3552 versions of Lustre. Please update all servers and clients to
3553 version 1.4.2 or later at the same time. You must also run
3554 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3555 to update the configuration logs.
3557 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3558 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3559 - swab llog records of type '0' so we get proper header size/idx (5861)
3560 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3561 - fix rename of one directory over another leaking an inode (5953)
3562 - avoid SetPageDirty on 2.6 (5981)
3563 - don't re-add just-being-destroyed locks to the waiting list (5653)
3564 - when creating new directories, inherit the parent's custom
3565 striping settings if present parent (3048)
3566 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3567 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3568 - don't hold client locks on temporary worklist from l_lru (5666)
3569 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3570 - automatically evict dead clients (5921)
3571 - Update file size properly in create+truncate+fstat case (6196)
3572 - Do not unhash mountpoint dentries, do not allow removal of
3574 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3575 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3576 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3577 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3578 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3579 - Correctly handle reads of files with no objects (6243)
3580 - lctl recover will also mark a device active if deactivate used (5933)
3582 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3583 - allow --write-conf on an MDS with different nettype than client (5619)
3584 - don't write config llogs to MDS for mounts not from that MDS (5617)
3585 - lconf should create multiple TCP connections from a client (5201)
3586 - init scripts are now turned off by default; run chkconfig --on
3587 lustre and chkconfig --on lustrefs to use them
3588 - upcalls are no longer needed for clients to recover to failover
3590 - add --abort-recovery option to lconf to abort recovery on device
3592 - add support for an arbitrary number of OSTs (3026)
3593 - Quota support protocol changes.
3594 - forward compatibility changes to wire structs (6007)
3595 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3596 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3597 - improved "lustre status" script
3598 - initialize blocksize for non-regular files (6062)
3599 - added --disable-server and --disable-client configure options (5782)
3600 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3601 - Vanilla 2.4.29 support
3602 - increase maximum number of obd devices to 520 (6242)
3603 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3604 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3605 SLES 9 2.6.5 kernels
3606 - limit stripes per file to 160 (the maximum EA size) (6093)
3608 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3611 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3612 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3613 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3614 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3615 - small ext3 extents cleanups and fixes (5733)
3616 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3617 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3618 - avoid needless client->OST connect, fix handle mismatch (5317)
3619 - fix DLM error path that led to out-of-sync client, long delays (5779)
3620 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3621 - fix several locking issues related to i_size (5492,5624,5654,5672)
3622 - don't move pending lock onto export if it is already evicted (5683)
3623 - fix kernel oops when creating .foo in unlinked directory (5548)
3624 - fix deadlock in obdfilter statistics vs. object create (5811)
3625 - use time_{before,after} to avoid timer jiffies wrap (5882)
3626 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3627 - Do not leak request if server was not able to process it (5154)
3628 - If mds_open unable to find parent dir, make that negative lookup(5154)
3629 - don't create new directories with extent-mapping (5909, 5936)
3631 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3632 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3633 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3634 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3635 - support for RHEL4 kernel on i686 (5773)
3636 - provide error messages when incompatible logs are encountered (5898)
3638 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3639 * version 1.4.0.10 (1.4.1 release candidate 1)
3641 - don't keep a lock reference when lock is not granted (4238)
3642 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3643 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3644 - drop import inflight refcount on signal_completed_replay error (5255)
3645 - unlock page after async write error during send (3677)
3646 - handle missing objects in filter_preprw_read properly (5265)
3647 - no transno return for symlink open, don't save no-trasno open (3440)
3648 - don't try to complete elan receive that already failed (4012)
3649 - free RPC server reply state on error (5406)
3650 - clean up thread from ptlrpc_start_thread() on error (5160)
3651 - readahead could read extra page into cache that wasn't ejected (5388)
3652 - prevent races in class_attach/setup/cleanup/detach (5260)
3653 - don't dereference de->d_inode after l_dput of de (5458)
3654 - use "int" for stripe value returned from lock_to_stripe (5544)
3655 - mballoc allocation and error-checking fixes in 2.6 (5504)
3656 - block device patches to fix I/O request sizes in 2.6 (5482)
3657 - look up hostnames for IB nals (5602)
3658 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3659 - don't start multiple acceptors for the same port (5277)
3660 - fix incorrect LASSERT in mds_getattr_name (5635)
3661 - export a proc file for general "ping" checking (5628)
3662 - fix "lfs check" to not block when the MDS is down (5628)
3664 - service request history (4965)
3665 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3666 - create an "evict_client" /proc entry on OSTs, like the MDS has
3667 - fix mount usage message, return errors per mount(8) (5168)
3668 - change grep [] to grep "[]" in tests so they work in more UMLs
3669 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3670 - remove ancient LOV_MAGIC_V0 EA support (5047)
3671 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3672 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3673 - allow readahead allocations to fail when low on memory (5383)
3674 - mmap locking landed again, after considerable improvement (2828)
3675 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3677 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3680 - send OST transaction number in read/write reply to free req (4966)
3681 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3682 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3683 - the watchdog thread now runs as interruptible (5246)
3684 - flock/lockf fixes (but it's still disabled, pending 5135)
3685 - don't use EXT3 constants in llite code (5094)
3686 - memory shortage at startup could cause assertion (5176)
3688 - reorganization of lov code
3689 - single portals codebase
3691 - add extents/mballoc support (5025)
3692 - direct I/O reads in the obdfilter (4048)
3693 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3695 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3698 - send OST transaction number in read/write reply to free req (4966)
3699 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3700 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3701 - the watchdog thread now runs as interruptible (5246)
3702 - handle missing objects in filter_preprw_read properly (5265)
3703 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3704 - drop import inflight refcount on signal_completed_replay error (5255)
3705 - unlock page after async write error during send (3677)
3706 - return original error code on reconstructed replies (3761)
3707 - no transno return for symlink open, don't save no-trasno open (3440)
3709 - add pid to ldlm debugging output (4922)
3710 - bump the watchdog timeouts -- we can't handle 30sec yet
3711 - extra debugging for orphan dentry/inode bug (5259)
3713 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3716 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3717 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3718 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3719 - fix NULL dereference in /proc/sys/portals/routes (4827)
3720 - allow failed mdc_close() operations to be interrupted (4561)
3721 - stop precreate on OST before MDS would time out on it (4778)
3722 - don't send partial-page writes before EOF from client (4410)
3723 - discard client grant for sub-page writes on large-page clients (4520)
3724 - don't free dentries not owned by NFS code, check generation (4806)
3725 - fix lsm leak if mds_create_objects() fails (4801)
3726 - limit debug_daemon file size, always print CERROR messages (4789)
3727 - use transno after validating reply (3892)
3728 - process timed out requests if import state changes (3754)
3729 - update mtime on OST during writes, return in glimpse (4829)
3730 - add mkfsoptions to LDAP (4679)
3731 - use ->max_readahead method instead of zapping global ra (5039)
3732 - don't interrupt __l_wait_event() during strace
3734 - add software watchdogs to catch hung threads quickly (4941)
3735 - make lustrefs init script start after nfs is mounted
3736 - fix CWARN/ERROR duplication (4930)
3737 - return async write errors to application if possible (2248)
3738 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3739 - print NAL number in %x format (4645)
3740 - update barely-supported suse-2.4.21-171 series (4842)
3741 - support for sles 9 %post scripts
3742 - support for building 2.6 kernel-source packages
3743 - support for sles km_* packages
3745 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3748 - ignore -ENOENT errors in osc_destroy (3639)
3749 - notify osc create thread that OSC is being cleaned up (4600)
3750 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3751 - reconstruct ost_handle() like mds_handle() (4657)
3752 - create a new thread to do import eviction to avoid deadlock (3969)
3753 - let lconf resolve symlinked-to devices (4629)
3754 - don't unlink "objects" from directory with default EA (4554)
3755 - hold socknal file ref over connect in case target is down (4394)
3756 - allow more than 32000 subdirectories in a single directory (3244)
3757 - fix blocks count for O_DIRECT writes (3751)
3758 - OST returns ENOSPC from object create when no space left (4539)
3759 - don't send truncate RPC if file size isn't changing (4410)
3760 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3761 - bind to privileged port in socknal and tcpnal (3689)
3763 - rate limit CERROR/CWARN console message to avoid overload (4519)
3764 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3765 - basic mmap support (3918)
3766 - kernel patch series update from b1_4 (4711)
3768 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3771 - avoid crash during MDS cleanup with OST shut down (2775)
3772 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3773 - don't use bad inodes on the MDS (3744)
3774 - dynamic object preallocation to improve recovery speed (4236)
3775 - don't hold spinlock over lock dumping or change debug flags (4401)
3776 - don't zero obd_dev when it is force cleaned (3651)
3777 - print grants to console if they go negative (4431)
3778 - "lctl deactivate" will stop automatic recovery attempts (3406)
3779 - look for existing locks in ldlm_handle_enqueue() (3764)
3780 - don't resolve lock handle twice in recovery avoiding race (4401)
3781 - revalidate should check working dir is a directory (4134)
3783 - don't always mark "slow" obdfilter messages as errors (4418)
3785 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3788 - don't close LustreDB during write_conf until it is done (3860)
3789 - fix typo in lconf for_each_profile (3821)
3790 - allow dumping logs from multiple threads at one time (3820)
3791 - don't allow multiple threads in OSC recovery (3812)
3792 - fix debug_size parameters (3864)
3793 - fix mds_postrecov to initialize import for llog ctxt (3121)
3794 - replace config semaphore with spinlock (3306)
3795 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3796 - don't allow enqueue to complete on a destroyed export (3822)
3797 - down write_lock before checking llog header bitmap (3825)
3798 - recover from lock replay timeout (3764)
3799 - up llog sem before sending rpc (3652)
3800 - reduce ns lock hold times when setting kms (3267)
3801 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3802 - fix NULL deref and obd_dev leak on setup error (3312)
3803 - replace some LBUG about llog ops with error handling (3841)
3804 - don't match INVALID dentries from d_lookup and spin (3784)
3805 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3806 - fix invalid assertion in ptlrpc_set_wait (3880)
3808 - add libwrap support for the TCP acceptor (3996)
3809 - add /proc/sys/portals/routes for non-root route listing (3994)
3810 - allow setting MDS UUID in .xml (2580)
3811 - print the stack of a process that LBUGs (4228)
3813 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3816 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3817 - make sure to unset replay flag from failed open requests (3440)
3818 - if default stripe count is 0, use OST count for inode size (3636)
3819 - update parent mtime/ctime on client for create/unlink (2611)
3820 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3821 - free recovery state on server during a forced cleanup (3571)
3822 - unregister_reply for resent reqs (3063)
3823 - loop back devices mounting and status check on 2.6 (3563)
3824 - fix resource-creation race that can provoke i_size == 0 (3513)
3825 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3826 - more debugging for page-accounting assertion (3746)
3827 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3828 - avoid deadlock after precreation failure (3758)
3829 - fix race and lock order deadlock in orphan handling (3450, 3750)
3830 - add validity checks when grabbing inodes from l_ast_data (3599)
3832 - add /proc/.../recovery_status to obdfilter (3428)
3833 - lightweight CDEBUG infrastructure, debug daemon (3668)
3834 - change default OSC RPC parameters to be better on small clusters
3835 - turn off OST read cache for files smaller than 32MB
3836 - install man pages and include them in rpms (3100)
3837 - add new init script for (un)mounting lustre filesystems (2593)
3838 - run chkconfig in %post for init scripts (3701)
3839 - drop scimac NAL (unmaintained)
3841 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3844 - clean kiobufs before and after use (3485)
3845 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3846 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3847 - change init script to fail more clearly if not run as root (1528)
3848 - allow clients to reconnect during replay (1742)
3849 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3850 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3851 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3852 - fix cond_resched (3554)
3853 - only evict unfinished clients after recovery (3515)
3854 - allow bulk resend, prevent data loss (3570)
3855 - dynamic ptlrpc request buffer allocation (2102)
3856 - don't allow unlinking open directory if it isn't empty (2904)
3857 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3858 - remove extraneous obd dereference causing LASSERT failure (3334)
3859 - don't use get_cycles() when creating temp. files on the mds (3156)
3860 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3861 - handle EEXIST for set-stripe, set proper directory name (3336)
3863 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3864 - fix ksocknal_fmb_callback() error messages (2918)
3866 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3869 - don't copy lvb into (possibly NULL) reply on error (2983)
3870 - don't deref dentry after dput, don't free lvb on error (2922)
3871 - use the kms to determine writeback rpc length (2947)
3872 - increment oti_logcookies when osc is inactive (2948)
3873 - update client's i_blocks count via lvb messages (2543)
3874 - handle intent open/close of special files properly (1557)
3875 - mount MDS with errors=remount-ro, like obdfilter (2009)
3876 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3877 - don't use cancelling-locks' kms values (2947)
3878 - use highest lock extent for kms, not last one (2925)
3879 - don't dereference ERR_PTR() dentry in error handling path (3107)
3880 - fix thread race in portals_debug_dumplog() (3122)
3881 - create lprocfs device entries at setup instead of at attach (1519)
3882 - common AST error handler, don't evict client on completion race (3145)
3883 - zero nameidata in detach_mnt in 2.6 (3118)
3884 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3885 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3886 - initialize RPC timeout timer earlier for 2.6 (3219)
3887 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3888 - print nal/nid for unknown nid (3258)
3889 - additional checks for oscc recovery before doing precreate (3284)
3890 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3891 - don't crash in mdc_close for bad permissions on open (3285)
3892 - zero i_rdev for non-device files (3147)
3893 - clear page->private before handing to FS, better assertion (3119)
3894 - tune the read pipeline (3236)
3895 - fix incorrect decref of invalidated dentry (2350)
3896 - provide read-ahead stats and refine rpc in flight stats (3328)
3897 - don't hold journal transaction open across create RPC (3313)
3898 - update atime on MDS at close time (3265)
3899 - close LDAP connection when recovering to avoid server load (3315)
3900 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3901 - don't leak open file on MDS after open resend (3325)
3902 - serialize filter_precreate and filter_destroy_precreated (3329)
3903 - loop device shouldn't call sync_dev() for nul device (3092)
3904 - clear page cache after eviction (2766)
3905 - resynchronize MDS->OST in background (2824)
3906 - refuse to mount the same filesystem twice on same mountpoint (3394)
3907 - allow llmount to create routes for mounting behind routers (3320)
3908 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3909 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3910 - fix rare problem with rename on htree directories (3417)
3912 - allow default OST striping configuration per directory (1414)
3913 - fix compilation for qswnal for 2.6 kernels (3125)
3914 - increase maximum number of MDS request buffers for large systems
3915 - change liblustreapi to be useful for external progs like lfsck (3098)
3916 - increase local configuration timeout for slow disks (3353)
3917 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3919 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3922 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3923 - don't overwrite extent policy data in reply if lock was blocked (2901)
3924 - drop filter export grants atomically with removal from device (2663)
3925 - del obd_self_export from work_list in class_disconnect_exports (2908)
3926 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3927 - swab reply message in mdc_close, other PPC fixes (2464)
3928 - fix destroying of named logs (2325)
3929 - overwrite old logs when running lconf --write_conf (2264)
3930 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3931 - fix race in target_handle_connect (2898)
3932 - mds_reint_create() should take same inode create lock (2926)
3933 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3934 - don't close files for self_export to avoid uninitialized obd (2936)
3935 - allow MDS with the same name as client node (2939)
3936 - hold dentry reference for closed log files for unlink (2325)
3937 - reserve space for all logs during transactions (2059)
3938 - don't evict page beyond end of stripe extent (2925)
3939 - don't oops on a deleted current working directory (2399)
3940 - handle hard links to targets without a parent properly (2517)
3941 - don't dereference NULL lock when racing during eviction (2867)
3942 - don't grow lock extents when lots of conflicting locks (2919)
3944 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3947 - account for cache space usage on clients to avoid data loss (974)
3948 - lfsck support in lustre kernel code (2349)
3949 - reduce journal credits needed for BRW writes (2370)
3950 - orphan handling to avoid losing space on client/server crashes
3951 - ptlrpcd can be blocked, stopping ALL progress (2477)
3952 - use lock value blocks to assist in proper KMS, faster stat (1021)
3953 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3954 - recovery for initial connections (2355)
3955 - fixes for mds_cleanup_orphans (1934)
3956 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3957 - block all file creations until orphan recovery completes (1901)
3958 - client remove rq_connection from request struct (2423)
3959 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3960 - recovery timer race (2670)
3961 - mdc_close recovey bug (2532)
3962 - ptlrpc cleanup bug (2710)
3963 - mds timeout on local locks (2588)
3964 - namespace lock held during RPCs (2431)
3965 - handle interrupted sync write properly (2503)
3966 - don't try to handle a message that hasn't been replied to (2699)
3967 - client assert failure during cleanup after abort recovery (2701)
3968 - leak mdc device after failed mount (2712)
3969 - ptlrpc_check_set allows timedout requests to complete (2714)
3970 - wait for inflight reqs when ptlrpcd finishes (2710)
3971 - make sure unregistered services are removed from the srv_list
3972 - reset bulk XID's when resending them (caught by 1138 test)
3973 - unregister_bulk after timeout
3974 - fix lconf error (2694)
3975 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3976 - readahead locks pages, leaves pending causing memory pressure (2673)
3977 - increase OST request buffers to 4096 on large machines (2729)
3978 - fix up permission of existing directories in simple_mkdir (2661)
3979 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3980 - don't assign transno to errored transactions (2742)
3981 - don't delete objects on OST if given a bogus objid from MDS (2751)
3982 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3983 - if rq_no_resend, then timeout request after recovery (2432)
3984 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3985 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3986 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3987 - MDS recovery shouldn't delete objects that it has given out (2730)
3988 - if enqueue arrives after completion, don't clobber LVB (2819)
3989 - don't unlock pages twice when trigger_group_io returns error (2814)
3990 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3991 - don't write pages to disk if there was an error (1450)
3992 - don't ping imports that have recovery disabled (2676)
3993 - take buffered bytes into account when balancing socknal conn (2817)
3994 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3995 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3996 - remove little-endian swabbing of llog records (1987)
3997 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3998 - retry reposting request buffers if they fail (1191)
3999 - grow extent at grant time to avoid granting a revoked lock (2809)
4000 - lock revoke doesn't evict page if covered by a second lock (2765)
4001 - disable VM readahead to avoid reading outside lock extents (2805)
4003 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4004 - updated kernel patches for hp-2.4.20 kernel (2681)
4006 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4009 - fix truncated write corruption (2366)
4010 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4012 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4013 - ensure len(uuid) < 37 in lmc (1171)
4014 - fix ia64 OOPS in llog_test (2255)
4015 - zero end of page at obdfilter for partial page writes (2648)
4016 - don't leave stale dentries around after renames (bug 2428)
4017 - fix timeouts when evicting a client with a single lock held (2642)
4018 - set deadline for the initial HELLO message to drain (2634)
4019 - print out dotted-quad IP addresses in the socknal (2302)
4021 - additional debugging for MDS client eviction problem (2443)
4022 - fix mkfsoptions support for osts (2603, 2604)
4024 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4027 - add series for the vanilla 2.6.0 kernel
4028 - add series for the vanilla 2.4.24 kernel
4029 - add series for a cray x86/64 UL kernel drop
4030 - fix xattr patches for the vanilla 2.4.19 series
4032 - generate true UUIDs in lmc (1171)
4033 - have portals stack dumping break in UML (2466)
4034 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4035 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4036 - wake callback waiting threads on client eviction (2460)
4037 - Add --ptldebug and --subsystem to lmc (1719)
4038 - update assertion to allow safe interrupt allocation
4039 - set rq_no_resend for cancel requests (2432)
4040 - recalculate ptlrpcd timeout after resend (2494)
4041 - call vfs_rmdir when removing pending directories (2368)
4042 - fix renaming a file to itself (2429)
4043 - lmc creates a default one-stripe lov (2454)
4044 - expand procfs space to handle large clusters (2326)
4045 - increase UML stack to avoid overflow
4046 - update lconf's list of debug and subsystem masks
4047 - fix lfs find --obd (2510)
4048 - /proc tunable for disabling filter read caching (2591)
4049 - stop rpm packages from altering slapd.conf (2301)
4050 - disable nagle in the socknal under 0conf (2578)
4051 - choose mds inode size based on stripe count (2572)
4052 - fix kernel-source rpm problems (2516)
4054 - add --disable-doc to avoid pdf generation (2421)
4055 - update documentation, tests, type-os, comments
4056 - avoid format warnings on ia64
4057 - remove the TOE NAL
4058 - tiny code cleanups by removing unused fields
4060 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4063 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4064 - remove the most blatant lies from BUILDING file (2371)
4065 - change default debug level to reasonable production setting
4066 - reduce client side cache size to reduce cache flush time
4067 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4068 - make TCP zerocopy and pinger support enabled by default (2476)
4069 - sync writes completed after process exits caused crashes (2319)
4070 - maintain correct mount count on the MDS (2356)
4071 - backout 1557, because 2316 wasn't really fixed
4072 - better file I/O statistics gathering in /proc
4073 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4074 - another kernel patch to fix zero-copy TCP function export
4075 - don't take duplicate lock when processing re-sent getattr (2420)
4076 - lctl uses obd_self_export instead of creating new conn (2353)
4077 - MDS/OST recovery case which requires object creation asserted (2425)
4078 - move lfs from /usr/sbin to /usr/bin in packages
4079 - fix race between mds_client_add and mds_client_free (2417)
4080 - use kmalloc instead of slabs in portals (2430)
4081 - don't create duplicate records when a failover MDS is present (2442)
4082 - remove unnecessary mount age check (2332)
4083 - don't remove directory inodes from locks prematurely (2451)
4084 - don't break if MDS service name is the same as hostname (2103)
4085 - fix races in client write RPC generation when cache full (2482)
4087 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4090 - remove now-unused request->rq_obd (278)
4091 - if an allocation fails, print out how much memory we've used (1933)
4092 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4093 - add the "configurable stack size" patch to most series files (1256)
4094 - ability to write large log records, for 100+ OST configs (2306)
4095 - fix NULL deref when filter_prep fails (2314)
4096 - fix operator precedence error in filter_sync
4097 - dynamic allocation of socknal TX descriptors (2315)
4098 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4099 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4100 - ensure that CWARN messages in Portals always get to the syslog
4101 - __init/__exit are not for prototype decls (ldlm_init/exit)
4102 - x86-64 compile warning fixes
4103 - fix gateway LMC keyword conflict (2318)
4104 - fix MDS lock inversions in getattr/reint paths (1844)
4105 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4106 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4107 - DLM race condition prevented some lock evictions (2328)
4108 - ENOMEM detection and retry on socknal sends (2230)
4109 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4110 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4111 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4112 - better error messages when a client is rejected during recovery (1505)
4113 - avoid cancelling locks which were never granted, after failure (2330)
4114 - fix i_sem/journal inversion in mds_client_add (2333)
4115 - fix truncate/getattr lock cycle deadlock (2334)
4116 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4117 - fix two rare exit paths which could leak an l_lock() ref (2321)
4118 - fencepost error in MDS/OST orphan recovery (2226)
4119 - make log record alignment 8 bytes (1988)
4120 - lstripe now fails when requested offset > ost_count (2237)
4121 - ensure that all kernel series have a complete list.h (1607)
4122 - fix crashes in special-file operations (2316)
4123 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4125 - allow configurable automake binary, for testing new versions
4126 - small update to the lfs documentation
4128 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4130 * fix negative export reference count in fsfilt_sync (2312)
4132 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4133 * release candidate 0.9.1
4135 - orphans are moved into the PENDING directory for possible recovery
4136 - replayed opens now open by fid for orphan/rename safety (1042)
4137 - last close of an orphan inode generates a transno (683)
4138 - chdir() and mount() now pin the directory entry (1020)
4139 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4140 - discard very old requests without processing them (1502)
4141 - remove some common, well-understood CERRORs (1505)
4142 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4143 - clear "grant" flags in OST replies until OST grant code lands (1644)
4144 - fix read performance by not clobbering i_blksize on client (1598)
4145 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4146 - make LRU size a /proc tunable, clears locks when reduced (707)
4147 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4148 - server locks take references on exports now (1558)
4149 - build fixes for 2.4.20-rh trees (1663)
4150 - return an error from lov_create if all OSCs are inactive (1751)
4151 - fix import levels when a reconnect happens without a timeout (1597)
4152 - exit early from mds_open if we get a lookup error (1749)
4153 - partial page read at EOF wouldn't wait for disk before sending (1642)
4154 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4155 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4156 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4157 - write/truncate lock inversion (1639)
4158 - Don't auto-load obdclass, portals modules during cleanup (1495)
4159 - fix timestamps from jumping to "now" (1763)
4160 - extra journal assertions (1648)
4161 - add an extra multiunlink test (1771)
4162 - fix read_record/write_record API (1776)
4163 - fix leak of offset_extent, possible incorrect i_size later (1772)
4164 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4165 - Debugging for the kqswnal_get_idle_tx problems (1820)
4166 - Allow recovery to be attempted multiple times (1536)
4167 - Write out MDS last_rcvd file after it is first created (1600)
4168 - Fix tx_descriptor leak in failed transmit situations (1827)
4169 - ext3 journaling fixes for assertion failure after IO error (1871)
4170 - class_export_put() on freed export after completion AST error (1896)
4171 - Fix revalidate looping in VFS (1322)
4172 - Don't access a freed export during MDS_REINT timeout (1521)
4173 - Add open-unlink recovery support on the MDS (1673,1764)
4174 - Return an error if no MDS data was read from last_rcvd (1946)
4175 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4176 - Don't setup a disk device that doesn't match exported UUID (317)
4177 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4178 - avoid committing NULL handle in force close
4179 - local.sh is now a one-stripe LOV configuration
4180 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4181 - fix ext3 htree duplicate directory entry corruption (1516)
4182 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4183 - update to new LOV EA format (2097)
4184 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4186 2003-06-15 Phil Schwan <phil@clusterfs.com>
4189 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4190 - per-import/export recovery handling (958, 931, 959)
4191 - multiple last-rcvd slots, for serving multiple FSes (949)
4192 - connections are again shared between multiple imp/exports (963, 964)
4193 - "umount -f" would hang if any requests needed to be sent (393, 978)
4194 - avoid pinning large req buffer by copying for queued messages (989)
4195 - add "uuid" to "lctl device" command to help upcalls (991)
4196 - "open" RPCs with transnos would confuse recovery counters (1037)
4197 - do proper endian conversion of all wire messages (288, 340, 891)
4198 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4199 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4200 - fix ptlrpc_connection leak in target_handle_connect (1174)
4201 - fix import refcounting bug in OST and MDS cleanup (1134)
4202 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4203 - fix very unlikely obd_types race condition (501)
4204 - remove osc_open hack for echo_client (1187)
4205 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4206 - a failure in read_inode2 leads to deadlock (1139)
4207 - cancel ack-locks as soon as transaction is committed (1072)
4208 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4209 - make sure to commitrw after any preprw to avoid deadlock (1162)
4210 - failing to execute a file in a lustre FS would lock inode (1203)
4211 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4212 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4213 - obd_uuid2tgt would walk off the end of the list (1255)
4214 - on IA64 the osc would give portals incorrect bulk size (1258)
4215 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4216 - fix lock inversion caused by new llite matching code (1282)
4217 - limit the number of dirty pages on a client to 10MB (1286)
4218 - timed out locks were not being corrected cancelled (1289)
4219 - fix O_DIRECT above 4GB on IA-32 (1292)
4220 * major user-visible changes
4221 - fail out/fail over policy now controlled by the upcall (993)
4223 - add OBD_PING to check server availability and failure (954)
4224 - lustre messages are now sent in sending host order (288, 340, 891)
4225 - add eadatalen to MDS getattr reply (340)
4226 - OST read replies may contain second buffer, with per-page status (593)
4228 2003-03-11 Phil Schwan <phil@clusterfs.com>
4231 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4232 - failed open()s could cause deadlock; fixed (867, 869)
4233 - stop cancelling OST locks when files are closed (481)
4234 - overlapping XID spaces caused network corruption (851, 853)
4235 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4236 - setattr_raw added, to do single-RPC, server-side setattrs
4237 - lmc/lconf syntax change for OST UUIDs
4238 - fix crashy race condition between ptlrpc_free_req and osc_close
4239 - don't use request in mdc_enqueue if we hit a timeout (889)
4240 - don't set the inode i_size for regular files from the MDS (896)
4241 - handle out of order completion AST (842)
4242 - don't LBUG if a lock request times out after receiving AST (913)
4243 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4244 - if a bad lock AST arrives, send an error instead of dropping entirely
4245 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4246 - fix leak in bulk IO when only partially completed (899, 900, 926)
4247 - fix O_DIRECT for ia64 (55)
4248 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4249 - C-z after timeout could hang a process forever; fixed (977)
4251 - client-side I/O cache (678, 924, 929, 941, 970)
4253 - READPAGE and SETATTRs which don't take server-side locks get
4256 2003-02-11 Phil Schwan <phil@clusterfs.com>
4259 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4260 - Fix fsfilt_extN_readpage() to read a full page of directory
4261 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4262 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4263 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4264 - Most intent operations were removed, in favour of a new RPC mode
4265 that does a single RPC to the server and bypasses most of the VFS
4266 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4267 - Aggressively cancel local locks on DLM servers
4268 - mds_reint_unlink sends EA to the client if it's the last nlink.
4269 client uses that EA to unlink OST objects.
4270 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4271 - recursive symlinks were fixed (439)
4272 - fixed NULL deref in DEBUG_REQ
4273 - filter_update_lastobjid no longer calls sync, which annoyed extN
4274 - fixed multi-client small-writes to a single file problem (445)
4275 - fixed mtime updates during file writes (607)
4276 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4277 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4278 - fixed Linux OST size reporting problem (444, 656)
4279 - OST now updates object mtime with writes or setattr (607, 619)
4280 - client verifies file size before zeroing page past EOF (445)
4281 - OST now writes last allocated objid to disk with allocation (108)
4282 - LOV on echo now works (409)
4284 - mds_reint_unlink sends a new buffer, with the EA included. this
4285 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4286 set if a regular file was being unlinked, and it was the last link
4287 - use PtlGet from the target for bulk writes (315)
4288 - OST now updates object mtime with writes or setattr (607, 619)
4289 - LDLM now has a grant-time callback to revalidate locked items, if
4291 - Many MDS operations were reorganized to combat race conditions
4293 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4294 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4296 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4299 - Fully reactivate OST imports after reconnection (512, others)
4300 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4301 - More graceful error handling for truncating on dead OST (515)
4302 - Don't error out unless we're actually accessing dead stripes (474)
4303 - Fix garbage sizes when stripes are missing (410)
4304 - LRU counters were broken, causing constant lock purge (433, 432)
4305 - garbage on read from stripes with failed OSTs (441)
4306 - mark OSCs as active before reconnecting during recovery (438)
4307 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4308 - lfind did not preserve OST order in output (443)
4309 - symlinks cause hung clients, incorrect data (439)
4310 - stop dereferencing request after dropping refcount (457)
4311 - don't LASSERT(spin_is_locked) on non-SMP (455)
4312 - fixes for many rename() bugs
4313 - fstat didn't correctly synchronize attributes (399)
4314 - server must handle lock cancellation during blocking AST prep (487)
4315 - bulk descriptors were free()d too soon (511)
4316 - fix paths in lconf, which would load incorrect modules (451, 507)
4317 - fix confusing lconf 'host not found' error message (386)
4318 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4319 - fix race condition in mdc_blocking_ast() for inode access (526)
4320 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4321 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4322 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4323 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4324 - fix lstripe in several ways (kernel side) (527)
4325 - fix request leak in ldlm_cli_enqueue (262)
4326 - incorrect OSC was marked inactive after OST failure
4327 - call mds_fs_cleanup before unmounting filesystem (524)
4328 - fix races between taking ns_lock and ldlm_lock_change_resource
4329 - fix races updating LOV export open file list
4330 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4331 - fix recovery NULL deref in ldlm_cli_cancel_unused
4332 - fix some DLM races by using new hash table for lock handles (419)
4333 - permit the client to specify desired inodes, at replay
4334 - duplicate requests when we queue them for replay reintegration
4335 - fix last_rcvd offset calculation
4336 - sync after each recovered transaction, so we always make progress
4337 - never, not always, ERESTART requests without transnos
4338 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4340 - skip replay if the MDS says that the client is already connected
4341 - don't check for a recovery-enabled export to match lctl's UUID
4342 - don't INC_USE_COUNT for phantom exports
4343 - don't crash when cleaning up phantom exports (567)
4344 - don't double-finish or set replay data for errored mdc_open requests
4345 - abort requests when they time out, so we don't get old replies
4346 - send/receive replies for AST messages again
4347 - if the client says that it doesn't have the lock, cancel it on the
4349 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4350 mark it as destroyed, it will all work out when decref is called
4351 - fix module use counts (22, 581)
4353 - ASTs now expect a reply (server cancels lock on error reply)
4355 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4358 - fix many simultaneous client startup (392)
4359 - fix dentry->d_it clobbering
4360 - credentials weren't being shipped for readdir/getattr operations
4361 - remove invalid assertions triggered during some concurrent MD
4363 - proper Lustre versions added (336, 389)
4364 - fix memory leak for create error case (398)
4365 - fix LOV locking bug that would get cli/srv out of sync
4366 - fix echo client over LOV (409)
4367 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4368 - fix double-O_EXCL intent crash (424)
4369 - avoid sending multiple lock CANCELs (352)
4371 - MDS can do multi-client recovery (modulo bugs in new code)
4373 - many updates, edits, cleanups
4375 2002-11-18 Phil Schwan <phil@clusterfs.com>
4378 - fix null d_it dereference (346)
4379 - fix full OST/dbench hang (333)
4380 - fix permission problem with file removal (286)
4381 - fix removal of OSCs from LOV when they fail
4382 - fix NULL deref during bulk timeout (214)
4383 - fix problems related to multiple filesystems on one MDS (241)
4384 - fixed serious subtle metadata locking bugs
4385 - free locks on clients when inodes are removed due to memory
4387 - fix inode pointer in lock data (285)
4388 - partial support for multiple MDS on a single host (241)
4389 - data locks weren't cancelled at clear_inode time (290, 311)
4390 - intent locks could lead to unbounded lock growth (205)
4391 - added a maximum lock count, an LRU list, and a flusher
4392 - fix multiple rename (365)
4393 - properly abstracted the echo client
4394 - OSC locked 1 byte too many; fixed
4395 - rewrote brw callback code:
4396 - fixed recovery bugs related to LOVs (306)
4397 - fixed too-many-pages-in-one-write crash (191)
4398 - fixed (again) crash in sync_io_timeout (214)
4399 - probably fixed callback-related race (385)
4401 - Add capability to MDS protocol
4402 - LDLM cancellations and callbacks on different portals
4404 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4407 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4408 - defer open object destruction to close time (601981, 138)
4409 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4410 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4411 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4412 - don't call obd_setattr in ll_file_release for destroyed objects
4413 * protocol change to lustre_msg: move |version| and add |flags|
4414 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4415 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4416 * added replay of create, unlink, link and rename operations during
4417 MDS failover; recovery should be much more robust now
4418 * remove failed OSCs from LOVs (only lov_create uses this so far)
4419 * the lustre-HOWTO was brought (more) up to date (582544)
4421 2002-10-23 Phil Schwan <phil@clusterfs.com>
4424 - in-use dentries weren't being reused properly (617851)
4425 - prevent multiple LDLM setup (599178)
4426 - fix LOV size calculations for truncate (617853)
4427 - fix client handling of MDS intent errors (POSIX)
4428 - fix permission bug in lovstripe.c test (624321)
4429 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4430 - truncate past end of file could corrupt data
4431 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4432 - a race in recovery could return ETIMEDOUT to apps (623947)
4433 - building outside the source directory was fixed
4434 * the lustre-HOWTO was brought (more) up to date (582544)
4435 * major progress was made on recovery functionality
4437 2002-10-10 Phil Schwan <phil@clusterfs.com>
4440 - recovery deadlock fix
4441 - rm -rf causes LBUG fix (617817)
4442 - file open by multiple tasks fix (618962)
4443 - directory permissions bugs (602707 and 620007)
4444 - journal_stop fixed with locking (611313)
4445 - O_APPEND failures resolved (618273, perhaps 614459)
4446 - lconf PATH fix (619770)
4447 - IA64 build fix (621450)
4448 - RPC buffer sizes scale with amount of memory
4450 2002-10-01 Phil Schwan <phil@clusterfs.com>
4453 - locks would be cancelled without throwing away data pages,
4454 resulting in inconsistent data (605627)
4455 - inode attributes were not always being refreshed (605627, 612449)
4456 - lconf now continues to cleanup after lctl reports an error
4457 - MDS now enforces user permissions (602707)
4458 - lprocfs cleanup fixed, but not yet enabled (614157)
4459 - fixed infinite server hang, should a client not respond to an AST
4460 - avoid going into recovery if user calls readlink() with a buffer
4461 that's too small (613941)
4462 - AST RPCs no longer require replies (614867) -- this may be changed
4463 - don't crash server if client sends an IOV that's too big (611336)
4464 - fixed lock conversion deadlock (611892)
4465 - fixed the following of symlinks (614622)
4466 * recovery: the server can remove locks from a client that dies, other
4467 clients can make progress
4468 * more extN patch fixes
4469 * compile-time configurable ptlrpc buffer allocations
4471 - collaborative read cache document
4472 - Lustre Lite Performance CDR document-in-progress
4474 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4477 - fix typo in patch-2.4.18
4479 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4482 - clear ptlrpc request each time in handle_incoming_request()
4483 - unlink of files now destroys the object on the OST
4485 2002-09-19 Peter Braam <braam@clusterfs.com>
4487 * add hard link support
4488 * change obdfile creation method
4489 * kernel patch changed
4491 2002-09-19 Peter Braam <braam@clusterfs.com>
4494 - stack overflow bug in extN fixed
4496 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4498 * documentation updates
4499 - add man pages for config tools
4500 - update tests/README to describe testing with new config tools
4501 - finish metadata API descriptions
4502 * bug fixes and cleanups
4503 - statfs workaround for 16TB limit
4504 - LOV stripe allocation improved, can stripe on subset of OSTs
4505 - LOV file size/IO offset was wrong for files > 4GB in size
4506 - object EA data was being dropped, caused files to be unreadable
4507 - memory overflow with non-LOV OST caused memory corruption
4508 - fixed regression tests to work with new config tools, obdfilter
4509 - fixed bug when directory size became larger than 1 block
4510 - fixed bug (for single client case) when PWD was deleted
4511 - invalidate local directory pages when doing intent-based ops
4512 - avoid LDLM oops when lock callback contained bad data
4514 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4516 * documentation updates
4517 * bug fixes and cleanups
4518 - configuration tools
4521 - 64-bit compile warnings
4522 - 64-bit internal statfs data
4524 * test_brw on persistent OST devices
4526 * lprocfs (disabled)
4528 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4530 * documentation updates
4531 * bug fixes and cleanups
4532 * configuration tools
4534 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4537 * many small fixes to 0_5_4
4538 * io/network handling
4539 * thinkos in MDS operations
4541 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4544 * crucial basic fixes to 0.5.3
4545 * IOR, Iozone work over Elan
4548 2002-08-07 Phil Schwan <phil@clusterfs.com>
4549 * version 0_5_3, our first alpha
4550 * we use the new Portals iovs
4551 * documentation updates
4552 * bug fixes and cleanups
4553 * small changes in the DLM wire protocol
4555 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4556 * version 0_5_1 with some initial stability,
4557 * locking on MD and file I/O.
4558 * documentation updates
4559 * several bug fixes since 0.5.0
4560 * small changes in wire protocol
4562 2002-07-18 Phil Schwan <phil@clusterfs.com>
4564 * delivered as Lustre Light Alpha
4565 * fixed a crash after handling invalid MDS requests
4566 * fixed directory pages for architectures with non-4k pages sizes
4568 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4569 * release version v0_4_4
4570 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4572 * Fixes a number of interruption problems with OST operations.
4573 * Update documentation for portals header changes
4574 * Move all wire protocol structs/defines to lustre_idl.h
4575 * Fixes symlink length bug.
4576 * Add tcpdump to repository.
4578 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4579 * release version v0_4_3
4580 * Fixes statfs for inodes on extN.
4581 * Fixes bug in runtests which would delete /etc/hosts.
4582 * Use 64-bit object IDs wherever possible (not into VFS though)
4583 Remove ost_get_info, which is unused by lustre, and out of date.
4585 2002-07-03 Peter Braam <braam@clusterfs.com>
4586 * release version v0_4_2 Fixes a lookup error (type not passed)
4587 * move forward to head of Portals
4588 * move forward to latest Lustre kernel
4590 2002-06-25 Peter Braam <braam@clusterfs.com>
4591 * release version v0_4_1. Hopefully stable on single node use.