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
834 Description: build for x2 fails
835 Details : fix compile issue on Cray systems.
839 Description: Properly match for duplicate locks
840 Details : Due to different lock order from skiplists code, we need to
841 traverse entire list for now
843 --------------------------------------------------------------------------------
845 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
847 * Support for kernels:
848 2.6.5-7.283 (SLES 9),
849 2.6.9-55.EL (RHEL 4),
850 2.6.16.46-0.14 (SLES 10),
851 2.6.18.8 vanilla (kernel.org)
852 * Client support for unpatched kernels:
853 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
854 2.6.16 - 2.6.22 vanilla (kernel.org)
855 * Due to recently discovered recovery problems, we do not recommend
856 using patchless RHEL 4 clients with this or any earlier release.
857 * Recommended e2fsprogs version: 1.39.cfs8
858 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
859 * Starting with this release, the ldiskfs backing filesystem required
860 by Lustre is now in its own package, lustre-ldiskfs. This package
861 should be installed. It is versioned separately from Lustre and
862 may be released separately in future.
866 Description: block reactivating mgc import until all deactivates complete
867 Details : Fix race when failing back MDT/MGS to itself (testing)
870 Frequency : at statup only
872 Description: mds_lov_synchronize race leads to various problems
873 Details : simultaneous MDT->OST connections at startup can cause the
874 sync to abort, leaving the OSC in a bad state.
876 Severity : enhancement
878 Description: add optional extra BUILD_VERSION info
879 Details : add a new environment variable (namely LUSTRE_VERS) which allows
880 to override the lustre version.
883 Frequency : 2.6.18 servers only
885 Description: ll_kern_mount() doesn't release the module reference
886 Details : The ldiskfs module reference count never drops down to 0
887 because ll_kern_mount() doesn't release the module reference.
892 Description: server LBUG when using old ost_num_threads parameter
893 Details : Accept the old ost_num_threads parameter but warn that it
894 is deprecated, and fix an off-by-one error that caused an LBUG.
899 Description: Transient SCSI error results in persistent IO issue
900 Details : iobuf->dr_error is not reinitialized to 0 between two
904 Frequency : sometimes when underlying device returns I/O errors
906 Description: OSTs not going read-only during write failures
907 Details : OSTs are not remounted read-only when the journal commit threads
908 get I/O errors because fsfilt_ext3 calls journal_start/stop()
909 instead of the ext3 wrappers.
913 Description: poor connect scaling with increasing client count
914 Details : Don't run filter_grant_sanity_check for more than 100 exports
915 to improve scaling for large numbers of clients.
918 Frequency : SLES10 only
920 Description: sanity-quota.sh quotacheck failed: rc = -22
921 Details : Quotas cannot be enabled on SLES10.
924 Frequency : liblustre clients only
926 Description: getdirentries does not give error when run on compute nodes
927 Details : getdirentries does not fail when the size specified as an argument
928 is too small to contain at least one entry
930 Severity : enhancement
932 Description: Add LNET router traceability for debug purposes
933 Details : If a checksum failure occurs with a router as part of the
934 IO path, the NID of the last router that forwarded the bulk data
935 is printed so it can be identified.
940 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
941 Details : Due to a race condition, liblustre clients were occasionally
944 Severity : enhancement
946 Description: lfs setstripe use optional parameters instead of postional
949 Severity : enhancement
951 Description: Nanosecond timestamp support for ldiskfs
952 Details : The on-disk ldiskfs filesystem has added support for nanosecond
953 resolution timestamps. There is not yet support for this at
954 the Lustre filesystem level.
957 Frequency : during server recovery
959 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
960 Details : request with rq_no_resend flag not awake l_wait_event if they get a
964 Frequency : nfs export on patchless client
966 Description: connectathon hang when test nfs export over patchless client
967 Details : Disconnected dentry cannot be found with lookup, so we do not need
968 to unhash it or make it invalid
971 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
972 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
973 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
978 Description: open req refcounting wrong on reconnect
979 Details : If reconnect happened between getting open reply from server and
980 call to mdc_set_replay_data in ll_file_open, we will schedule
981 replay for unreferenced request that we are about to free.
982 Subsequent close will crash in variety of ways.
983 Check that request is still eligible for replay in
984 mdc_set_replay_data().
989 Description: disable writes to filesystem when reading health_check file
990 Details : the default for reading the health_check proc file has changed
991 to NOT do a journal transaction and write to disk, because this
992 can cause reads of the /proc file to hang and block HA state
993 checking on a healthy but otherwise heavily loaded system. It
994 is possible to return to the previous behaviour during configure
995 with --enable-health-write.
997 Severity : enhancement
999 Description: 64-bit inode version
1000 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1001 to the inode. This will be required for version-based recovery.
1006 Description: MDS fails to start if a duplicate client export is detected
1007 Details : in some rare cases it was possible for a client to connect to
1008 an MDS multiple times. Upon recovery the MDS would detect this
1009 and fail during startup. Handle this more gracefully.
1011 Severity : enhancement
1013 Description: Add -o localflock option to simulate old noflock
1015 Details : This will achieve local-only flock/fcntl locks
1021 Description: log_commit_thread vs filter_destroy race leads to crash
1022 Details : Take import reference before releasing llog record semaphore
1027 Description: Wrong request locking in request set processing
1028 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1029 imp_delayed_list, in this place should be used imp_lock.
1032 Frequency : when reconnection
1034 Description: Grant Leak when osc reconnect to OST
1035 Details : When osc reconnect ost, OST(filter) should check whether it
1036 should grant more space to client by comparing fed_grant and
1037 cl_avail_grant, and return the granted space to client instead
1038 of "new granted" space, because client will call osc_init_grant
1039 to update the client grant space info.
1042 Frequency : when client reconnect to OST
1044 Description: Grant Leak when osc do resend and replay bulk write
1045 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1046 bulk write request, because the grant info while be sync between
1047 OSC and OST when reconnect, and we should ignore the grant info
1048 these of resend/replay write req.
1053 Description: Grant space more than avaiable left space sometimes.
1054 Details : When then OST is about to be full, if two bulk writing from
1055 different clients came to OST. Accord the avaliable space of the
1056 OST, the first req should be permitted, and the second one
1057 should be denied by ENOSPC. But if the seconde arrived before
1058 the first one is commited. The OST might wrongly permit second
1059 writing, which will cause grant space > avaiable space.
1062 Frequency : when client is evicted
1064 Description: Grant might be wrongly erased when osc is evicted by OST
1065 Details : when the import is evicted by server, it will fork another
1066 thread ptlrpc_invalidate_import_thread to invalidate the
1067 import, where the grant will be set to 0. While the original
1068 thread will update the grant it got when connecting. So if
1069 the former happened latter, the grant will be wrongly errased
1070 because of this race.
1075 Description: Checking Stale with correct fid
1076 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1077 to check whether it is stale, because sometimes, we want the
1078 enqueue happened anyway, and op_data.fid2 will not be initialized.
1080 Severity : enhancement
1082 Description: update patchless client
1083 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1086 Frequency : only with 2.4 kernel
1088 Description: random memory corruption
1089 Details : size of struct ll_inode_info is to big for union inode.u and this
1090 can be cause of random memory corruption.
1095 Description: Memory leak in recovery
1096 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1097 It should also check obd_fail before fsfilt_start, otherwise if
1098 fsfilt_start return -EROFS,(failover mds during mds recovery).
1099 then the req will return with repmsg->transno = 0 and rc = EROFS.
1100 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1101 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1102 be freed no matter whether fsfilt_commit success or not.
1105 Frequency : only with huge count clients
1107 Description: Prevents from taking the superblock lock in llap_from_page for
1109 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1110 for llap_from_page call in ll_removepage prevents from taking the
1111 superblock lock for a soon died page.
1116 Description: Not check open intent error before release open handle
1117 Details : in some rare cases, the open intent error is not checked before
1118 release open handle, which may cause
1119 ASSERTION(open_req->rq_transno != 0), because it tries to release
1120 the failed open handle.
1125 Description: Set cat log bitmap only after create log success.
1126 Details : in some rare cases, the cat log bitmap is set too early. and it
1127 should be set only after create log success.
1131 Description: Accessing a block bevice can re-enable I/O when Lustre is
1132 tearing down a device.
1133 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1138 Description: service threads may hog cpus when there are a lot of requests
1139 Details : Insert cond_resched to give other threads a chance to use some CPU
1144 Description: the cat log was not initialized in recovery
1145 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1146 unlink log on mds will not be initialized until mds post
1147 recovery. And also in mds post recovery, the unlink log will
1148 initialization will be done asynchronausly, so there will be race
1149 between add unlink log and unlink log initialization.
1153 Description: brw_stats were being printed incorrectly
1154 Details : brw_stats were being printed as log2 but all of them were not
1155 recorded as log2. Also remove some code duplication arising from
1156 filter_tally_{read,write}.
1160 Frequency : rare, only in recovery.
1161 Description: ASSERTION(req->rq_type != LI_POISON) failed
1162 Details : imp_lock should be held while iterating over imp_sending_list for
1163 prevent destroy request after get timeout in ptlrpc_queue_wait.
1167 Description: replay-single.sh test 52 fails
1168 Details : A lock's skiplist need to be cleanup when it being unlinked
1169 from its resource list.
1173 Description: Short directio read returns full requested size rather than
1175 Details : Direct I/O operations should return actual amount of bytes
1176 transferred rather than requested size.
1178 Severity : enhancement
1180 Description: metadata RPC reduction (e.g. for rm performance)
1181 Details : decrease the amount of synchronous RPC between clients and servers
1182 by canceling conflicing lock before the operation on the client
1183 and packing thier handles into the main operation RPC to server.
1185 Severity : enhancement
1187 Description: Async OSC create to avoid the blocking unnecessarily.
1188 Details : If a OST has no remain object, system will block on the creating
1189 when need to create a new object on this OST. Now, ways use
1190 pre-created objects when available, instead of blocking on an
1191 empty osc while others are not empty. If we must block, we block
1192 for the shortest possible period of time.
1194 Severity : enhancement
1196 Description: refine locking for avoid write wrong info into lov_objid file
1197 Details : fix possible races with add new target and write/update data in
1200 --------------------------------------------------------------------------------
1202 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1207 Frequency : on some architectures
1209 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1210 Details : Uninitialized flags sometimes cause configuration commands to
1214 Frequency : patchless clients only
1216 Description: missing __iget() symbol export
1217 Details : The __iget() symbol export is missing. To avoid the need for
1218 this on patchless clients the deathrow inode reaper is turned
1219 off, and we depend on the VM to clean up old inodes. This
1220 dependency was during via the fix for bug 12181.
1222 --------------------------------------------------------------------------------
1224 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1226 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1227 INTEROPERATE with older versions automatically. In many cases a
1228 special upgrade step is needed. Please read the
1229 user documentation before upgrading any part of a live system.
1230 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
1231 version will not interoperate with 1.6 betas before beta5 (1.5.95).
1232 * WARNING: Lustre configuration and startup changes are required with
1233 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1239 Severity : enhancement
1241 Description: MountConf
1242 Details : Lustre configuration is now managed via mkfs and mount
1243 commands instead of lmc and lconf. New obd types (MGS, MGC)
1244 are added for dynamic configuration management. See
1245 https://mail.clusterfs.com/wikis/lustre/MountConf for
1248 Severity : enhancement
1250 Description: dynamic OST addition
1251 Details : OSTs can now be added to a live filesystem
1253 Severity : enhancement
1255 Description: startup order invariance
1256 Details : MDTs and OSTs can be started in any order. Clients only
1257 require the MDT to complete startup.
1259 Severity : enhancement
1261 Description: parallel, asynchronous orphan cleanup
1262 Details : orphan cleanup is now performed in separate threads for each
1263 OST, allowing parallel non-blocking operation.
1265 Severity : enhancement
1267 Description: optimized stripe assignment
1268 Details : stripe assignments are now made based on ost space available,
1269 ost previous usage, and OSS previous usage, in order to try
1270 to optimize storage space and networking resources.
1272 Severity : enhancement
1274 Description: Permanently set tunables
1275 Details : All writable /proc/fs/lustre tunables can now be permanently
1276 set on a per-server basis, at mkfs time or on a live system.
1278 Severity : enhancement
1280 Description: Lustre message v2
1281 Details : Add lustre message format v2.
1283 Severity : enhancement
1285 Description: client OST exclusion list
1286 Details : Clients can be started with a list of OSTs that should be
1287 declared "inactive" for known non-responsive OSTs.
1291 Description: SPEC SFS validation failure on NFS v2 over lustre.
1292 Details : Changes the blocksize for regular files to be 2x RPC size,
1293 and not depend on stripe size.
1295 Severity : enhancement
1297 Description: Multiple MD RPCs in flight.
1298 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1299 To avoid overly-overloading MDS, introduce a limit on number of
1300 MDS RPCs in flight for a single client and add /proc controls
1301 to adjust this limit.
1303 Severity : enhancement
1305 Description: client read/write statistics
1306 Details : Add client read/write call usage stats for performance
1307 analysis of user processes.
1308 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1309 file access. extents_stats shows chunk size distribution.
1310 extents_stats_per_process show chunk size distribution per
1313 Severity : enhancement
1315 Description: mds statistics
1316 Details : Add detailed mds operations statistics in
1317 /proc/fs/lustre/mds/*/stats.
1321 Description: Failure of copying files with lustre special EAs.
1322 Details : Client side always return success for setxattr call for lustre
1323 special xattr (currently only "trusted.lov").
1328 Description: Refcount LNET uuids
1329 Details : The global LNET uuid list grew linearly with every startup;
1330 refcount repeated list entries instead of always adding to
1333 Severity : enhancement
1335 Description: Dynamic service threads
1336 Details : Within a small range, start extra service threads
1337 automatically when the request queue builds up.
1340 Frequency : mixed-endian client/server environments
1342 Description: mixed-endian crashes
1343 Details : The new msg_v2 system had some failures in mixed-endian
1346 Severity : enhancement
1348 Description: Easy OST removal
1349 Details : OSTs can be permanently deactivated with e.g. 'lctl
1350 conf_param lustre-OST0001.osc.active=0'
1352 Severity : enhancement
1354 Description: MGS proc entries
1355 Details : Added basic proc entries for the MGS showing what filesystems
1358 Severity : enhancement
1360 Description: provide MGS failover
1361 Details : Added config lock reacquisition after MGS server failover.
1363 Severity : enhancement
1365 Description: add Linux 2.4 support
1366 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1370 Description: a large application tries to do I/O to the same resource and dies
1371 in the middle of it.
1372 Details : Check the req->rq_arrival time after the call to
1373 ost_brw_lock_get(), but before we do anything about
1374 processing it & sending the BULK transfer request. This
1375 should help move old stale pending locks off the queue as
1376 quickly as obd_timeout.
1379 Frequency : when an incorrect nid is specified during startup
1381 Description: ptlrpc connect to non-existant node causes kernel crash
1382 Details : LNET can't be re-entered from an event callback, which
1383 happened when we expire a message after the export has been
1384 cleaned up. Instead, hand the zombie cleanup off to another
1387 Severity : enhancement
1389 Description: plain/inodebits lock performance improvement
1390 Details : Grouping plain/inodebits in granted list by their request modes
1391 and bits policy, thus improving the performance of search through
1395 Frequency : only if OST filesystem is corrupted
1397 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1398 Details : for a short time RPCs with bulk IO are in the replay list,
1399 but replay of bulk IOs is unimplemented. If the OST filesystem
1400 is corrupted due to disk cache incoherency and then replay is
1401 started it is possible to trip an assertion. Avoid putting
1402 committed RPCs into the replay list at all to avoid this issue.
1405 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1407 Description: System hang on startup
1408 Details : This bug allowed the liblustre (e.g. catamount) client to
1409 return to the app before handling all startup RPCs. This
1410 could leave the node unresponsive to lustre network traffic
1411 and manifested as a server ptllnd timeout.
1413 Severity : enhancement
1415 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1416 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1417 boolean to control whether to print peer debug info when a
1418 client's RPC times out.
1421 Frequency : only for kernels with patches from Lustre below 1.4.3
1423 Description: Remove old rdonly API
1424 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1427 Frequency : only for devices with external journals
1429 Description: Set external device read-only also
1430 Details : During a commanded failover stop, we set the disk device
1431 read-only while the server shuts down. We now also set any
1432 external journal device read-only at the same time.
1435 Frequency : when upgrading from 1.4 while trying to change parameters
1437 Description: The wrong (new) MDC name was used when setting parameters for
1438 upgraded MDT's. Also allows changing of OSC (and MDC)
1439 parameters if --writeconf is specified at tunefs upgrade time.
1442 Frequency : when setting specific ost indicies
1444 Description: QOS code breaks on skipped indicies
1445 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1446 created with --index need not be sequential.
1448 Severity : enhancement
1450 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1451 Details : The uninit_groups feature works in conjunction with the kernel
1452 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1453 up the pass1 processing of e2fsck. This is a read-only feature
1454 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1455 mount filesystems that have had this feature enabled.
1457 Severity : enhancement
1459 Description: Improve multi-block allocation algorithm to avoid fragmentation
1460 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1461 allocation locality and avoid filesystem fragmentation.
1463 ------------------------------------------------------------------------------
1465 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1467 * Support for kernels:
1468 2.4.21-47.0.1.EL (RHEL 3)
1469 2.6.5-7.283 (SLES 9)
1470 2.6.9-42.0.10.EL (RHEL 4)
1471 2.6.12.6 vanilla (kernel.org)
1472 2.6.16.27-0.9 (SLES 10)
1473 * Recommended e2fsprogs version: 1.39.cfs5
1475 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1479 Frequency : occasional, depends on client load and configuration
1480 Bugzilla : 12181, 12203
1481 Description: data loss for recently-modified files
1483 Details : In some cases it is possible that recently written or created
1484 files may not be written to disk in a timely manner (this should
1485 normally be within 30s unless client IO load is very high).
1486 The problem appears as zero-length files or files that are a
1487 multiple of 1MB in size after a client crash or client eviction
1488 that are missing data at the end of the file.
1490 This problem is more likely to be hit on clients where files are
1491 repeatedly created and unlinked in the same directory, clients
1492 have a large amount of RAM, have many CPUs, the filesystem has
1493 many OSTs, the clients are rebooted frequently, and/or the files
1494 are not accessed by other nodes after being written.
1496 The presence of the problem can be detected by looking at
1497 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1498 smaller than the second (nr_unused) then dirty files will not
1499 be flushed automatically to disk. "sync; sleep 10" should be
1500 run several times on the node before unmounting it to update
1501 Lustre (this is also safe to run on nodes without this problem).
1503 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1504 that can cause this same problem, so customers using that kernel
1505 also need to update the kernel in addition to Lustre. In order
1506 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1508 It is normal that files written just before a client crash (less
1509 than 30s) may not yet have been flushed to disk, even for local
1513 Frequency : frequent on thin XT3 nodes
1515 Description: UUID collision on thin XT3 Linux nodes
1516 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1517 randomly, since we relied on an insufficiently-seeded PRNG.
1522 Description: OSS hangs after "All ost request buffers busy"
1523 Details : A deadlock between quota and journal operations caused OSS
1524 hangs after printing "All ost request buffers busy."
1527 Frequency : always on liblustre builds
1529 Description: Cleanup compiler warnings on liblustre
1532 Frequency : always on liblustre builds on XT3
1534 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1538 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1539 > 32000 subdirectories
1540 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1541 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1542 subdirectory count crosses 32000. This will aid e2fsck to
1543 correctly handle more than 32000 subdirectories.
1546 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1548 Description: System hang on startup
1549 Details : This bug allowed the liblustre (e.g. catamount) client to
1550 return to the app before handling all startup RPCs. This
1551 could leave the node unresponsive to lustre network traffic
1552 and manifested as a server ptllnd timeout.
1554 Severity : enhancement
1556 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1557 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1558 boolean to control whether to print peer debug info when a
1559 client's RPC times out.
1564 Description: make O_SYNC working on 2.6 kernels
1565 Details : 2.6 kernels use different method for mark pages for write,
1566 so need add a code to lustre for O_SYNC work.
1571 Description: Failure to close file and release space on NFS
1572 Details : Put inode details into lock acquired in ll_intent_file_open.
1573 Use mdc_intent_lock in ll_intent_open to properly
1574 detect all kind of errors unhandled by mdc_enqueue.
1579 Description: proc file read during shutdown sometimes raced obd removal,
1581 Details : Add lock to prevent obd access after proc file removal.
1584 Frequency : Only for files larger than 4GB on 32-bit clients.
1586 Description: improperly doing page alignment of locks
1587 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1588 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1593 Description: RPCs being resent when they shouldn't be
1594 Details : Some RPCs that should not be resent are being resent. This
1595 can cause inconsistencies in the RPC state machine. Do not
1596 resend such requests.
1599 Frequency : rare, only with NFS export
1601 Description: Crash on NFS re-export node
1602 Details : under very unusual load conditions an assertion is hit in
1603 ll_intent_file_open()
1606 Frequency : only if OST filesystem is corrupted
1608 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1609 Details : for a short time RPCs with bulk IO are in the replay list,
1610 but replay of bulk IOs is unimplemented. If the OST filesystem
1611 is corrupted due to disk cache incoherency and then replay is
1612 started it is possible to trip an assertion. Avoid putting
1613 committed RPCs into the replay list at all to avoid this issue.
1618 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1619 Details : Large single O_DIRECT read and write calls can fail to allocate
1620 a sufficiently large buffer to process the request. In case of
1621 allocation failure the allocation is retried with a smaller
1622 buffer and broken into smaller requests.
1624 Severity : enhancement
1626 Description: Add -o localflock option to simulate old noflock behaviour.
1627 Details : This will achieve local-only flock/fcntl locks coherentness.
1632 Description: versioning check is incomplete
1633 Details : Checking the version difference of client vs. server, report
1634 error if the gap is too big.
1639 Description: add support PG_writeback bit
1640 Details : add support for PG_writeback bit for Lustre, for more carefull
1641 work with page cache in 2.6 kernel. This also fix some deadlocks
1642 and remove hack for work O_SYNC with 2.6 kernel.
1644 Severity : enhancement
1646 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1647 Details : The uninit_groups feature works in conjunction with the kernel
1648 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1649 up the pass1 processing of e2fsck. This is a read-only feature
1650 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1651 mount filesystems that have had this feature enabled.
1653 Severity : enhancement
1655 Description: Improve multi-block allocation algorithm to avoid fragmentation
1656 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1657 allocation locality and avoid filesystem fragmentation.
1659 ------------------------------------------------------------------------------
1661 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1663 * Support for kernels:
1664 2.6.9-42.0.3.EL (RHEL 4)
1665 2.6.5-7.276 (SLES 9)
1666 2.4.21-47.0.1.EL (RHEL 3)
1667 2.6.12.6 vanilla (kernel.org)
1668 2.6.16.21-0.8 (SLES10)
1669 * Recommended e2fsprogs version: 1.39.cfs2-0
1671 * The backwards-compatible /proc/sys/portals symlink has been removed
1672 in this release. Before upgrading, please ensure that you change
1673 any configuration scripts or /etc/sysctl.conf files that access
1674 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1675 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1676 in advance of the upgrade on any system running Lustre 1.4.6 or
1677 newer, since /proc/sys/lnet was added in that version.
1678 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1682 Frequency : only when quota is used
1684 Description: avoid scanning export list for quota master
1685 Details : Change the algorithms to avoid scanning export list in order
1686 to improve the efficiency.
1689 Frequency : MDS failover only, very rarely
1691 Description: "went back in time" messages on mds failover
1692 Details : The greatest transno may be lost when the current operation
1693 finishes with an error (transno==0) and the client's last_rcvd
1694 record is over-written. Save the greatest transno in the
1695 mds_last_transno for this case.
1698 Frequency : always for specific kernels and striping counts
1700 Description: client may get "Matching packet too big" without ACL support
1701 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1702 when trying to access files in certain configurations. The
1703 clients should in fact be denied when mounting because they do
1704 not understand ACLs.
1707 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1709 Description: many clients connecting with IO in progress causes connect timeouts
1710 Details : Avoid synchronous journal commits to avoid delays caused by many
1711 clients connecting/disconnecting when bulk IO is in progress.
1712 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1713 OST_IO_PORTAL to avoid delays behind potentially many pending
1717 Frequency : occasionally with multiple writers to a single file
1719 Description: shared writes to file may result in wrong size reported by stat()
1720 Details : Allow growing of kms when extent lock is cancelled
1723 Frequency : always with random mmap IO to multi-striped file
1725 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1726 Details : Only if the hole is at the end of OST object so that kms is too
1727 small. Fix is to increase kms accordingly in ll_nopage.
1730 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1732 Description: writes to a missing object would leak memory on the OST
1733 Details : If there is an inconsistency between the MDS and OST filesystems,
1734 such that the MDS references an object that doesn't exist, writes
1735 to that object will leak memory due to incorrect cleanup in the
1736 error handling path, eventually running out of memory on the OST.
1741 Description: Creating too long symlink causes lustre errors
1742 Details : Check symlink and name lengths before sending requests to MDS.
1745 Frequency : only if flock is enabled (not on by default)
1747 Description: posix locks not released on fd closure on 2.6.9+
1748 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1749 kernels, this caused such locks not to be released on fd close and
1750 then assertions on fs unmount about still used locks.
1753 Frequency : MDS failover only, very rarely
1755 Description: clients may get ASSERTION(granted_lock != NULL)
1756 Details : When request was taking a long time, and a client was resending
1757 a getattr by name lock request. The were multiple lock requests
1758 with the same client lock handle and
1759 mds_getattr_name->fixup_handle_for_resent_request found one of the
1760 lock handles but later failed with ASSERTION(granted_lock != NULL).
1765 Description: handle->h_buffer_credits > 0, assertion failure
1766 Details : h_buffer_credits is zero after truncate, causing assertion
1767 failure. This patch extends the transaction or creates a new
1771 Frequency : NFS re-export or patchless client
1772 Bugzilla : 11179, 10796
1773 Description: Crash on NFS re-export node (__d_move)
1774 Details : We do not want to hash the dentry if we don't have a lock.
1775 But if this dentry is later used in d_move, we'd hit uninitialised
1776 list head d_hash, so we just do this to init d_hash field but
1777 leave dentry unhashed.
1780 Frequency : NFS re-export or patchless client
1782 Description: NFS exports has problem with symbolic link
1783 Details : lustre client didn't properly install dentry when re-exported
1784 to NFS or running patchless client.
1787 Frequency : NFS re-export or patchless client
1789 Description: Various nfs/patchless fixes.
1790 Details : fixes reuse disconected alias for lookup process - this fixes
1791 warning "find_exported_dentry: npd != pd",
1792 fix permission error with open files at nfs.
1793 fix apply umask when do revalidate.
1796 Frequency : occasional
1798 Description: Crash on NFS re-export node
1799 Details : calling clear_page() on the wrong pointer triggered oops in
1800 generic_mapping_read().
1803 Frequency : rarely, using O_DIRECT IO
1805 Description: unaligned directio crashes client with LASSERT
1806 Details : check for unaligned buffers before trying any requests.
1809 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1811 Description: stale data returned from RAID cache
1812 Details : If only a small amount of IO is done to the RAID device before
1813 reading it again it is possible to get stale data from the RAID
1814 cache instead of reading it from disk.
1817 Frequency : always for sles10 kernel
1819 Description: sles10 support
1820 Details : ll_follow_link: compile fixes and using of nd_set_link
1821 under newer kernels.
1824 Frequency : depends on arch, kernel and compiler version, always on sles10
1827 Description: recursive or deep enough symlinks cause stack overflow
1828 Details : getting rid of large stack-allocated variable in
1832 Frequency : depends on hardware
1834 Description: lustre write performance loss in the SLES10 kernel
1835 Details : the performance loss is caused by using of write barriers in the
1836 ext3 code. The SLES10 kernel turns barrier support on by
1837 default. The fix is to undo that change for ldiskfs.
1839 ------------------------------------------------------------------------------
1841 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1843 * Support for kernels:
1844 2.6.9-42.0.3EL (RHEL 4)
1845 2.6.5-7.276 (SLES 9)
1846 2.4.21-47.0.1.EL (RHEL 3)
1847 2.6.12.6 vanilla (kernel.org)
1851 Frequency : quota enabled and large files being deleted
1853 Description: releasing more than 4GB of quota at once hangs OST
1854 Details : If a user deletes more than 4GB of files on a single OST it
1855 will cause the OST to spin in an infinite loop. Release
1856 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1861 Description: statfs data retrieved from /proc may be stale or zero
1862 Details : When reading per-device statfs data from /proc, in the
1863 {kbytes,files}_{total,free,avail} files, it may appear
1864 as zero or be out of date.
1867 Frequency : systems with MD RAID1 external journal devices
1869 Description: lconf's call to blkid is confused by RAID1 journal devices
1870 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1871 of returning all block devices that match the journal UUID.
1874 Frequency : always, for aggregate stripe size over 4GB
1876 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1877 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1878 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1881 Frequency : always if "lfs find" used on a local file/directory
1883 Description: "lfs find" segfaults if used on a local file/directory
1884 Details : The case where a directory component was not specified wasn't
1885 handled correctly. Handle this properly.
1888 Frequency : always on ppc64
1890 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1891 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1892 perform endianness conversion.
1895 Frequency : rarely (truncate to non-zero file size after write under load)
1896 Bugzilla : 10730, 10687
1897 Description: Files padded with zeros to next 4K multiple
1898 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1899 it is possible that files that are truncated to a non-zero size
1900 immediately after being written are filled with zero bytes beyond
1901 the truncated size. No file data is lost.
1903 Severity : enhancement
1905 Description: Allow recovery/failover for liblustre clients.
1906 Details : liblustre clients were unaware of failover configurations until
1909 Severity : enhancement
1911 Description: user file locks should fail when not mounting with flock option
1912 Details : Set up an error-returning stub in ll_file_operations.lock field
1913 to prevent incorrect behaviour when client is mounted without
1914 flock option. Also, set up properly f_op->flock field for
1918 Frequency : always on ia64
1920 Description: "lfs df" loops on printing out MDS statfs information
1921 Details : The obd_ioctl_data was not initialized and in some systems
1922 this caused a failure during the ioctl that did not return
1923 an error. Initialize the struct and return an error on failure.
1926 Frequency : SLES 9 only
1928 Description: Error of copying files with lustre special EAs as root
1929 Details : Client side always return success for setxattr call for lustre
1930 special xattr (currently only "trusted.lov").
1933 Frequency : rarely on clusters with both ia64+i386 clients
1935 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1936 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1937 clients are doing concurrent writes to the same file it is
1938 possible that the ia64 clients may LASSERT because the OST
1939 extent locks are not PAGE_SIZE aligned. Ensure that grown
1940 locks are always aligned on the request boundary.
1943 Frequency : specific use, occasional
1945 Description: Overwriting in use executable truncates on-disk binary image
1946 Details : If one node attempts to overwrite an executable in use by
1947 another node, we now correctly return ETXTBSY instead of
1948 truncating the file.
1950 Severity : enhancement
1952 Description: Async OSC create to avoid the blocking unnecessarily.
1953 Details : If a OST has no remain object, system will block on the creating
1954 when need to create a new object on this OST. Now, ways use
1955 pre-created objects when available, instead of blocking on an
1956 empty osc while others are not empty. If we must block, we block
1957 for the shortest possible period of time.
1962 Description: chmod on Lustre root is propagated to other clients
1963 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1964 invalid by the follow_mount time.
1967 Frequency : liblustre clients only
1969 Description: Race in 'instant cancel' lock handling could lead to such locks
1970 never to be granted in case of SMP MDS
1971 Details : Do not destroy not yet granted but cbpending locks in
1975 Frequency : replay/resend of open
1977 Description: non null lock assetion failure in mds_intent_policy
1978 Details : Trying to replay/resend lockless open requests resulted in
1979 mds_open() returning 0 with no lock. Now it sets a flag if
1980 a lock is going to be returned.
1982 Severity : enhancement
1984 Description: Checksum enhancements
1985 Details : New checksum enhancements allow for resending RPCs that failed
1988 Severity : enhancement
1990 Description: Tunables on number of dirty pages in cacche
1991 Details : Allow to set limit on number of dirty pages cached.
1996 Description: client crash on unmount - lock still has references
1997 Details : In some error handling cases it was possible to leak a lock
1998 reference on a client while accessing a file. This was not
1999 harmful to the client during operation, but would cause the
2000 client to crash when the filesystem is unmounted.
2003 Frequency : specific case, rare
2005 Description: ETXTBSY on mds though file not in use by client
2006 Details : ETXTBSY is no longer incorrectly returned when attempting to
2007 chmod or chown a directory that the user previously tried to
2008 execute or a currently-executing binary.
2011 Frequency : extremely rare except on liblustre-based clients
2013 Description: Lustre space not freed when files are deleted
2014 Details : Clean up open-unlinked files after client eviction. Previously
2015 the unlink was skipped and the files remained as orphans.
2020 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2021 Details : In some cases it was possible to send a blocking callback to
2022 a client doing a glimpse, even though that client didn't get
2023 a lock granted. When the glimpse lock is cancelled on the OST
2024 the freed lock is left on the waiting list and corrupted the list.
2027 Frequency : all core dumps
2029 Description: Broke core dumps to lustre
2030 Details : Negative dentry may be unhashed if parent does not have UPDATE
2031 lock, but some callers, e.g. do_coredump, expect dentry to be
2032 hashed after successful create, hash it in ll_create_it.
2034 ------------------------------------------------------------------------------
2036 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2038 * Support for kernels:
2039 2.6.9-42.0.2.EL (RHEL 4)
2040 2.6.5-7.276 (SLES 9)
2041 2.4.21-40.EL (RHEL 3)
2042 2.6.12.6 vanilla (kernel.org)
2046 Frequency : always on RHEL 3
2048 Description: Number of open files grows over time
2049 Details : The number of open files grows over time, whether or not
2050 Lustre is started. This was due to a filp leak introduced
2051 by one of our kernel patches.
2053 ------------------------------------------------------------------------------
2055 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2057 * Support for kernels:
2058 2.6.9-42.EL (RHEL 4)
2059 2.6.5-7.276 (SLES 9)
2060 2.4.21-40.EL (RHEL 3)
2061 2.6.12.6 vanilla (kernel.org)
2066 Bugzilla : 5719, 9635, 9792, 9684
2067 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2068 Details : If a server is under heavy load and cannot reply to new
2069 connection requests before the client resends the (re)connect,
2070 the connection handling code can behave badly if two service
2071 threads are concurrently handing separate (re)connections from
2072 the same client. Add better locking to the connection handling
2073 code, and ensure that only a single connection will be processed
2074 for a given client UUID, even if the lock is dropped.
2076 Severity : enhancement
2078 Description: add TCP zero-copy support to kernel
2079 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2080 sends if the hardware supports scatter-gather and checksumming.
2081 This allows socklnd to do client-write and server-read more
2082 efficiently and reduce CPU utilization from skbuf copying.
2085 Frequency : only if NFS exporting from client
2087 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2088 Details : If setting attributes on a file created under NFS that had
2089 never been opened it would be possible to oops the client
2090 if the file had no objects.
2093 Frequency : always for liblustre
2095 Description: liblustre client does MDS+OSTs setattr RPC for each write
2096 Details : When doing a write from a liblustre client, the client
2097 incorrectly issued an RPC to the MDS and each OST the file was
2098 striped over in order to update the timestamps. When writing
2099 with small chunks and many clients this could overwhelm the MDS
2100 with RPCs. In all cases it would slow down the write because
2101 these RPCs are unnecessary.
2103 Severity : enhancement
2105 Description: allow number of MDS service threads to be changed at module load
2106 Details : It is now possible to change the number of MDS service threads
2107 running. Adding "options mds mds_num_threads={N}" to the MDS's
2108 /etc/modprobe.conf will set the number of threads for the next
2109 time Lustre is restarted (assuming the "mds" module is also
2110 reloaded at that time). The default number of threads will
2111 stay the same, 32 for most systems.
2116 Description: OST crash if filesystem is unformatted or corrupt
2117 Details : If an OST is started on a device that has never been formatted
2118 or if the filesystem is corrupt and cannot even mount then the
2119 error handling cleanup routines would dereference a NULL pointer.
2124 Description: NULL pointer deref in llap_from_page.
2125 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2126 incorrect) mapping if the page was truncated/reclaimed while it was
2127 searched for. Check for this condition and skip such pages when
2128 doing readahead. Introduce extra check to llap_from_page() to
2129 verify page->mapping->host is non-NULL (so page is not anonymous).
2132 Frequency : Sometimes when using sys_sendfile
2134 Description: "page not covered by a lock" warnings from ll_readpage
2135 Details : sendfile called ll_readpage without right page locks present.
2136 Now we introduced ll_file_sendfile that does necessary locking
2137 around call to generic_file_sendfile() much like we do in
2141 Frequency : with certain MDS communication failures at client mount time
2143 Description: NULL pointer deref after failed client mount
2144 Details : a client connection request may delayed by the network layer
2145 and not be sent until after the PTLRPC layer has timed out the
2146 request. If the client fails the mount immediately it will try
2147 to clean up before the network times out the request. Add a
2148 reference from the request import to the obd device and delay
2149 the cleanup until the network drops the request.
2152 Frequency : occasionally during client (re)connect
2154 Description: assertion failure during client (re)connect
2155 Details : processing a client connection request may be delayed by the
2156 client or server longer than the client connect timeout. This
2157 causes the client to resend the connection request. If the
2158 original connection request is replied in this interval, the
2159 client may trip an assertion failure in ptlrpc_connect_interpret()
2160 which thought it would be the only running connect process.
2163 Frequency : only with obd_echo servers and clients that are rebooted
2165 Description: kernel BUG accessing uninitialized data structure
2166 Details : When running an obd_echo server it did not start the ping_evictor
2167 thread, and when a client was evicted an uninitialized data
2168 structure was accessed. Start the ping_evictor in the RPC
2169 service startup instead of the OBD startup.
2171 Severity : enhancement
2172 Bugzilla : 10193 (patchless)
2173 Description: Remove dependency on various unexported kernel interfaces.
2174 Details : No longer need reparent_to_init, exit_mm, exit_files,
2175 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2178 Frequency : rare (only users of deprecated and unsupported LDAP config)
2180 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2181 Details : LDAP apparently contains 'lustreName' attributes instead of
2182 'name'. A simple remapping of the name is sufficient.
2185 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2187 Description: waiting_locks_callback trips kernel BUG if client is evicted
2188 Details : Running with the dump_on_timeout debug flag turned on makes
2189 it possible that the waiting_locks_callback() can try to dump
2190 the Lustre kernel debug logs from an interrupt handler. Defer
2191 this log dumping to the expired_lock_main() thread.
2193 Severity : enhancement
2195 Description: Support NFS exporting on 2.6 kernels.
2196 Details : Implement non-rawops metadata methods for NFS server to use without
2197 changing NFS server code.
2200 Frequency : very rare (synthetic metadata workload only)
2202 Description: two racing renames might cause an MDS thread to deadlock
2203 Details : Running the "racer" program may cause one MDS thread to rename
2204 a file from being the source of a rename to being the target of
2205 a rename at exactly the same time that another thread is doing
2206 so, and the second thread has already enqueued these locks after
2207 doing a lookup of the target and is trying to relock them in
2208 order. Ensure that we don't try to re-lock the same resource.
2211 Frequency : only very large systems with liblustre clients
2213 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2214 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2215 all OSTs in parallel. This allows the eviction of stale liblustre
2216 clients to proceed much faster than if they were done in series,
2217 and also offers similar improvements for other set_info RPCs.
2222 Description: excessive CPU usage during initial read phase on client
2223 Details : During the initial read phase on a client, it would agressively
2224 retry readahead on the file, consuming too much CPU and impacting
2225 performance (since 1.4.5.8). Improve the readahead algorithm
2226 to avoid this, and also improve some other common cases (read
2227 of small files in particular, where "small" is files smaller than
2228 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2233 Description: MDS crash when receiving packet with unknown intent.
2234 Details : Do not LBUG in unknown intent case, just return -EFAULT
2236 Severity : enhancement
2237 Bugzilla : 9293, 9385
2238 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2239 Details : Do not serialize getattr (non-intent version) and statfs.
2242 Frequency : occasional, when OST network is overloaded/intermittent
2244 Description: client evicted by OST after bulk IO timeout
2245 Details : If a client sends a bulk IO request (read or write) the OST
2246 may evict the client if it is unresposive to its data GET/PUT
2247 request. This is incorrect if the network is overloaded (takes
2248 too long to transfer the RPC data) or dropped the OST GET/PUT
2249 request. There is no need to evict the client at all, since
2250 the pinger and/or lock callbacks will handle this, and the
2251 client can restart the bulk request.
2254 Frequency : Always when mmapping file with no objects
2256 Description: client crashes when mmapping file with no objects
2257 Details : Check that we actually have objects in a file before doing any
2258 operations on objects in ll_vm_open, ll_vm_close and
2264 Description: Request leak when working with deleted CWD
2265 Details : Introduce advanced request refcount tracking for requests
2266 referenced from lustre intent.
2268 Severity : Enhancement
2270 Description: Cache open file handles on client.
2271 Details : MDS now will return special lock along with openhandle, if
2272 requested and client is allowed to hold openhandle, even if unused,
2273 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2274 closing files for every read/write openration.
2276 Severity : Enhancement
2278 Description: Cache open negative dentries on client when possible.
2279 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2280 negative dentries on lock revocation.
2285 Description: Remounting a client read-only wasn't possible with a zconf mount
2286 Details : It wasn't possible to remount a client read-only with llmount.
2288 Severity : enhancement
2289 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2290 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2291 MPICH is included to add Lustre-specific ADIO interfaces.
2292 This is based closely on the UFS ADIO layer and only differs
2293 in file creation, in order to allow the OST striping to be set.
2294 This is user-contributed code and not supported by CFS.
2299 Description: extended inode attributes (immutable, append-only) work improperly
2300 when 2.4 and 2.6 kernels are used on client/server or vice versa
2301 Details : Introduce kernel-independent values for these flags.
2303 Severity : enhancement
2306 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2307 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2308 granularity. Now, Fractional MB granularity is be supported,
2309 this is very useful for low memory system.
2311 Severity : enhancement
2313 Description: Getattr by fid
2314 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2315 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2316 have locks and avoids getattr rpc requests for it on every stat.
2319 Frequency : Always, for filesystems larger than 2TB
2321 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2322 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2323 filesystems larger than 2TB to be mounted with mballoc enabled.
2326 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2328 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2329 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2330 filesystems larger than 2TB is mounted then the kernel will
2331 silently corrupt the start of the filesystem. CONFIG_LBD is
2332 enabled for all CFS-supported kernels, but the possibility of
2333 this happening with a modified kernel config exists.
2335 Severity : enhancement
2337 Description: add client O_DIRECT support for 2.6 kernels
2338 Details : It is now possible to do O_DIRECT reads and writes to files
2339 in the Lustre client mountpoint on 2.6 kernel clients.
2341 Severity : enhancement
2343 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2344 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2345 parallel, not waiting for response from every OST before sending
2346 a rpc to the next OST.
2351 Description: setattr vs write race when updating file timestamps
2352 Details : Client processes that update a file timestamp into the past
2353 right after writing to the file (e.g. tar) it is possible that
2354 the updated file modification time can be reset to the current
2355 time due to a race between processing the setattr and write RPC.
2357 Severity : enhancement
2359 Description: Bring 'lfs find' closer in line with regular Linux find.
2360 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2361 -print0 options and obtains all the needed info through the lustre
2364 Severity : enhancement
2366 Description: support up to 1024 configured devices on one node
2367 Details : change obd_dev array from statically allocated to dynamically
2368 allocated structs as they are first used to reduce memory usage
2373 Description: Flush dirty partially truncated pages during truncate
2374 Details : Immediatelly flush partially truncated pages in filter_setattr,
2375 this way we completely avoid having any pages in page cache on OST
2376 and can retire ugly workarounds during writes to flush such pages.
2381 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2382 Details : protect inode from truncation within vfs_unlink() context
2383 just take a reference before calling vfs_unlink() and release it
2384 when parent's i_sem is free.
2389 Description: last_id value checked outside lock on OST caused LASSERT failure
2390 Details : If there were multiple MDS->OST object precreate requests in
2391 flight, it was possible that the OST's last object id was checked
2392 outside a lock and incorrectly tripped an assertion. Move checks
2393 inside locks, and discard old precreate requests.
2396 Frequency : always, if extents are used on OSTs
2398 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2399 Details : OSTs using the extents format would not zero the high 16 bits of
2400 the index physical block number. This is not a problem for any
2401 OST filesystems smaller than 16TB, and no kernels support ext3
2402 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2403 new/modified files) and can be fixed for existing filesystems
2404 with e2fsprogs-1.39-cfs1.
2409 Description: import connection selection may be incorrect if timer wraps
2410 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2411 import connections to be ignored if the 32-bit jiffies counter
2412 wraps. Use a 64-bit jiffies counter.
2415 Frequency : very large clusters immediately after boot
2417 Description: LNET request buffers exhausted under heavy short-term load
2418 Details : If a large number of client requests are generated on a service
2419 that has previously never seen so many requests it is possible
2420 that the request buffer growth cannot keep up with the spike in
2421 demand. Instead of dropping incoming requests, they are held in
2422 the LND until the RPC service can accept more requests.
2425 Frequency : Sometimes during replay
2427 Description: Assertion failure in ll_local_open after replay.
2428 Details : If replay happened on an open request reply before we were able
2429 to set replay handler, reply will become not swabbed tripping the
2430 assertion in ll_local_open. Now we set the handler right after
2431 recognising of open request
2434 Frequency : very rare
2436 Description: kernel reports "badness in vsnprintf"
2437 Details : Reading from the "recovery_status" /proc file in small chunks
2438 may cause a negative length in lprocfs_obd_rd_recovery_status()
2439 call to vsnprintf() (which is otherwise harmless). Exit early
2440 if there is no more space in the output buffer.
2442 Severity : enhancement
2444 Description: clear OBD RPC statistics by writing to them
2445 Details : It is now possible to clear the OBD RPC statistics by writing
2446 to the "stats" file.
2451 Description: Client mtime is not the same on different clients after utimes
2452 Details : In some cases, the client was using the utimes() syscall on
2453 a file cached on another node. The clients now validate the
2454 ctime from the MDS + OSTs to determine which one is right.
2459 Description: Inability to activate failout mode
2460 Details : lconf script incorrectly assumed that in pythong string's numeric
2461 value is used in comparisons.
2464 Frequency : always with multiple stripes per file
2466 Description: Inefficient object allocation for mutli-stripe files
2467 Details : When selecting which OSTs to stripe files over, for files with
2468 a stripe count that divides evenly into the number of OSTs,
2469 the MDS is always picking the same starting OST for each file.
2470 Return the OST selection heuristic to the original design.
2475 Description: mount failures may take full timeout to return an error
2476 Details : Under some heavy load conditions it is possible that a
2477 failed mount can wait for the full obd_timeout interval,
2478 possibly several minutes, before reporting an error.
2479 Instead return an error as soon as the status is known.
2481 Frequency : quota enabled and large files being deleted
2483 Description: releasing more than 4GB of quota at once hangs OST
2484 Details : If a user deletes more than 4GB of files on a single OST it
2485 will cause the OST to spin in an infinite loop. Release
2486 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2491 Description: statfs data retrieved from /proc may be stale or zero
2492 Details : When reading per-device statfs data from /proc, in the
2493 {kbytes,files}_{total,free,avail} files, it may appear
2494 as zero or be out of date.
2497 Frequency : systems with MD RAID1 external journal devices
2499 Description: lconf's call to blkid is confused by RAID1 journal devices
2500 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2501 of returning all block devices that match the journal UUID.
2504 Frequency : always, for aggregate stripe size over 4GB
2506 Description: assertion fails when trying to use 4GB stripe size
2507 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2508 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2511 Frequency : always on ppc64
2513 Description: the first write on an ext3 filesystem with mballoc got stuck
2514 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2515 perform endianness conversion.
2517 ------------------------------------------------------------------------------
2519 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2521 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2522 INTEROPERATE with older versions automatically. Please read the
2523 user documentation before upgrading any part of a live system.
2524 * WARNING: Lustre networking configuration changes are required with
2525 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2528 * Support for kernels:
2529 2.6.9-22.0.2.EL (RHEL 4)
2530 2.6.5-7.244 (SLES 9)
2531 2.6.12.6 vanilla (kernel.org)
2534 Severity : enhancement
2535 Bugzilla : 7981/8208
2536 Description: Introduced Lustre Networking (LNET)
2537 Details : LNET is new networking infrastructure for Lustre, it includes
2538 a reorganized network configuration mode (see the user
2539 documentation for full details) as well as support for routing
2540 between different network fabrics. Lustre Networking Devices
2541 (LNDs) for the supported network fabrics have also been
2542 created for this new infrastructure.
2544 Severity : enhancement
2545 Description: Introduced Access control lists
2546 Details : clients can set ACLs on files and directories in order to have
2547 more fine-grained permissions than the standard Unix UGO+RWX.
2548 The MDS must be started with the "-o acl" mount option.
2550 Severity : enhancement
2551 Description: Introduced filesystem quotas
2552 Details : Administrators may now establish per-user quotas on the
2555 Severity : enhancement
2557 Description: Configuration change for the XT3
2558 The PTLLND is now used to run Lustre over Portals on the XT3
2559 The configure option(s) --with-cray-portals are no longer used.
2560 Rather --with-portals=<path-to-portals-includes> is used to
2561 enable building on the XT3. In addition to enable XT3 specific
2562 features the option --enable-cray-xt3 must be used.
2567 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2568 Details : A race between allocating a new llcd and re-getting the llcd_lock
2569 allowed another thread to grab newly-allocated llcd.
2571 Severity : enhancement
2573 Description: 2.6 OST async journal commit and locking fix to improve performance
2574 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2575 2.6 kernels are now async as they already were in 2.4 kernels so
2576 that they can commit concurrently with the network bulk transfer.
2577 For block-allocated files the filter allocation semaphore is held
2578 to avoid filesystem fragmentation during allocation. BKL lock
2579 removed for 2.6 xattr operations where it is no longer needed.
2584 Description: lconf incorrectly determined whether two IP networks could talk
2585 Details : In some more complicated routing and multiple-network
2586 configurations, lconf will avoid trying to make a network
2587 connection to a disjoint part of the IP space. It was doing the
2588 math incorrectly for one set of cases.
2593 Description: Fix for potential infinite loop processing records in an llog.
2594 Details : If an llog record is corrupted/zeroed, it is possible to loop
2595 forever in llog_process(). Validate the llog record length
2596 and skip the remainder of the block on error.
2599 Frequency : occasional (liblustre only)
2601 Description: liblustre could not open files whose last component is a symlink
2602 Details : sysio_path_walk() would incorrectly pass the open intent to
2603 intermediate path components.
2606 Frequency : rare (liblustre only with non-standard tuning)
2607 Bugzilla : 7201 (7350)
2608 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2609 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2610 after it has been released, only do this for referenced locks
2612 Severity : enhancement
2614 Description: specifying an (invalid) directory default stripe_size of -1
2615 would reset the directory default striping
2616 Details : stripe_size -1 was used internally to signal directory stripe
2617 removal, now use "all default" to signal dir stripe removal
2618 as a directory striping of "all default" is not useful
2621 Frequency : common for large clusters running liblustre clients
2623 Description: doing an ls when liblustre clients are running is slow
2624 Details : sending a glimpse AST to a liblustre client waits for every AST
2625 to time out, as liblustre clients will not respond. Since they
2626 cannot cache data we refresh the OST lock LVB from disk instead.
2628 Severity : enhancement
2630 Description: doing an ls at the same time as file IO can be slow
2631 Details : enqueue and other "small" requests can be blocked behind many
2632 large IO requests. Create a new OST IO portal for non-IO
2633 requests so they can be processed faster.
2636 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2638 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2639 created file with mode 000
2640 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2641 allow an ftruncate() as a non-root user to a file with mode 000.
2642 Lustre now respects this flag to disable mode checks when
2643 truncating a file owned by the user
2646 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2649 Description: Revoking locks from clients that went dead or catatonic might take
2651 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2652 cancellation of such locks instant on servers without waiting for
2653 any reply from clients. Clients drops these locks when cancel
2654 notification from server is received without replying.
2657 Frequency : liblustre-only, when liblustre client dies or becomes busy
2659 Description: Doing ls on Linux clients can take a long time with active
2661 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2662 granting such locks to it in the first place if possible. Locks
2663 are taken by proxy on the OST during the read or write and
2664 dropped immediately afterward. Add connect flags handling, do
2665 not grant locks to liblustre clients for glimpse ASTs.
2667 Severity : enhancement
2669 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2670 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2671 sequential reads and behaves badly with applications doing random
2672 reads. Improve it by reading ahead at least read region, and
2673 avoiding excessive large RPC for small reads.
2675 Severity : enhancement
2677 Description: Creating more than 1000 files for a single job may cause a load
2678 imbalance on the OSTs if there are also a large number of OSTs.
2679 Details : qos_prep_create() uses an OST index reseed value that is an
2680 even multiple of the number of available OSTs so that if the
2681 reseed happens in the middle of the object allocation it will
2682 still utilize the OSTs as uniformly as possible.
2687 Description: OST or MDS may oops in ping_evictor_main()
2688 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2689 but doesn't restart at beginning of obd_exports_timed list
2692 Severity : enhancement
2694 Description: improve by-nid export eviction on the MDS and OST
2695 Details : allow multiple exports with the same NID to be evicted at one
2696 time without re-searching the exports list.
2699 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2701 Description: MDS may oops in groups_free()
2702 Details : in rare race conditions a newly allocated group_info struct is
2703 freed again, and this can be NULL. The 2.4 compatibility code
2704 for groups_free() checked for a NULL pointer, but 2.6 did not.
2707 Frequency : common for liblustre clients doing little filesystem IO
2708 Bugzilla : 9352, 7313
2709 Description: server may evict liblustre clients accessing contended locks
2710 Details : if a client is granted a lock or receives a completion AST
2711 with a blocking AST already set it would not reply to the AST
2712 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2715 Frequency : lfs setstripe, only systems with more than 160 OSTs
2717 Description: unable to set striping with a starting offset beyond OST 160
2718 Details : llapi_create_file() incorrectly limited the starting stripe
2719 index to the maximum single-file stripe count.
2722 Frequency : LDAP users only
2724 Description: lconf did not handle in-kernel recovery with LDAP properly
2725 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2727 Severity : enhancement
2729 Description: bind OST threads to NUMA nodes to improve performance
2730 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2731 node and do their allocations there to localize memory access
2733 Severity : enhancement
2735 Description: llmount can determine client NID directly from Myrinet (GM)
2736 Details : the client NID code from gmnalnid was moved directly into
2737 llmount, removing the need to use this or specifying the
2738 client NID explicitly when mounting GM clients with zeroconf
2741 Frequency : if client is started with down MDS
2743 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2744 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2745 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2746 ptlrpc_import_delayed_req() because we didn't check if the
2747 request was interrupted, and we also didn't break out of the
2748 event loop if there was no timeout
2753 Description: data loss during non-page-aligned writes to a single file from
2754 both multiple nodes and multiple threads on one node at same time
2755 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2756 inconsistency led to false short-reads, that were cached and later
2757 used by ->prepare_write() to fill in partially written page,
2758 leading to data loss.
2761 Frequency : always, if lconf --abort_recovery used
2763 Description: lconf --abort_recovery fails with 'Operation not supported'
2764 Details : lconf was attempting to abort recovery on the MDT device and not
2767 Severity : enhancement
2769 Description: remove cleanup logs
2770 Details : replace lconf-generated cleanup logs with lustre internal
2771 cleanup routines. Eliminates the need for client-cleanup and
2774 Severity : enhancement
2776 Description: add support for EAs (user and system) on lustre filesystems
2777 Details : it is now possible to store extended attributes in the Lustre
2778 client filesystem, and with the user_xattr mount option it
2779 is possible to allow users to store EAs on their files also
2781 Severity : enhancement
2783 Description: Add possibility (config option) to show minimal available OST free
2785 Details : When compiled with --enable-mindf configure option, statfs(2)
2786 (and so, df) will return least minimal free space available from
2787 all OSTs as amount of free space on FS, instead of summary of
2788 free spaces of all OSTs.
2790 Severity : enhancement
2792 Description: do not expand extent locks acquired on OST-side
2793 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2794 by server: they are not cached anyway.
2797 Frequency : when mmap is used/binaries executed from Lustre
2799 Description: Unmmap pages before throwing them away from read cache.
2800 Details : llap_shrink cache now attempts to unmap pages before discarding
2801 them (if unmapping failed - do not discard). SLES9 kernel has
2802 extra checks that trigger if this unmapping is not done first.
2807 Description: lconf didn't resolve symlinks before checking to see whether a
2808 given mountpoint was already in use
2811 Frequency : when migrating failover services
2812 Bugzilla : 6395, 9514
2813 Description: When migrating a subset of services from a node (e.g. failback
2814 from a failover service node) the remaining services would
2815 time out and evict clients.
2816 Details : lconf --force (implied by --failover) sets the global obd_timeout
2817 to 5 seconds in order to quickly disconnect, but this caused
2818 other RPCs to time out too quickly. Do not change the global
2819 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2821 Severity : enhancement
2822 Frequency : if MDS is started with down OST
2823 Bugzilla : 9439,5706
2824 Description: Allow startup/shutdown of an MDS without depending on the
2825 availability of the OSTs.
2826 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2827 Add appropriate locking and lov-osc refcounts for safe
2828 cleaning. Add osc abort_inflight calls in case the
2829 synchronize never started.
2832 Frequency : occasional (Cray XT3 only)
2834 Description: root not authorized to access files in CRAY_PORTALS environment
2835 Details : The client process capabilities were not honoured on the MDS in
2836 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2837 been accessed by an authorized user then root was able to access
2838 the file on the local client also. The root user capabilities
2839 are now allowed on the MDS, as this environment has secure UID.
2842 Frequency : occasional
2844 Description: ldiskfs "too long searching" message happens too often
2845 Details : A debugging message (otherwise harmless) prints too often on
2846 the OST console. This has been reduced to only happen when
2847 there are fragmentation problems on the filesystem.
2852 Description: Division by zero in statfs when all OSCs are inactive
2853 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2854 lov_statfs divides by value returned by lov_get_stripecnt().
2858 Bugzilla : 9489, 3273
2859 Description: First write from each client to each OST was only 4kB in size,
2860 to initialize client writeback cache, which caused sub-optimal
2861 RPCs and poor layout on disk for the first writen file.
2862 Details : Clients now request an initial cache grant at (re)connect time
2863 and so that they can start streaming writes to the cache right
2864 away and always do full-sized RPCs if there is enough data.
2865 If the OST is rebooted the client also re-establishes its grant
2866 so that client cached writes will be honoured under the grant.
2871 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2872 Details : Now I/O RPCs go to different portal number and (presumably) fast
2873 lock requests (and glimses) and other RPCs get their own service
2874 threads pool that should be able to service those RPCs
2877 Severity : enhancement
2879 Description: Ability to exchange lustre version between client and servers and
2880 issue warnings at client side if client is too old. Also for
2881 liblustre clients there is ability to refuse connection of too old
2883 Details : New 'version' field is added to connect data structure that is
2884 filled with version info. That info is later checked by server and
2888 Frequency : rare, liblustre only.
2889 Bugzilla : 9296, 9581
2890 Description: Two simultaneous writes from liblustre at offset within same page
2891 might proceed at the same time overwriting eachother with stale
2893 Details : I/O lock withing llu_file_prwv was released too early, before data
2894 actually was hitting the wire. Extended lock-holding time until
2895 server acknowledges receiving data.
2898 Frequency : extremely rare. Never observed in practice.
2900 Description: avoid generating lustre_handle cookie of 0.
2901 Details : class_handle_hash() generates handle cookies by incrementing
2902 global counter, and can hit 0 occasionaly (this is unlikely, but
2903 not impossible, because initial value of cookie counter is
2904 selected randonly). Value of 0 is used as a sentinel meaning
2905 "unassigned handle" --- avoid it. Also coalesce two critical
2906 sections in this function into one.
2908 Severity : enhancement
2910 Description: allow liblustre clients to delegate truncate locking to OST
2911 Details : To avoid overhead of locking, liblustre client instructs OST to
2912 take extent lock in ost_punch() on client's behalf. New connection
2913 flag is added to handle backward compatibility.
2915 Severity : enhancement
2916 Bugzilla : 4928, 7341, 9758
2917 Description: allow number of OST service threads to be specified
2918 Details : a module parameter allows the number of OST service threads
2919 to be specified via "options ost ost_num_threads={N}" in the
2920 OSS's /etc/modules.conf or /etc/modprobe.conf.
2924 Bugzilla : 6146, 9635, 9895
2925 Description: servers crash with bad pointer in target_handle_connect()
2926 Details : In rare cases when a client is reconnecting it was possible that
2927 the connection request was the last reference for that export.
2928 We would temporarily drop the export reference and get a new
2929 one, but this may have been the last reference and the export
2930 was just destroyed. Get new reference before dropping old one.
2932 Severity : enhancement
2933 Frequency : if client is started with failover MDS
2935 Description: Allow multiple MDS hostnames in the mount command
2936 Details : Try to read the configuration from all specified MDS
2937 hostnames during a client mount in case the "primary"
2940 Severity : enhancement
2942 Description: Stop sending data to evicted clients as soon as possible.
2943 Details : Check if the client we are about to send or are sending data to
2944 was evicted already. (Check is done every second of waiting,
2945 for which l_wait_event interface was extended to allow checking
2946 of exit condition at specified intervals).
2949 Frequency : rare, normally only when NFS exporting is done from client
2951 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2953 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2956 Frequency : always, if srand() is called before liblustre initialization
2958 Description: Liblustre uses system PRNG disturbing its usage by user application
2959 Details : Introduce internal to lustre fast and high-quality PRNG for
2960 lustre usage and make liblustre and some other places in generic
2961 lustre code to use it.
2963 Severity : enhancement
2964 Bugzilla : 9477, 9557, 9870
2965 Description: Verify that the MDS configuration logs are updated when xml is
2966 Details : Check if the .xml configuration logs are newer than the config
2967 logs stored on the MDS and report an error if this is the case.
2968 Request --write-conf, or allow starting with --old_conf.
2970 Severity : enhancement
2972 Description: Handle symlinks in the path when checking if Lustre is mounted.
2973 Details : Resolve intermediate symlinks when checking if a client has
2974 mounted a filesystem to avoid duplicate client mounts.
2979 Description: lconf can hit an error exception but still return success.
2980 Details : The lconf command catches the Command error exception at the top
2981 level script context and will exit with the associated exit
2982 status, but doesn't ensure that this exit status is non-zero.
2987 Description: failure of ptlrpc thread startup can cause oops
2988 Details : Starting a ptlrpc service thread can fail if there are a large
2989 number of threads or the server memory is very fragmented.
2990 Handle this without oopsing.
2993 Frequency : always, only if liblustre and non-default acceptor port was used
2995 Description: liblustre cannot connect to servers with non-default acceptor port
2996 Details : tcpnal_set_default_params() was not called and was therefore
2997 ignoring the environment varaible TCPNAL_PORT, as well as other
2998 TCPNAL_ environment variables
3003 Description: two objects could be created on the same OST for a single file
3004 Details : If an OST is down, in some cases it was possible to create two
3005 objects on a single OST for a single file. No problems other
3006 than potential performance impact and spurious error messages.
3010 Bugzilla : 5681, 9562
3011 Description: Client may oops in ll_unhash_aliases
3012 Details : Client dcache may become inconsistent in race condition.
3013 In some cases "getcwd" can fail if the current directory is
3019 Description: Inode refcounting problems in NFS export code
3020 Details : link_raw functions used to call d_instantiate without obtaining
3021 extra inode reference first.
3025 Bugzilla : 9942, 9903
3026 Description: Referencing freed requests leading to crash, memleaks with NFS.
3027 Details : We used to require that call to ll_revalidate_it was always
3028 followed by ll_lookup_it. Also with revalidate_special() it is
3029 possible to call ll_revalidate_it() twice for the same dentry
3030 even if first occurence returned success. This fix changes semantic
3031 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3032 reference on a request referred from the intent.
3033 ll_intent_release() then releases such a request.
3036 Frequency : rare, normally benchmark loads only
3038 Description: unlinked inodes were kept in memory on the client
3039 Details : If a client is repeatedly creating and unlinking files it
3040 can accumulate a lot of stale inodes in the inode slab cache.
3041 If there is no other client load running this can cause the
3042 client node to run out of memory. Instead flush old inodes
3043 from client cache that have the same inode number as a new inode.
3046 Frequency : SLES9 2.6.5 kernel and long filenames only
3047 Bugzilla : 9969, 10379
3048 Description: utime reports stale NFS file handle
3049 Details : SLES9 uses out-of-dentry names in some cases, which confused
3050 the lustre dentry revalidation. Change it to always use the
3054 Frequency : rare, unless heavy write-truncate concurrency is continuous
3055 Bugzilla : 4180, 6984, 7171, 9963, 9331
3056 Description: OST becomes very slow and/or deadlocked during object unlink
3057 Details : filter_destroy() was holding onto the parent directory lock
3058 while truncating+unlinking objects. For very large objects this
3059 may block other threads for a long time and slow overall OST
3060 responsiveness. It may also be possible to get a lock ordering
3061 deadlock in this case, or run out of journal credits because of
3062 the combined truncate+unlink. Solution is to do object truncate
3063 first in one transaction without parent lock, and then do the
3064 final unlink in a new transaction with the parent lock. This
3065 reduces the lock hold time dramatically.
3068 Frequency : rare, 2.4 kernels only
3070 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3071 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3072 before unmounting their filesystem in order to avoid deadlock.
3073 The kernel_locked() function in 2.4 kernels only checks whether
3074 the kernel lock is held, not whether it is this process that is
3075 holding it as 2.6 kernels do.
3080 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3081 Details : The client ptlrpc code may be trying to reconnect to a down
3082 server before a previous connection attempt has timed out.
3083 Increase the reconnect interval to be longer than the connection
3084 timeout interval to avoid sending duplicate connections to
3088 Frequency : echo_client brw_test command
3090 Description: fix echo_client to work with OST preallocated code
3091 Details : OST preallocation code (5137) didn't take echo_client IO path
3092 into account: echo_client calls filter methods outside of any
3093 OST thread and, hence, there is no per-thread preallocated
3094 pages and buffers to use. Solution: hijack pga pages for IO. As
3095 a byproduct, this avoids unnecessary data copying.
3099 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3100 Description: Client can oops in mdc_commit_close() after open replay
3101 Details : It was possible for the MDS to return an open request with no
3102 transaction number in mds_finish_transno() if the client was
3103 evicted, but without actually returning an error. Clients
3104 would later try to replay that open and may trip an assertion
3105 Simplify the client close codepath, and always return an error
3106 from the MDS in case the open is not successful.
3109 Frequency : rare, 2.6 OSTs only
3111 Description: OST may deadlock under high load on fragmented files
3112 Details : If there was a heavy load and highly-fragmented OST filesystems
3113 it was possible to have all the OST threads deadlock waiting on
3114 allocation of biovecs, because the biovecs were not released
3115 until the entire RPC IO was completed. Instead, release biovecs
3116 as soon as they are complete to ensure forward IO progress.
3118 Severity : enhancement
3120 Description: Support for specifying external journal device at mount
3121 Details : If an OST or MDS device is formatted with an external journal
3122 device, this device major/minor is stored in the ext3 superblock
3123 and may not be valid for failover. Allow detecting and
3124 specifying the external journal at mount time.
3129 Description: Mounting an MDS with pending unlinked files may cause oops
3130 Details : target_finish_recovery() calls mds_postrecov() which returned
3131 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3132 considers this an error and immediately begins cleaning up the
3133 lov, just after starting the mds_lov process
3135 Severity : enhancement
3137 Description: Implement 'lfs df' to report actual free space on per-OST basis
3138 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3139 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3140 usage of objects; '-h' to report in human readable format.
3142 ------------------------------------------------------------------------------
3144 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
3151 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3152 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3153 at startup during orphan unlinking. Instead, skip buddy update
3154 before setup, as it will be regenerated after recovery is complete.
3159 Description: If an OST is inactive, its locks might reference stale inodes.
3160 Details : lov_change_cbdata() must iterate over all namespaces, even if
3161 they are inactive to clear inode references from the lock.
3163 Severity : enhancement
3164 Frequency : occasional, if non-standard max_dirty_mb used
3166 Description: Client will block write RPCs if not enough grant
3167 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3168 then the client will block writes while waiting for another RPC
3169 to complete instead of consuming its dirty limit. With change
3170 we get improved performance when max_dirty_mb is small.
3172 Severity : enhancement
3173 Bugzilla : 3389, 6253
3174 Description: Add support for supplementary groups on the MDS.
3175 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3176 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3177 lookups for user supplementary groups into a cache.
3181 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3182 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3183 for liblustre O_EXCL is enforced by MDS.
3187 Description: readdir never returns NULL in liblustre.
3188 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3189 dentry in struct dirent.
3193 Frequency : liblustre only, depends on application IO pattern
3194 Description: liblustre clients evicted if not contacting servers
3195 Details : Don't put liblustre clients into the ping_evictor list, so
3196 they will not be evicted by the pinger ever.
3198 Severity : enhancement
3200 Description: Add ability to evict clients by NID from MDS.
3201 Details : By echoing "nid:$NID" string into
3202 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3203 $NID would be instantly evicted from this MDS and from all active
3204 OSTs connected to it.
3208 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3209 Details : lookup_it_finish() used to query file size from OSTs that was not
3214 Description: service threads change working directory to that of init
3215 Details : Starting lustre service threads may pin the working directory
3216 of the parent thread, making that filesystem busy. Threads
3217 now change to the working directory of init to avoid this.
3221 Frequency : during shutdown only
3222 Description: shutdown with a failed MDS or OST can cause unmount to hang
3223 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3228 Frequency : 2.6 only
3229 Description: chmod/chown may include an extra supplementary group
3230 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3231 supplementary group and if none is specified this is used.
3234 Bugzilla : 5479 (6816)
3236 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3237 Details : If lookup is in progress on a file that is unlinked we might try
3238 to revalidate the inode and fail in revalidate after lookup is
3239 complete and ll_file_open() enqueues the open again but
3240 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3243 Frequency : always, if lconf --abort_recovery used
3245 Description: lconf --abort_recovery fails with 'Operation not supported'
3246 Details : lconf was attempting to abort recovery on the MDT device and not
3249 ------------------------------------------------------------------------------
3251 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3256 Frequency : rare (only unsupported configurations with a node running as an
3258 Bugzilla : 6514, 5137
3259 Description: Mounting a Lustre file system on a node running as an OST could
3261 Details : OSTs now preallocates memory needed to write out data at
3262 startup, instead of when needed, to avoid having to
3263 allocate memory in possibly low memory situations.
3264 Specifically, if the file system is mounted on on OST,
3265 memory pressure could force it to try to write out data,
3266 which it needed to allocate memory to do. Due to the low
3267 memory, it would be unable to do so and the node would
3268 become unresponsive.
3270 Severity : enhancement
3272 Description: Addition of lconf --service command line option
3273 Details : lconf now accepts a '--service <arg>' option, which is
3274 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3276 Severity : enhancement
3278 Description: Failover mode is now the default for OSTs.
3279 Details : By default, OSTs will now run in failover mode. To return to
3280 the old behaviour, add '--failout' to the lmc line for OSTs.
3282 Severity : enhancement
3284 Description: Health checks are now provided for MDS and OSTs
3285 Details : Additional detailed health check information on MSD and OSTs
3286 is now provided through the procfs health_check value.
3289 Frequency : occasional, depends on IO load
3291 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3292 after numerous create/delete cycles
3293 Details : The ext3 inode allocation policy would not allocate new inodes
3294 very well on the OSTs because there are no new directories
3295 being created. Instead we look for groups with free space if
3296 the parent directories are nearly full.
3301 Description: Network or server problems during mount may cause partially
3302 mounted clients instead of returning an error.
3303 Details : The config llog parsing code may overwrite the error return
3304 code during mount error handling, returning success instead
3310 Description: MDS can fail to allocate large reply buffers
3311 Details : After long uptimes the MDS can fail to allocate large reply
3312 buffers (e.g. zconf client mount config records) due to memory
3313 fragmentation or consumption by the buffer cache. Preallocate
3314 some large reply buffers so that these replies can be sent even
3315 under memory pressure.
3319 Frequency : rare (liblustre)
3320 Description: fsx running with liblustre complained that using truncate() to
3321 extend the file doesn't work. This patch corrects that issue.
3322 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3323 ATTR_SIZE and lsm use in llu_setattr_raw.
3327 Frequency : rare, only 2.6 kernels
3328 Description: Unusual file access patterns on the MDS may result in inode
3329 data being lost in very rare circumstances.
3330 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3331 code in the 2.6 kernel caused the inode and/or EA data not to
3332 be read from disk, causing single-file corruption.
3336 Frequency : rare, only 2.6 filesystems using extents
3337 Description: Heavy concurrent write and delete load may cause data corruption.
3338 Details : It was possible under high-load situations to have an extent
3339 metadata block in the block device cache from a just-unlinked
3340 file overwrite a newly-allocated data block. We now unmap any
3341 metadata buffers that alias just-allocated data blocks.
3345 Frequency : filesystems with default stripe_count larger than 77
3346 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3347 Details : lconf specifies an inode size of 4096 bytes when the default
3348 stripe_count is larger than 77. This conflicts with the default
3349 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3350 this case to avoid pinning too much memory for large EAs.
3352 ------------------------------------------------------------------------------
3354 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3359 Frequency : rare (extremely heavy IO load with hundreds of clients)
3361 Description: Client is evicted, gets IO error writing to file
3362 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3363 caused clients to be evicted for AST timeouts. The fixes in
3364 bug 5192 mean we no longer need to have such short AST timeouts
3365 so ldlm_timeout has been increased.
3368 Frequency : occasional during --force or --failover shutdown under load
3369 Bugzilla : 5949, 4834
3370 Description: Server oops/LBUG if stopped with --force or --failover under load
3371 Details : a collection of import/export refcount and cleanup ordering
3372 issues fixed for safer force cleanup
3375 Frequency : only filesystems larger than 120 OSTs
3376 Bugzilla : 5990, 6223
3377 Description: lfs getstripe would oops on a very large filesystem
3378 Details : lov_getconfig used kfree on vmalloc'd memory
3381 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3382 Bugzilla : 6242, 6243
3383 Description: reading from files that had been truncated to a non-zero size
3384 but never opened returned no data
3385 Details : ll_file_read() reads zeros from no-object files to EOF
3390 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3391 Details : There's little harm in aborting MDS/OSS recovery and letting it
3392 try again, so I removed the LASSERT and return an error instead.
3394 Severity : enhancement
3396 Description: New debugging infrastructure for tracking down data corruption
3397 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3398 (b) cover more of the client-side code path, and (c) try to narrow
3399 down where problems occurred
3403 Bugzilla : 3819, 4364, 4397, 6313
3404 Description: Racing close and eviction MDS could cause assertion in mds_close
3405 Details : It was possible to get multiple mfd references during close and
3406 client eviction, leading to one thread referencing a freed mfd.
3408 Severity: : enhancement
3409 Bugzilla : 3262, 6359
3410 Description: Attempts to reconnect to servers are now more aggressive.
3411 Details : This builds on the enhanced upcall-less recovery that was added
3412 in 1.4.2. When trying to reconnect to servers, clients will
3413 now try each server in the failover group every 10 seconds. By
3414 default, clients would previously try one server every 25 seconds.
3419 Description: After recovery, certain operations trigger a failed
3420 assertion on a client.
3421 Details : Failing over an mds, using lconf -d --failover, while a
3422 client was doing a readdir() call would cause the client to
3423 LBUG after recovery completed and the readdir() was resent.
3425 Severity : enhancement
3427 Description: Default groups are now added by lconf
3428 Details : You can now run lconf --group <servicename> without having to
3429 manually add groups with lmc.
3432 Frequency : occasional
3434 Description: Nodes with an elan id of 0 trigger a failed assertion
3437 Frequency : always when accessing e.g. tty/console device nodes
3439 Description: tty and some other devices nodes cannot be used on lustre
3440 Details : file's private_data field is used by device data and lustre
3441 values in there got lost. New field was added to struct file to
3442 store fs-specific private data.
3445 Frequency : when exporting Lustre via NFS
3447 Description: NFSD failed occasionally when looking up a path component
3448 Details : NFSD is looking up ".." which was broken in ext3 directories
3449 that had grown large enough to become hashed.
3452 Frequency : Clusters with multiple interfaces not on the same subnet
3454 Description: Nodes will repeatedly try to reconnect to an interface which it
3455 cannot reach and report an error to the log.
3456 Details : Extra peer list entries will be created by lconf with some peers
3457 unreachable. lconf now validates the peer before adding it.
3460 Frequency : Only if a default stripe is set on the filesystem root.
3462 Description: Setting a default stripe on the filesystem root prevented the
3463 filesystem from being remounted.
3464 Details : The client was sending extra request flags in the root getattr
3465 request and did not allocate a reply buffer for the dir EA.
3468 Frequency : occasional, higher if lots of files are accessed by one client
3469 Bugzilla : 6159, 6097
3470 Description: Client trips assertion regarding lsm mismatch/magic
3471 Details : While revalidating inodes the VFS looks up inodes with ifind()
3472 and in rare cases can find an inode that is being freed.
3473 The ll_test_inode() code will free the lsm during ifind()
3474 when it finds an existing inode and then the VFS later attaches
3475 this free lsm to a new inode.
3479 Bugzilla : 6422, 7030
3480 Description: MDS deadlock between mkdir and client eviction
3481 Details : Creating a new file via mkdir or mknod (starting a transaction
3482 and getting the ns lock) can deadlock with client eviction
3483 (gets ns lock and trying to finish a synchronous transaction).
3486 Frequency : occasional
3487 Description: While starting a server, the fsfilt_ext3 module could not be
3489 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3490 kernels. Previously, lconf would still use the ext3 name
3491 when trying to load modules. Now, it will correctly use
3492 ext3 on 2.4 and ldiskfs on 2.6.
3494 Severity : enhancement
3495 Description: The default stripe count has been changed to 1
3496 Details : The interpretation of the default stripe count (0, to lfs
3497 or lmc) has been changed to mean striping across a single
3498 OST, rather than all available. For general usage we have
3499 found a stripe count of 1 or 2 works best.
3501 Severity : enhancement
3502 Description: Add support for compiling against Cray portals.
3503 Details : Conditional compiling for some areas that are different
3507 Frequency : occasional
3508 Bugzilla : 6409, 6834
3509 Description: Creating files with an explicit stripe count may lead to
3510 a failed assertion on the MDS
3511 Details : If some OSTs are full or unavailable, creating files may
3512 trigger a failed assertion on the MDS. Now, Lustre will
3513 try to use other servers or return an error to the
3517 Frequency : occasional
3519 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3520 caused lock timeout during readahead (since 1.4.2).
3521 Details : Processes doing readahead might match a lock that hasn't been
3522 granted yet if there are overlapping and conflicting lock
3523 requests. The readahead process waits on ungranted lock
3524 (original lock is CBPENDING), while OST waits for that process
3525 to cancel CBPENDING read lock and eventually evicts client.
3527 Severity : enhancement
3529 Description: Initial enabling of flock support for clients
3530 Details : Implements fcntl advisory locking and file status functions.
3531 This feature is provided as an optional mount flag (default
3532 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3533 locking are implemented yet, and those that are are not guaranteed
3534 to be completely correct in production environments.
3535 mount -t lustre -o [flock|noflock] ...
3538 Frequency : occasional
3540 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3541 assertion in the ext3 JBD (journaling) layer.
3542 Details : The b_committed_data struct is protected by the big kernel lock
3543 in 2.4 kernels, serializing journal_commit_transaction() and
3544 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3545 access to this struct. In 2.6 kernels there is finer grained
3546 locking to improve SMP performance of the JBD layer.
3550 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3555 Description: merge and cleanup kernel patches.
3558 -----------------------------------------------------------------------------
3560 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3562 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3563 versions of Lustre. Please update all servers and clients to
3564 version 1.4.2 or later at the same time. You must also run
3565 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3566 to update the configuration logs.
3568 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3569 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3570 - swab llog records of type '0' so we get proper header size/idx (5861)
3571 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3572 - fix rename of one directory over another leaking an inode (5953)
3573 - avoid SetPageDirty on 2.6 (5981)
3574 - don't re-add just-being-destroyed locks to the waiting list (5653)
3575 - when creating new directories, inherit the parent's custom
3576 striping settings if present parent (3048)
3577 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3578 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3579 - don't hold client locks on temporary worklist from l_lru (5666)
3580 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3581 - automatically evict dead clients (5921)
3582 - Update file size properly in create+truncate+fstat case (6196)
3583 - Do not unhash mountpoint dentries, do not allow removal of
3585 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3586 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3587 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3588 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3589 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3590 - Correctly handle reads of files with no objects (6243)
3591 - lctl recover will also mark a device active if deactivate used (5933)
3593 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3594 - allow --write-conf on an MDS with different nettype than client (5619)
3595 - don't write config llogs to MDS for mounts not from that MDS (5617)
3596 - lconf should create multiple TCP connections from a client (5201)
3597 - init scripts are now turned off by default; run chkconfig --on
3598 lustre and chkconfig --on lustrefs to use them
3599 - upcalls are no longer needed for clients to recover to failover
3601 - add --abort-recovery option to lconf to abort recovery on device
3603 - add support for an arbitrary number of OSTs (3026)
3604 - Quota support protocol changes.
3605 - forward compatibility changes to wire structs (6007)
3606 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3607 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3608 - improved "lustre status" script
3609 - initialize blocksize for non-regular files (6062)
3610 - added --disable-server and --disable-client configure options (5782)
3611 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3612 - Vanilla 2.4.29 support
3613 - increase maximum number of obd devices to 520 (6242)
3614 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3615 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3616 SLES 9 2.6.5 kernels
3617 - limit stripes per file to 160 (the maximum EA size) (6093)
3619 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3622 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3623 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3624 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3625 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3626 - small ext3 extents cleanups and fixes (5733)
3627 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3628 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3629 - avoid needless client->OST connect, fix handle mismatch (5317)
3630 - fix DLM error path that led to out-of-sync client, long delays (5779)
3631 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3632 - fix several locking issues related to i_size (5492,5624,5654,5672)
3633 - don't move pending lock onto export if it is already evicted (5683)
3634 - fix kernel oops when creating .foo in unlinked directory (5548)
3635 - fix deadlock in obdfilter statistics vs. object create (5811)
3636 - use time_{before,after} to avoid timer jiffies wrap (5882)
3637 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3638 - Do not leak request if server was not able to process it (5154)
3639 - If mds_open unable to find parent dir, make that negative lookup(5154)
3640 - don't create new directories with extent-mapping (5909, 5936)
3642 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3643 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3644 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3645 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3646 - support for RHEL4 kernel on i686 (5773)
3647 - provide error messages when incompatible logs are encountered (5898)
3649 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3650 * version 1.4.0.10 (1.4.1 release candidate 1)
3652 - don't keep a lock reference when lock is not granted (4238)
3653 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3654 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3655 - drop import inflight refcount on signal_completed_replay error (5255)
3656 - unlock page after async write error during send (3677)
3657 - handle missing objects in filter_preprw_read properly (5265)
3658 - no transno return for symlink open, don't save no-trasno open (3440)
3659 - don't try to complete elan receive that already failed (4012)
3660 - free RPC server reply state on error (5406)
3661 - clean up thread from ptlrpc_start_thread() on error (5160)
3662 - readahead could read extra page into cache that wasn't ejected (5388)
3663 - prevent races in class_attach/setup/cleanup/detach (5260)
3664 - don't dereference de->d_inode after l_dput of de (5458)
3665 - use "int" for stripe value returned from lock_to_stripe (5544)
3666 - mballoc allocation and error-checking fixes in 2.6 (5504)
3667 - block device patches to fix I/O request sizes in 2.6 (5482)
3668 - look up hostnames for IB nals (5602)
3669 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3670 - don't start multiple acceptors for the same port (5277)
3671 - fix incorrect LASSERT in mds_getattr_name (5635)
3672 - export a proc file for general "ping" checking (5628)
3673 - fix "lfs check" to not block when the MDS is down (5628)
3675 - service request history (4965)
3676 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3677 - create an "evict_client" /proc entry on OSTs, like the MDS has
3678 - fix mount usage message, return errors per mount(8) (5168)
3679 - change grep [] to grep "[]" in tests so they work in more UMLs
3680 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3681 - remove ancient LOV_MAGIC_V0 EA support (5047)
3682 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3683 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3684 - allow readahead allocations to fail when low on memory (5383)
3685 - mmap locking landed again, after considerable improvement (2828)
3686 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3688 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3691 - send OST transaction number in read/write reply to free req (4966)
3692 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3693 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3694 - the watchdog thread now runs as interruptible (5246)
3695 - flock/lockf fixes (but it's still disabled, pending 5135)
3696 - don't use EXT3 constants in llite code (5094)
3697 - memory shortage at startup could cause assertion (5176)
3699 - reorganization of lov code
3700 - single portals codebase
3702 - add extents/mballoc support (5025)
3703 - direct I/O reads in the obdfilter (4048)
3704 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3706 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3709 - send OST transaction number in read/write reply to free req (4966)
3710 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3711 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3712 - the watchdog thread now runs as interruptible (5246)
3713 - handle missing objects in filter_preprw_read properly (5265)
3714 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3715 - drop import inflight refcount on signal_completed_replay error (5255)
3716 - unlock page after async write error during send (3677)
3717 - return original error code on reconstructed replies (3761)
3718 - no transno return for symlink open, don't save no-trasno open (3440)
3720 - add pid to ldlm debugging output (4922)
3721 - bump the watchdog timeouts -- we can't handle 30sec yet
3722 - extra debugging for orphan dentry/inode bug (5259)
3724 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3727 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3728 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3729 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3730 - fix NULL dereference in /proc/sys/portals/routes (4827)
3731 - allow failed mdc_close() operations to be interrupted (4561)
3732 - stop precreate on OST before MDS would time out on it (4778)
3733 - don't send partial-page writes before EOF from client (4410)
3734 - discard client grant for sub-page writes on large-page clients (4520)
3735 - don't free dentries not owned by NFS code, check generation (4806)
3736 - fix lsm leak if mds_create_objects() fails (4801)
3737 - limit debug_daemon file size, always print CERROR messages (4789)
3738 - use transno after validating reply (3892)
3739 - process timed out requests if import state changes (3754)
3740 - update mtime on OST during writes, return in glimpse (4829)
3741 - add mkfsoptions to LDAP (4679)
3742 - use ->max_readahead method instead of zapping global ra (5039)
3743 - don't interrupt __l_wait_event() during strace
3745 - add software watchdogs to catch hung threads quickly (4941)
3746 - make lustrefs init script start after nfs is mounted
3747 - fix CWARN/ERROR duplication (4930)
3748 - return async write errors to application if possible (2248)
3749 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3750 - print NAL number in %x format (4645)
3751 - update barely-supported suse-2.4.21-171 series (4842)
3752 - support for sles 9 %post scripts
3753 - support for building 2.6 kernel-source packages
3754 - support for sles km_* packages
3756 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3759 - ignore -ENOENT errors in osc_destroy (3639)
3760 - notify osc create thread that OSC is being cleaned up (4600)
3761 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3762 - reconstruct ost_handle() like mds_handle() (4657)
3763 - create a new thread to do import eviction to avoid deadlock (3969)
3764 - let lconf resolve symlinked-to devices (4629)
3765 - don't unlink "objects" from directory with default EA (4554)
3766 - hold socknal file ref over connect in case target is down (4394)
3767 - allow more than 32000 subdirectories in a single directory (3244)
3768 - fix blocks count for O_DIRECT writes (3751)
3769 - OST returns ENOSPC from object create when no space left (4539)
3770 - don't send truncate RPC if file size isn't changing (4410)
3771 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3772 - bind to privileged port in socknal and tcpnal (3689)
3774 - rate limit CERROR/CWARN console message to avoid overload (4519)
3775 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3776 - basic mmap support (3918)
3777 - kernel patch series update from b1_4 (4711)
3779 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3782 - avoid crash during MDS cleanup with OST shut down (2775)
3783 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3784 - don't use bad inodes on the MDS (3744)
3785 - dynamic object preallocation to improve recovery speed (4236)
3786 - don't hold spinlock over lock dumping or change debug flags (4401)
3787 - don't zero obd_dev when it is force cleaned (3651)
3788 - print grants to console if they go negative (4431)
3789 - "lctl deactivate" will stop automatic recovery attempts (3406)
3790 - look for existing locks in ldlm_handle_enqueue() (3764)
3791 - don't resolve lock handle twice in recovery avoiding race (4401)
3792 - revalidate should check working dir is a directory (4134)
3794 - don't always mark "slow" obdfilter messages as errors (4418)
3796 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3799 - don't close LustreDB during write_conf until it is done (3860)
3800 - fix typo in lconf for_each_profile (3821)
3801 - allow dumping logs from multiple threads at one time (3820)
3802 - don't allow multiple threads in OSC recovery (3812)
3803 - fix debug_size parameters (3864)
3804 - fix mds_postrecov to initialize import for llog ctxt (3121)
3805 - replace config semaphore with spinlock (3306)
3806 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3807 - don't allow enqueue to complete on a destroyed export (3822)
3808 - down write_lock before checking llog header bitmap (3825)
3809 - recover from lock replay timeout (3764)
3810 - up llog sem before sending rpc (3652)
3811 - reduce ns lock hold times when setting kms (3267)
3812 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3813 - fix NULL deref and obd_dev leak on setup error (3312)
3814 - replace some LBUG about llog ops with error handling (3841)
3815 - don't match INVALID dentries from d_lookup and spin (3784)
3816 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3817 - fix invalid assertion in ptlrpc_set_wait (3880)
3819 - add libwrap support for the TCP acceptor (3996)
3820 - add /proc/sys/portals/routes for non-root route listing (3994)
3821 - allow setting MDS UUID in .xml (2580)
3822 - print the stack of a process that LBUGs (4228)
3824 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3827 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3828 - make sure to unset replay flag from failed open requests (3440)
3829 - if default stripe count is 0, use OST count for inode size (3636)
3830 - update parent mtime/ctime on client for create/unlink (2611)
3831 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3832 - free recovery state on server during a forced cleanup (3571)
3833 - unregister_reply for resent reqs (3063)
3834 - loop back devices mounting and status check on 2.6 (3563)
3835 - fix resource-creation race that can provoke i_size == 0 (3513)
3836 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3837 - more debugging for page-accounting assertion (3746)
3838 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3839 - avoid deadlock after precreation failure (3758)
3840 - fix race and lock order deadlock in orphan handling (3450, 3750)
3841 - add validity checks when grabbing inodes from l_ast_data (3599)
3843 - add /proc/.../recovery_status to obdfilter (3428)
3844 - lightweight CDEBUG infrastructure, debug daemon (3668)
3845 - change default OSC RPC parameters to be better on small clusters
3846 - turn off OST read cache for files smaller than 32MB
3847 - install man pages and include them in rpms (3100)
3848 - add new init script for (un)mounting lustre filesystems (2593)
3849 - run chkconfig in %post for init scripts (3701)
3850 - drop scimac NAL (unmaintained)
3852 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3855 - clean kiobufs before and after use (3485)
3856 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3857 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3858 - change init script to fail more clearly if not run as root (1528)
3859 - allow clients to reconnect during replay (1742)
3860 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3861 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3862 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3863 - fix cond_resched (3554)
3864 - only evict unfinished clients after recovery (3515)
3865 - allow bulk resend, prevent data loss (3570)
3866 - dynamic ptlrpc request buffer allocation (2102)
3867 - don't allow unlinking open directory if it isn't empty (2904)
3868 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3869 - remove extraneous obd dereference causing LASSERT failure (3334)
3870 - don't use get_cycles() when creating temp. files on the mds (3156)
3871 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3872 - handle EEXIST for set-stripe, set proper directory name (3336)
3874 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3875 - fix ksocknal_fmb_callback() error messages (2918)
3877 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3880 - don't copy lvb into (possibly NULL) reply on error (2983)
3881 - don't deref dentry after dput, don't free lvb on error (2922)
3882 - use the kms to determine writeback rpc length (2947)
3883 - increment oti_logcookies when osc is inactive (2948)
3884 - update client's i_blocks count via lvb messages (2543)
3885 - handle intent open/close of special files properly (1557)
3886 - mount MDS with errors=remount-ro, like obdfilter (2009)
3887 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3888 - don't use cancelling-locks' kms values (2947)
3889 - use highest lock extent for kms, not last one (2925)
3890 - don't dereference ERR_PTR() dentry in error handling path (3107)
3891 - fix thread race in portals_debug_dumplog() (3122)
3892 - create lprocfs device entries at setup instead of at attach (1519)
3893 - common AST error handler, don't evict client on completion race (3145)
3894 - zero nameidata in detach_mnt in 2.6 (3118)
3895 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3896 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3897 - initialize RPC timeout timer earlier for 2.6 (3219)
3898 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3899 - print nal/nid for unknown nid (3258)
3900 - additional checks for oscc recovery before doing precreate (3284)
3901 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3902 - don't crash in mdc_close for bad permissions on open (3285)
3903 - zero i_rdev for non-device files (3147)
3904 - clear page->private before handing to FS, better assertion (3119)
3905 - tune the read pipeline (3236)
3906 - fix incorrect decref of invalidated dentry (2350)
3907 - provide read-ahead stats and refine rpc in flight stats (3328)
3908 - don't hold journal transaction open across create RPC (3313)
3909 - update atime on MDS at close time (3265)
3910 - close LDAP connection when recovering to avoid server load (3315)
3911 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3912 - don't leak open file on MDS after open resend (3325)
3913 - serialize filter_precreate and filter_destroy_precreated (3329)
3914 - loop device shouldn't call sync_dev() for nul device (3092)
3915 - clear page cache after eviction (2766)
3916 - resynchronize MDS->OST in background (2824)
3917 - refuse to mount the same filesystem twice on same mountpoint (3394)
3918 - allow llmount to create routes for mounting behind routers (3320)
3919 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3920 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3921 - fix rare problem with rename on htree directories (3417)
3923 - allow default OST striping configuration per directory (1414)
3924 - fix compilation for qswnal for 2.6 kernels (3125)
3925 - increase maximum number of MDS request buffers for large systems
3926 - change liblustreapi to be useful for external progs like lfsck (3098)
3927 - increase local configuration timeout for slow disks (3353)
3928 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3930 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3933 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3934 - don't overwrite extent policy data in reply if lock was blocked (2901)
3935 - drop filter export grants atomically with removal from device (2663)
3936 - del obd_self_export from work_list in class_disconnect_exports (2908)
3937 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3938 - swab reply message in mdc_close, other PPC fixes (2464)
3939 - fix destroying of named logs (2325)
3940 - overwrite old logs when running lconf --write_conf (2264)
3941 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3942 - fix race in target_handle_connect (2898)
3943 - mds_reint_create() should take same inode create lock (2926)
3944 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3945 - don't close files for self_export to avoid uninitialized obd (2936)
3946 - allow MDS with the same name as client node (2939)
3947 - hold dentry reference for closed log files for unlink (2325)
3948 - reserve space for all logs during transactions (2059)
3949 - don't evict page beyond end of stripe extent (2925)
3950 - don't oops on a deleted current working directory (2399)
3951 - handle hard links to targets without a parent properly (2517)
3952 - don't dereference NULL lock when racing during eviction (2867)
3953 - don't grow lock extents when lots of conflicting locks (2919)
3955 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3958 - account for cache space usage on clients to avoid data loss (974)
3959 - lfsck support in lustre kernel code (2349)
3960 - reduce journal credits needed for BRW writes (2370)
3961 - orphan handling to avoid losing space on client/server crashes
3962 - ptlrpcd can be blocked, stopping ALL progress (2477)
3963 - use lock value blocks to assist in proper KMS, faster stat (1021)
3964 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3965 - recovery for initial connections (2355)
3966 - fixes for mds_cleanup_orphans (1934)
3967 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3968 - block all file creations until orphan recovery completes (1901)
3969 - client remove rq_connection from request struct (2423)
3970 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3971 - recovery timer race (2670)
3972 - mdc_close recovey bug (2532)
3973 - ptlrpc cleanup bug (2710)
3974 - mds timeout on local locks (2588)
3975 - namespace lock held during RPCs (2431)
3976 - handle interrupted sync write properly (2503)
3977 - don't try to handle a message that hasn't been replied to (2699)
3978 - client assert failure during cleanup after abort recovery (2701)
3979 - leak mdc device after failed mount (2712)
3980 - ptlrpc_check_set allows timedout requests to complete (2714)
3981 - wait for inflight reqs when ptlrpcd finishes (2710)
3982 - make sure unregistered services are removed from the srv_list
3983 - reset bulk XID's when resending them (caught by 1138 test)
3984 - unregister_bulk after timeout
3985 - fix lconf error (2694)
3986 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3987 - readahead locks pages, leaves pending causing memory pressure (2673)
3988 - increase OST request buffers to 4096 on large machines (2729)
3989 - fix up permission of existing directories in simple_mkdir (2661)
3990 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3991 - don't assign transno to errored transactions (2742)
3992 - don't delete objects on OST if given a bogus objid from MDS (2751)
3993 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3994 - if rq_no_resend, then timeout request after recovery (2432)
3995 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3996 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3997 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3998 - MDS recovery shouldn't delete objects that it has given out (2730)
3999 - if enqueue arrives after completion, don't clobber LVB (2819)
4000 - don't unlock pages twice when trigger_group_io returns error (2814)
4001 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4002 - don't write pages to disk if there was an error (1450)
4003 - don't ping imports that have recovery disabled (2676)
4004 - take buffered bytes into account when balancing socknal conn (2817)
4005 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4006 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4007 - remove little-endian swabbing of llog records (1987)
4008 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4009 - retry reposting request buffers if they fail (1191)
4010 - grow extent at grant time to avoid granting a revoked lock (2809)
4011 - lock revoke doesn't evict page if covered by a second lock (2765)
4012 - disable VM readahead to avoid reading outside lock extents (2805)
4014 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4015 - updated kernel patches for hp-2.4.20 kernel (2681)
4017 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4020 - fix truncated write corruption (2366)
4021 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4023 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4024 - ensure len(uuid) < 37 in lmc (1171)
4025 - fix ia64 OOPS in llog_test (2255)
4026 - zero end of page at obdfilter for partial page writes (2648)
4027 - don't leave stale dentries around after renames (bug 2428)
4028 - fix timeouts when evicting a client with a single lock held (2642)
4029 - set deadline for the initial HELLO message to drain (2634)
4030 - print out dotted-quad IP addresses in the socknal (2302)
4032 - additional debugging for MDS client eviction problem (2443)
4033 - fix mkfsoptions support for osts (2603, 2604)
4035 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4038 - add series for the vanilla 2.6.0 kernel
4039 - add series for the vanilla 2.4.24 kernel
4040 - add series for a cray x86/64 UL kernel drop
4041 - fix xattr patches for the vanilla 2.4.19 series
4043 - generate true UUIDs in lmc (1171)
4044 - have portals stack dumping break in UML (2466)
4045 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4046 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4047 - wake callback waiting threads on client eviction (2460)
4048 - Add --ptldebug and --subsystem to lmc (1719)
4049 - update assertion to allow safe interrupt allocation
4050 - set rq_no_resend for cancel requests (2432)
4051 - recalculate ptlrpcd timeout after resend (2494)
4052 - call vfs_rmdir when removing pending directories (2368)
4053 - fix renaming a file to itself (2429)
4054 - lmc creates a default one-stripe lov (2454)
4055 - expand procfs space to handle large clusters (2326)
4056 - increase UML stack to avoid overflow
4057 - update lconf's list of debug and subsystem masks
4058 - fix lfs find --obd (2510)
4059 - /proc tunable for disabling filter read caching (2591)
4060 - stop rpm packages from altering slapd.conf (2301)
4061 - disable nagle in the socknal under 0conf (2578)
4062 - choose mds inode size based on stripe count (2572)
4063 - fix kernel-source rpm problems (2516)
4065 - add --disable-doc to avoid pdf generation (2421)
4066 - update documentation, tests, type-os, comments
4067 - avoid format warnings on ia64
4068 - remove the TOE NAL
4069 - tiny code cleanups by removing unused fields
4071 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4074 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4075 - remove the most blatant lies from BUILDING file (2371)
4076 - change default debug level to reasonable production setting
4077 - reduce client side cache size to reduce cache flush time
4078 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4079 - make TCP zerocopy and pinger support enabled by default (2476)
4080 - sync writes completed after process exits caused crashes (2319)
4081 - maintain correct mount count on the MDS (2356)
4082 - backout 1557, because 2316 wasn't really fixed
4083 - better file I/O statistics gathering in /proc
4084 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4085 - another kernel patch to fix zero-copy TCP function export
4086 - don't take duplicate lock when processing re-sent getattr (2420)
4087 - lctl uses obd_self_export instead of creating new conn (2353)
4088 - MDS/OST recovery case which requires object creation asserted (2425)
4089 - move lfs from /usr/sbin to /usr/bin in packages
4090 - fix race between mds_client_add and mds_client_free (2417)
4091 - use kmalloc instead of slabs in portals (2430)
4092 - don't create duplicate records when a failover MDS is present (2442)
4093 - remove unnecessary mount age check (2332)
4094 - don't remove directory inodes from locks prematurely (2451)
4095 - don't break if MDS service name is the same as hostname (2103)
4096 - fix races in client write RPC generation when cache full (2482)
4098 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4101 - remove now-unused request->rq_obd (278)
4102 - if an allocation fails, print out how much memory we've used (1933)
4103 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4104 - add the "configurable stack size" patch to most series files (1256)
4105 - ability to write large log records, for 100+ OST configs (2306)
4106 - fix NULL deref when filter_prep fails (2314)
4107 - fix operator precedence error in filter_sync
4108 - dynamic allocation of socknal TX descriptors (2315)
4109 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4110 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4111 - ensure that CWARN messages in Portals always get to the syslog
4112 - __init/__exit are not for prototype decls (ldlm_init/exit)
4113 - x86-64 compile warning fixes
4114 - fix gateway LMC keyword conflict (2318)
4115 - fix MDS lock inversions in getattr/reint paths (1844)
4116 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4117 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4118 - DLM race condition prevented some lock evictions (2328)
4119 - ENOMEM detection and retry on socknal sends (2230)
4120 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4121 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4122 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4123 - better error messages when a client is rejected during recovery (1505)
4124 - avoid cancelling locks which were never granted, after failure (2330)
4125 - fix i_sem/journal inversion in mds_client_add (2333)
4126 - fix truncate/getattr lock cycle deadlock (2334)
4127 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4128 - fix two rare exit paths which could leak an l_lock() ref (2321)
4129 - fencepost error in MDS/OST orphan recovery (2226)
4130 - make log record alignment 8 bytes (1988)
4131 - lstripe now fails when requested offset > ost_count (2237)
4132 - ensure that all kernel series have a complete list.h (1607)
4133 - fix crashes in special-file operations (2316)
4134 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4136 - allow configurable automake binary, for testing new versions
4137 - small update to the lfs documentation
4139 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4141 * fix negative export reference count in fsfilt_sync (2312)
4143 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4144 * release candidate 0.9.1
4146 - orphans are moved into the PENDING directory for possible recovery
4147 - replayed opens now open by fid for orphan/rename safety (1042)
4148 - last close of an orphan inode generates a transno (683)
4149 - chdir() and mount() now pin the directory entry (1020)
4150 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4151 - discard very old requests without processing them (1502)
4152 - remove some common, well-understood CERRORs (1505)
4153 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4154 - clear "grant" flags in OST replies until OST grant code lands (1644)
4155 - fix read performance by not clobbering i_blksize on client (1598)
4156 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4157 - make LRU size a /proc tunable, clears locks when reduced (707)
4158 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4159 - server locks take references on exports now (1558)
4160 - build fixes for 2.4.20-rh trees (1663)
4161 - return an error from lov_create if all OSCs are inactive (1751)
4162 - fix import levels when a reconnect happens without a timeout (1597)
4163 - exit early from mds_open if we get a lookup error (1749)
4164 - partial page read at EOF wouldn't wait for disk before sending (1642)
4165 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4166 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4167 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4168 - write/truncate lock inversion (1639)
4169 - Don't auto-load obdclass, portals modules during cleanup (1495)
4170 - fix timestamps from jumping to "now" (1763)
4171 - extra journal assertions (1648)
4172 - add an extra multiunlink test (1771)
4173 - fix read_record/write_record API (1776)
4174 - fix leak of offset_extent, possible incorrect i_size later (1772)
4175 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4176 - Debugging for the kqswnal_get_idle_tx problems (1820)
4177 - Allow recovery to be attempted multiple times (1536)
4178 - Write out MDS last_rcvd file after it is first created (1600)
4179 - Fix tx_descriptor leak in failed transmit situations (1827)
4180 - ext3 journaling fixes for assertion failure after IO error (1871)
4181 - class_export_put() on freed export after completion AST error (1896)
4182 - Fix revalidate looping in VFS (1322)
4183 - Don't access a freed export during MDS_REINT timeout (1521)
4184 - Add open-unlink recovery support on the MDS (1673,1764)
4185 - Return an error if no MDS data was read from last_rcvd (1946)
4186 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4187 - Don't setup a disk device that doesn't match exported UUID (317)
4188 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4189 - avoid committing NULL handle in force close
4190 - local.sh is now a one-stripe LOV configuration
4191 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4192 - fix ext3 htree duplicate directory entry corruption (1516)
4193 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4194 - update to new LOV EA format (2097)
4195 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4197 2003-06-15 Phil Schwan <phil@clusterfs.com>
4200 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4201 - per-import/export recovery handling (958, 931, 959)
4202 - multiple last-rcvd slots, for serving multiple FSes (949)
4203 - connections are again shared between multiple imp/exports (963, 964)
4204 - "umount -f" would hang if any requests needed to be sent (393, 978)
4205 - avoid pinning large req buffer by copying for queued messages (989)
4206 - add "uuid" to "lctl device" command to help upcalls (991)
4207 - "open" RPCs with transnos would confuse recovery counters (1037)
4208 - do proper endian conversion of all wire messages (288, 340, 891)
4209 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4210 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4211 - fix ptlrpc_connection leak in target_handle_connect (1174)
4212 - fix import refcounting bug in OST and MDS cleanup (1134)
4213 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4214 - fix very unlikely obd_types race condition (501)
4215 - remove osc_open hack for echo_client (1187)
4216 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4217 - a failure in read_inode2 leads to deadlock (1139)
4218 - cancel ack-locks as soon as transaction is committed (1072)
4219 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4220 - make sure to commitrw after any preprw to avoid deadlock (1162)
4221 - failing to execute a file in a lustre FS would lock inode (1203)
4222 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4223 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4224 - obd_uuid2tgt would walk off the end of the list (1255)
4225 - on IA64 the osc would give portals incorrect bulk size (1258)
4226 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4227 - fix lock inversion caused by new llite matching code (1282)
4228 - limit the number of dirty pages on a client to 10MB (1286)
4229 - timed out locks were not being corrected cancelled (1289)
4230 - fix O_DIRECT above 4GB on IA-32 (1292)
4231 * major user-visible changes
4232 - fail out/fail over policy now controlled by the upcall (993)
4234 - add OBD_PING to check server availability and failure (954)
4235 - lustre messages are now sent in sending host order (288, 340, 891)
4236 - add eadatalen to MDS getattr reply (340)
4237 - OST read replies may contain second buffer, with per-page status (593)
4239 2003-03-11 Phil Schwan <phil@clusterfs.com>
4242 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4243 - failed open()s could cause deadlock; fixed (867, 869)
4244 - stop cancelling OST locks when files are closed (481)
4245 - overlapping XID spaces caused network corruption (851, 853)
4246 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4247 - setattr_raw added, to do single-RPC, server-side setattrs
4248 - lmc/lconf syntax change for OST UUIDs
4249 - fix crashy race condition between ptlrpc_free_req and osc_close
4250 - don't use request in mdc_enqueue if we hit a timeout (889)
4251 - don't set the inode i_size for regular files from the MDS (896)
4252 - handle out of order completion AST (842)
4253 - don't LBUG if a lock request times out after receiving AST (913)
4254 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4255 - if a bad lock AST arrives, send an error instead of dropping entirely
4256 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4257 - fix leak in bulk IO when only partially completed (899, 900, 926)
4258 - fix O_DIRECT for ia64 (55)
4259 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4260 - C-z after timeout could hang a process forever; fixed (977)
4262 - client-side I/O cache (678, 924, 929, 941, 970)
4264 - READPAGE and SETATTRs which don't take server-side locks get
4267 2003-02-11 Phil Schwan <phil@clusterfs.com>
4270 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4271 - Fix fsfilt_extN_readpage() to read a full page of directory
4272 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4273 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4274 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4275 - Most intent operations were removed, in favour of a new RPC mode
4276 that does a single RPC to the server and bypasses most of the VFS
4277 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4278 - Aggressively cancel local locks on DLM servers
4279 - mds_reint_unlink sends EA to the client if it's the last nlink.
4280 client uses that EA to unlink OST objects.
4281 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4282 - recursive symlinks were fixed (439)
4283 - fixed NULL deref in DEBUG_REQ
4284 - filter_update_lastobjid no longer calls sync, which annoyed extN
4285 - fixed multi-client small-writes to a single file problem (445)
4286 - fixed mtime updates during file writes (607)
4287 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4288 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4289 - fixed Linux OST size reporting problem (444, 656)
4290 - OST now updates object mtime with writes or setattr (607, 619)
4291 - client verifies file size before zeroing page past EOF (445)
4292 - OST now writes last allocated objid to disk with allocation (108)
4293 - LOV on echo now works (409)
4295 - mds_reint_unlink sends a new buffer, with the EA included. this
4296 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4297 set if a regular file was being unlinked, and it was the last link
4298 - use PtlGet from the target for bulk writes (315)
4299 - OST now updates object mtime with writes or setattr (607, 619)
4300 - LDLM now has a grant-time callback to revalidate locked items, if
4302 - Many MDS operations were reorganized to combat race conditions
4304 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4305 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4307 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4310 - Fully reactivate OST imports after reconnection (512, others)
4311 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4312 - More graceful error handling for truncating on dead OST (515)
4313 - Don't error out unless we're actually accessing dead stripes (474)
4314 - Fix garbage sizes when stripes are missing (410)
4315 - LRU counters were broken, causing constant lock purge (433, 432)
4316 - garbage on read from stripes with failed OSTs (441)
4317 - mark OSCs as active before reconnecting during recovery (438)
4318 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4319 - lfind did not preserve OST order in output (443)
4320 - symlinks cause hung clients, incorrect data (439)
4321 - stop dereferencing request after dropping refcount (457)
4322 - don't LASSERT(spin_is_locked) on non-SMP (455)
4323 - fixes for many rename() bugs
4324 - fstat didn't correctly synchronize attributes (399)
4325 - server must handle lock cancellation during blocking AST prep (487)
4326 - bulk descriptors were free()d too soon (511)
4327 - fix paths in lconf, which would load incorrect modules (451, 507)
4328 - fix confusing lconf 'host not found' error message (386)
4329 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4330 - fix race condition in mdc_blocking_ast() for inode access (526)
4331 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4332 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4333 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4334 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4335 - fix lstripe in several ways (kernel side) (527)
4336 - fix request leak in ldlm_cli_enqueue (262)
4337 - incorrect OSC was marked inactive after OST failure
4338 - call mds_fs_cleanup before unmounting filesystem (524)
4339 - fix races between taking ns_lock and ldlm_lock_change_resource
4340 - fix races updating LOV export open file list
4341 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4342 - fix recovery NULL deref in ldlm_cli_cancel_unused
4343 - fix some DLM races by using new hash table for lock handles (419)
4344 - permit the client to specify desired inodes, at replay
4345 - duplicate requests when we queue them for replay reintegration
4346 - fix last_rcvd offset calculation
4347 - sync after each recovered transaction, so we always make progress
4348 - never, not always, ERESTART requests without transnos
4349 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4351 - skip replay if the MDS says that the client is already connected
4352 - don't check for a recovery-enabled export to match lctl's UUID
4353 - don't INC_USE_COUNT for phantom exports
4354 - don't crash when cleaning up phantom exports (567)
4355 - don't double-finish or set replay data for errored mdc_open requests
4356 - abort requests when they time out, so we don't get old replies
4357 - send/receive replies for AST messages again
4358 - if the client says that it doesn't have the lock, cancel it on the
4360 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4361 mark it as destroyed, it will all work out when decref is called
4362 - fix module use counts (22, 581)
4364 - ASTs now expect a reply (server cancels lock on error reply)
4366 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4369 - fix many simultaneous client startup (392)
4370 - fix dentry->d_it clobbering
4371 - credentials weren't being shipped for readdir/getattr operations
4372 - remove invalid assertions triggered during some concurrent MD
4374 - proper Lustre versions added (336, 389)
4375 - fix memory leak for create error case (398)
4376 - fix LOV locking bug that would get cli/srv out of sync
4377 - fix echo client over LOV (409)
4378 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4379 - fix double-O_EXCL intent crash (424)
4380 - avoid sending multiple lock CANCELs (352)
4382 - MDS can do multi-client recovery (modulo bugs in new code)
4384 - many updates, edits, cleanups
4386 2002-11-18 Phil Schwan <phil@clusterfs.com>
4389 - fix null d_it dereference (346)
4390 - fix full OST/dbench hang (333)
4391 - fix permission problem with file removal (286)
4392 - fix removal of OSCs from LOV when they fail
4393 - fix NULL deref during bulk timeout (214)
4394 - fix problems related to multiple filesystems on one MDS (241)
4395 - fixed serious subtle metadata locking bugs
4396 - free locks on clients when inodes are removed due to memory
4398 - fix inode pointer in lock data (285)
4399 - partial support for multiple MDS on a single host (241)
4400 - data locks weren't cancelled at clear_inode time (290, 311)
4401 - intent locks could lead to unbounded lock growth (205)
4402 - added a maximum lock count, an LRU list, and a flusher
4403 - fix multiple rename (365)
4404 - properly abstracted the echo client
4405 - OSC locked 1 byte too many; fixed
4406 - rewrote brw callback code:
4407 - fixed recovery bugs related to LOVs (306)
4408 - fixed too-many-pages-in-one-write crash (191)
4409 - fixed (again) crash in sync_io_timeout (214)
4410 - probably fixed callback-related race (385)
4412 - Add capability to MDS protocol
4413 - LDLM cancellations and callbacks on different portals
4415 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4418 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4419 - defer open object destruction to close time (601981, 138)
4420 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4421 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4422 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4423 - don't call obd_setattr in ll_file_release for destroyed objects
4424 * protocol change to lustre_msg: move |version| and add |flags|
4425 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4426 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4427 * added replay of create, unlink, link and rename operations during
4428 MDS failover; recovery should be much more robust now
4429 * remove failed OSCs from LOVs (only lov_create uses this so far)
4430 * the lustre-HOWTO was brought (more) up to date (582544)
4432 2002-10-23 Phil Schwan <phil@clusterfs.com>
4435 - in-use dentries weren't being reused properly (617851)
4436 - prevent multiple LDLM setup (599178)
4437 - fix LOV size calculations for truncate (617853)
4438 - fix client handling of MDS intent errors (POSIX)
4439 - fix permission bug in lovstripe.c test (624321)
4440 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4441 - truncate past end of file could corrupt data
4442 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4443 - a race in recovery could return ETIMEDOUT to apps (623947)
4444 - building outside the source directory was fixed
4445 * the lustre-HOWTO was brought (more) up to date (582544)
4446 * major progress was made on recovery functionality
4448 2002-10-10 Phil Schwan <phil@clusterfs.com>
4451 - recovery deadlock fix
4452 - rm -rf causes LBUG fix (617817)
4453 - file open by multiple tasks fix (618962)
4454 - directory permissions bugs (602707 and 620007)
4455 - journal_stop fixed with locking (611313)
4456 - O_APPEND failures resolved (618273, perhaps 614459)
4457 - lconf PATH fix (619770)
4458 - IA64 build fix (621450)
4459 - RPC buffer sizes scale with amount of memory
4461 2002-10-01 Phil Schwan <phil@clusterfs.com>
4464 - locks would be cancelled without throwing away data pages,
4465 resulting in inconsistent data (605627)
4466 - inode attributes were not always being refreshed (605627, 612449)
4467 - lconf now continues to cleanup after lctl reports an error
4468 - MDS now enforces user permissions (602707)
4469 - lprocfs cleanup fixed, but not yet enabled (614157)
4470 - fixed infinite server hang, should a client not respond to an AST
4471 - avoid going into recovery if user calls readlink() with a buffer
4472 that's too small (613941)
4473 - AST RPCs no longer require replies (614867) -- this may be changed
4474 - don't crash server if client sends an IOV that's too big (611336)
4475 - fixed lock conversion deadlock (611892)
4476 - fixed the following of symlinks (614622)
4477 * recovery: the server can remove locks from a client that dies, other
4478 clients can make progress
4479 * more extN patch fixes
4480 * compile-time configurable ptlrpc buffer allocations
4482 - collaborative read cache document
4483 - Lustre Lite Performance CDR document-in-progress
4485 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4488 - fix typo in patch-2.4.18
4490 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4493 - clear ptlrpc request each time in handle_incoming_request()
4494 - unlink of files now destroys the object on the OST
4496 2002-09-19 Peter Braam <braam@clusterfs.com>
4498 * add hard link support
4499 * change obdfile creation method
4500 * kernel patch changed
4502 2002-09-19 Peter Braam <braam@clusterfs.com>
4505 - stack overflow bug in extN fixed
4507 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4509 * documentation updates
4510 - add man pages for config tools
4511 - update tests/README to describe testing with new config tools
4512 - finish metadata API descriptions
4513 * bug fixes and cleanups
4514 - statfs workaround for 16TB limit
4515 - LOV stripe allocation improved, can stripe on subset of OSTs
4516 - LOV file size/IO offset was wrong for files > 4GB in size
4517 - object EA data was being dropped, caused files to be unreadable
4518 - memory overflow with non-LOV OST caused memory corruption
4519 - fixed regression tests to work with new config tools, obdfilter
4520 - fixed bug when directory size became larger than 1 block
4521 - fixed bug (for single client case) when PWD was deleted
4522 - invalidate local directory pages when doing intent-based ops
4523 - avoid LDLM oops when lock callback contained bad data
4525 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4527 * documentation updates
4528 * bug fixes and cleanups
4529 - configuration tools
4532 - 64-bit compile warnings
4533 - 64-bit internal statfs data
4535 * test_brw on persistent OST devices
4537 * lprocfs (disabled)
4539 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4541 * documentation updates
4542 * bug fixes and cleanups
4543 * configuration tools
4545 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4548 * many small fixes to 0_5_4
4549 * io/network handling
4550 * thinkos in MDS operations
4552 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4555 * crucial basic fixes to 0.5.3
4556 * IOR, Iozone work over Elan
4559 2002-08-07 Phil Schwan <phil@clusterfs.com>
4560 * version 0_5_3, our first alpha
4561 * we use the new Portals iovs
4562 * documentation updates
4563 * bug fixes and cleanups
4564 * small changes in the DLM wire protocol
4566 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4567 * version 0_5_1 with some initial stability,
4568 * locking on MD and file I/O.
4569 * documentation updates
4570 * several bug fixes since 0.5.0
4571 * small changes in wire protocol
4573 2002-07-18 Phil Schwan <phil@clusterfs.com>
4575 * delivered as Lustre Light Alpha
4576 * fixed a crash after handling invalid MDS requests
4577 * fixed directory pages for architectures with non-4k pages sizes
4579 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4580 * release version v0_4_4
4581 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4583 * Fixes a number of interruption problems with OST operations.
4584 * Update documentation for portals header changes
4585 * Move all wire protocol structs/defines to lustre_idl.h
4586 * Fixes symlink length bug.
4587 * Add tcpdump to repository.
4589 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4590 * release version v0_4_3
4591 * Fixes statfs for inodes on extN.
4592 * Fixes bug in runtests which would delete /etc/hosts.
4593 * Use 64-bit object IDs wherever possible (not into VFS though)
4594 Remove ost_get_info, which is unused by lustre, and out of date.
4596 2002-07-03 Peter Braam <braam@clusterfs.com>
4597 * release version v0_4_2 Fixes a lookup error (type not passed)
4598 * move forward to head of Portals
4599 * move forward to latest Lustre kernel
4601 2002-06-25 Peter Braam <braam@clusterfs.com>
4602 * release version v0_4_1. Hopefully stable on single node use.