1 tbd Sun Microsystems, Inc.
4 2.6.16.54-0.2.5 (SLES 10),
5 2.6.18-53.1.14.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org).
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: 1.40.7-sun1
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 : on ppc only
18 Description: not convert ost objects for directory because it's not exist.
19 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
23 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
25 Severity : enhancement
27 Description: Update to support 2.6.22.14 vanilla kernel.
31 Frequency : rare, on recovery
32 Description: read procfs can produce deadlock in some situation
33 Details : Holding lprocfs lock which send rpc can produce block for destroy
34 obd objects and this also block reconnect with -EALREADY. This isn't
35 fix all lprocfs bugs - but make it rare.
37 Severity : enhancement
39 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
42 Frequency : frequent on X2 node
44 Description: mdc_set_open_replay_data LBUG
45 Details : Set replay data for requests that are eligible for replay.
49 Description: lustre_mgs: operation 101 on unconnected MGS
50 Details : When MGC is disconnected from MGS long enough, MGS will evict the
51 MGC, and late on MGC cannot successfully connect to MGS and a lot
52 of the error messages complaining that MGS is not connected.
55 Frequency : on start mds
57 Description: Implement get_info(last_id) in obdfilter.
60 Frequency : occasional
62 Description: Correctly check stale fid, not start epoch if ost not support SOM
63 Details : open with flag O_CREATE need set old fid in op_fid3 because op_fid2
64 overwrited with new generated fid, but mds can anwer with one of these
65 two fids and both is not stale. setattr incorectly start epoch and
66 assume will be called done_writeting, but without SOM done_writing
70 Frequency : rare, depends on device drivers and load
72 Description: MDS or OSS nodes crash due to stack overflow
73 Details : Code changes in 1.8.0 increased the stack usage of some functions.
74 In some cases, in conjunction with device drivers that use a lot
75 of stack the MDS (or possibly OSS) service threads could overflow
76 the stack. One change which was identified to consume additional
77 stack has been reworked to avoid the extra stack usage.
80 Frequency : occasional
82 Description: Do not fail import if osc_interpret_create gets -EAGAIN
83 Details : If osc_interpret_create got -EAGAIN it immediately exits and
84 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
85 oscc_has_objects and see OSC has no objests and call
86 oscc_internal_create to resend create request.
88 Severity : enhancement
90 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
92 Severity : enhancement
94 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
99 Description: Fix replay if there is an un-replied request and open
100 Details : In some cases, older replay request will revert the
101 mcd->mcd_last_xid on MDS which is used to record the client's
104 Severity : enhancement
106 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
108 Severity : enhancement
110 Description: Add rhel5 support to HEAD.
112 Serverity : enhancement
114 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
119 Description: Don't allow skipping OSTs if index has been specified.
120 Details : Don't allow skipping OSTs if index has been specified, make locking
121 in internal create lots better.
125 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
126 Details : access to bitfield in structure is always rounded to long
127 and this produce problem with not atomic change any bit.
131 Description: Lustre make rpms failed.
132 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
133 builds Lustre rpms from tarball.
136 Frequency : rare on shutdown ost
138 Description: If llog cancel was not send before clean_exports phase, this can
139 produce deadlock in llog code.
140 Details : If llog thread has last reference to obd and call class_import_put
141 this produce deadlock because llog_cleanup_commit_master wait when
142 last llog_commit_thread exited, but this never success because was
143 called from llog_commit_thread.
147 Description: allow userland application know is lost one of stripes.
148 Details : fill lvb_blocks with error code on ost and return it to
149 application if error flag found.
153 Description: NULL lov_tgts causing MDS oops
154 Details : more safe checks for NULL lov_tgts for avoid oops.
156 Severity : enhancement
158 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
162 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
163 Details : If an OST is down the MDS will hang indefinitely in
164 obd_statfs_rqset() waiting for the statfs data. While for
165 MDS QOS usage of statfs, it should not stuck in waiting.
167 Severity : enhancement
169 Description: remote_acl support
170 Details : Support ACL-based permission check for remote user.
171 Support setfacl/getfacl for remote user with the utils
172 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
173 the system "{s,g}etfacl" utils.
175 Severity : enhancement
177 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
179 Severity : enhancement
181 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
185 Description: make mgs_setparam() handle fsname containing dash
186 Details : fsname containing a dash does not work with lctl conf_param
188 Severity : enhancement
190 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
192 Severity : enhancement
194 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
198 Frequency : rare, at shutdown
199 Description: access already free / zero obd_namespace.
200 Details : if client_disconnect_export was called without force flag set,
201 and exist connect request in flight, this can produce access to
202 NULL pointer (or already free pointer) when connect_interpret
203 store ocd flags in obd_namespace.
207 Frequency : only at startup
208 Description: not alloc memory with spinlock held.
209 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
210 if any spinlock held.
212 Severity : enhancement
214 Description: make lustre randomly fail allocating memory
215 Details : Make lustre randomly failed allocating memory for testing purpose.
217 Severity : enhancement
219 Description: lost problems with lov objid file
220 Details : Fixes some scability and access to not inited memory problems
221 in work with lov objdid file.
226 Description: lfs find does not continue on file error
227 Details : Continue other files processing when a file/dir is absent.
231 Description: Inconsistent usage of lustre_pack_reply()
232 Details : Standardize the usage of lustre_pack_reply() such that it
233 always generate a CERROR on failure.
236 Frequency : occasional
238 Description: MDS hang or stay in waiting lock
239 Details : If client receive lock with CBPENDING flag ldlm need send lock
240 cancel as separate rpc, to avoid situation when cancel request
241 can't processed due all i/o threads stay in wait lock.
245 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
249 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
253 Description: Update to SLES9 kernel-2.6.5-7.287.3.
257 Description: Update to RHEL4 latest kernel.
259 Severity : enhancement
261 Description: Build SLES10 patchless client fails
262 Details : The configure was broken by run ./configure with
263 --with-linux-obj=.... argument for patchless client. When the
264 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
265 can't search header adequately. Use absolute path such as
266 -I($LINUX)/include instead.
270 Description: interrupt oig_wait produce painc on resend.
271 Details : brw_redo_request can be used for resend requests from ptlrpcd and
272 private set, and this produce situation when rq_ptlrpcd_data not
273 copyed to new allocated request and triggered LBUG on assert
274 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
275 ptlrpcd set if request is changed and can be safety changed to use
280 Description: LASSERT_{REQ,REP}SWAB macros are buggy
281 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
282 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
283 these macros and replace them with their difinitions instead.
287 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
288 Details : Update which_patch & target file for SLES10 latest kernel.
291 Description: Remove client patches from SLES 10 kernel.
292 Details : This causes SLES 10 clients to behave as patchless clients
293 even on a Lustre-patched (server) kernel.
295 Severity : enhancement
297 Description: self-adjustable client's lru lists
298 Details : use adaptive algorithm for managing client cached locks lru
299 lists according to current server load, other client's work
300 pattern, memory activities, etc. Both, server and client
301 side namespaces provide number of proc tunables for controlling
304 Severity : enhancement
306 Description: light-weight GSS support
307 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
308 and reduce performance overhead.
310 Severity : enhancement
312 Description: Linux keyring support
313 Details : Support using service of Linux keyring for Lustre GSS internal
314 context refresh/cache mechanism.
318 Description: Fix errors in lfs documentation
319 Details : Fixes man pages
323 Description: don't use GFP_* in generic Lustre code.
324 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
328 Description: obdclass is limited by single OBD_ALLOC(idarray)
329 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
333 Description: warnings with build patchless client with vanila 2.6.19 and up
334 Details : change old ctl_table style and replace ctl_table/ctl_table_header
335 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
339 Frequency : Only for SLES
340 Description: sanity_quota fail test_1
341 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
342 file, which makes TSTID[2] inunique.
346 Frequency : Only for SLES9
347 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
348 Details : Update target/ChangeLog/which_patch .
352 Frequency : Only for test_10 in sanity-quota.sh
353 Description: a bug in quota test code
354 Details : checking a test flag in a wrong place
358 Frequency : Only for test_13 in sanity-quota.sh
359 Description: a bug in quota test script
360 Details : shouldn't check overfull allocated quota
364 Description: jbd statistics
365 Details : Port older jbd statistics patch for sles10
367 Severity : enhancement
369 Description: Add printing inode info into message about error in writepage.
374 Description: reply_lock_interpret crash due to race with it and lock cancel.
375 Details : Do not replay locks that are being cancelled. Do not reference
376 locks by their address during replay, just by their handle.
380 Frequency : When flocks are used.
381 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
382 Details : Flock locks might destroy just granted lock if it could be merged
383 with another existing flock, this is done in completion handler,
384 so teach ldlm_cli_enquque_fini that this is a valid case for
390 Description: Oops in read and write path when failing to allocate lock.
391 Details : Check if lock allocation failed and return error back.
395 Description: lstripe command fails for valid OST index
396 Details : The stripe offset is compared to lov->desc.ld_tgt_count
397 instead of lov->desc.ld_active_tgt_count.
401 Description: sanity.sh failed test 103
402 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
407 Description: df doesn't work properly if diskfs blocksize != 4K
408 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
412 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
413 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
417 Description: obd_health_check_timeout too short
418 Details : set obd_health_check_timeout as 1.5x of obd_timeout
422 Description: llapi_file_create() does not allow some changes
423 Details : add llapi_file_open() that allows specifying the mode and
424 open flags, and also returns an open file handle.
428 Description: merge and cleanup kernel patches.
429 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
433 Description: Add journal checksum support.(Kernel part)
434 Details : The journal checksum feature adds two new flags i.e
435 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
436 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
437 indicates that the commit block contains the checksum for
438 the blocks described by the descriptor blocks. Now commit
439 record can be sent to disk without waiting for descriptor
440 blocks to be written to disk. This behavior is controlled
441 using JBD2_FEATURE_ASYNC_COMMIT flag.
445 Description: OSS needs mutliple precreate threads
446 Details : Add ability to start more than one create thread per OSS.
450 Description: Sanity.sh test_65k failed: setstripe should have succeeded
451 Details : execute lfs setstripe on client
455 Description: mds_obd_create error creating tmp object
456 Details : When the user sets quota on root, llog will be affected and can't
457 create files and write files.
460 Frequency : Always on ia64 patchless client, and possibly others.
462 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
463 Details : This allows the patchless client to be loaded on architectures
468 Description: RedHat Update kernel for RHEL5
469 Details : Add the kernel config file for RHEL5.
473 Description: RedHat Update kernel for RHEL5
474 Details : Modify the kernel config file more closer RHEL5.
478 Description: Build failure against Centos5 (RHEL5)
479 Details : Define PAGE_SIZE when it isn't present.
483 Description: test_7 on sanity-quota.sh can't be used on separate machines
484 Details : add facet to handle it
488 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
489 Details : with NFS, the anon dentry's parent was set to itself in
490 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
491 corresponding dentry other than use rec->ur_name.
493 Severity : enhancement
495 Description: lfs setstripe enhancement
496 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
500 Description: enable data checksumming by default
501 Details : enable checksum by default, allow --disable-checksum
502 configure option and "-o nochecksum" mount option. Checksums
503 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
507 Description: lustre support for RHEL5
508 Details : Add support for RHEL5.
512 Description: Client eviction due to failover config
513 Details : after a connection loss, the lustre client should attempt to
514 reconnect to the last active server first before trying the
515 other potential connections.
519 Description: when mds and osts use different quota unit(32bit and 64bit),
520 quota will be released repeatly.
521 Details : void sending multiple quota reqs to mds, which will keep the
522 status between the reqs.
526 Description: osts not allocated evenly to files
527 Details : change the condition to increase offset_idx
530 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
532 Description: Data corruption for OSTs that are formatted larger than 2TB
534 Details : When generating the bio request for lustre file writes the
535 sector number would overflow a temporary variable before being
536 used for the IO. The data reads correctly from Lustre (which
537 will overflow in a similar manner) but other file data or
538 filesystem metadata may be corrupted in some cases.
542 Description: Tune the kernel for good SCSI performance.
543 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
544 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
549 Description: rewrite ext2-derived code in obdclass/uuid.c
550 Details : rewrite inherited code (uuid parsing code from ext2 utils)
551 from scratch preserving functionality.
555 Description: "lfs find -obd UUID" prints directories
556 Details : "lfs find -obd UUID" will return all directory names instead
557 of just file names. It is incorrect because the directories
558 do not reside on the OSTs.
562 Description: use i_size_read and i_size_write in 2.6 port
563 Details : replace inode->i_size access with i_size_read/write()
566 Frequency : when removing large files
568 Description: scheduling issue during removal of large Lustre files
569 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
570 It causes scheduling issues when removing large files (17TB in the
574 Frequency : only with liblustre clients on XT3
576 Description: evictions taking too long
577 Details : allow llrd to evict clients directly on OSTs
580 Frequency : only on ppc
582 Description: /proc/fs/lustre/devices broken on ppc
583 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
584 We should make sure the type of 'index' is loff_t and then cast
585 explicitly as needed below. Do not assign an explicitly cast
589 Frequency : only for rhel5
591 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
592 Details : Modify the target file & which_kernel.
594 Severity : enhancement
596 Description: omit set fsid for export NFS
597 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
598 as FS_REQUIRES_DEV for avoid problems with generate fsid.
602 Frequency : Always, for kernels after 2.6.16
603 Description: Fix warning idr_remove called for id=.. which is not allocated.
604 Details : Last kernels save old s_dev before kill super and not allow
605 to restore from callback - restore it before call kill_anon_super.
609 Description: buffer overruns could theoretically occur
610 Details : llapi_semantic_traverse() modifies the "path" argument by
611 appending values to the end of the origin string, and a
612 overrun may occur. Adding buffer overrun check in liblustreapi.
616 Description: Fix error on 'ls .' at the top of the Lustre mount.
617 Details : Don't revalidate dentry if it is a root dentry.
621 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
622 Details : Modify vm-tunables-rhel4.patch.
626 Description: Kernel config for 2.6.18-vanilla.
627 Details : Modify targets/2.6-vanilla.target.in.
628 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
629 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
630 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
631 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
635 Description: improve handling recoverable errors
636 Details : if request processig with error which can be recoverable on server
637 request should be resend, otherwise page released from cache and
642 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
643 Details : Modify target file & which_patch.
644 A flaw was found in the IA32 system call emulation provided
645 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
646 value could be stored in the %RAX register, which could trigger an
647 out-of-bounds system call table access. An untrusted local user
648 could exploit this flaw to run code in the kernel
649 (ie a root privilege escalation). (CVE-2007-4573).
653 Description: O_DIRECT bypasses client statistics.
654 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
655 read_ahead_stats were updated. copied the stats section from
656 osc_send_oap_rpc() into async_internal().
660 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
664 Description: change order of libsysio includes
665 Details : '#include sysio.h' should always come before '#include xtio.h'
669 Description: handle "serious error: objid * already exists" more gracefully
670 Details : If LAST_ID value on disk is smaller than the objects existing in
671 the O/0/d* directories, it indicates disk corruption and causes an
672 LBUG(). If the object is 0-length, then we should use the existing
673 object. This will help to avoid a full fsck in most cases.
675 Severity : enhancement
677 Description: adapt the lustre_config script to support the upgrade case
678 Details : Add "-u" option for lustre_config script to support upgrading 1.4
679 server targets to 1.6 in parallel.
683 Description: To avoid grant space > avaible space when the disk is almost
684 full. Without this patch you might see the error "grant XXXX >
685 available" or some LBUG about grant, when the disk is almost
687 Details : In filter_check_grant, for non_grant cache write, we should
688 check the left space by if (*left > ungranted + bytes), instead
689 of (*left > ungranted), because only we are sure the left space
690 is enough for another "bytes", then the ungrant space should be
691 increase. In client, we should update cl_avail_grant only there
692 is OBD_MD_FLGRANT in the reply.
696 Description: Update RHEL 4 kernel to fix local root privilege escalation.
697 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
698 described in CVE-2007-4573. This problem could allow untrusted
699 local users to gain root access.
702 Frequency : when using O_DIRECT and quotas
704 Description: Incorrect file ownership on O_DIRECT output files
705 Details : block usage reported by 'lfs quota' does not take into account
706 files that have been written with O_DIRECT.
711 Description: touch file failed when fs is not full
712 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
713 otherwise we can get ENOSP while fs is not full.
717 Description: parallel lock callbacks
718 Details : Instead of sending blocking and completion callbacks as separated
719 requests, adding them to a set and sending in parallel.
722 Frequency : only for Cray XT3
723 Bugzilla : 12829/13455
724 Description: Changing primary group doesn't change the group lustre assigns to
726 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
727 overridden with the primary group provided by the group upcall,
728 whereas the supplied fsgid can be trusted if it is in the list of
729 supplementary groups returned by the group upcall.
731 Severity : enhancement
733 Description: Allow masking D_WARNING, D_ERROR messages from console
734 Details : Console messages can now be disabled via lnet.printk.
738 Description: User code with malformed file open parameter crashes client node
739 Details : Before packing join_file req, all the related reference should be
740 checked carefully in case some malformed flags cause fake
741 join_file req on client.
745 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
746 Details : ldlm_completion_ast() assumes that a lock is granted when the req
747 mode is equal to the granted mode. However, it should also check
748 that LDLM_FL_CP_REQD is not set.
752 Description: Heavy nfs access might result in deadlocks
753 Details : After ELC code landed, it is now improper to enqueue any mds
754 locks under och_sem, because enqueue might want to decide to
755 cancel open locks for same inode we are holding och_sem for.
759 Description: Client eviction while running blogbench
760 Details : A lot of unlink operations with concurrent I/O can lead to a
761 deadlock causing evictions. To address the problem, the number of
762 oustanding OST_DESTROY requests is now throttled to
763 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
764 callbacks are processed in priority.
768 Description: enable ACLs on MDS by default
769 Details : ACLs must be enabled on MDS by default.
772 Frequency : PPC/PPC64 only
774 Description: conflicts between asm-ppc64/types.h and lustre_types.h
775 Details : fix duplicated definitions between asm-ppc64/types.h and
776 lustre_types.h on PPC.
779 Frequency : PPC/PPC64 only
781 Description: asm-ppc/segment.h does not exist
782 Details : fix compile issue on PPC.
786 Description: better handle error messages in extents code
789 Frequency : RHEL4 only
791 Description: mkfs is very slow on IA64/RHEL4
792 Details : A performance regression has been discovered in the MPT Fusion
793 driver between versions 3.02.73rh and 3.02.99.00rh. As a
794 consequence, we have downgraded the MPT Fusion driver in the RHEL4
795 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
797 Severity : enhancement
799 Description: SNMP support enhancement
800 Details : Adding total number of sampled request for an MDS node in snmp
803 Severity : enhancement
805 Description: Optimize ldlm waiting list processing for PR extent locks
806 Details : When processing waiting list for read extent lock and meeting read
807 lock that is same or wider to it that is not contended, skip
808 processing rest of the list and immediatelly return current
809 status of conflictness, since we are guaranteed there are no
810 conflicting locks in the rest of the list.
814 Description: Time out and refuse to reconnect
815 Details : When the failover node is the primary node, it is possible
816 to have two identical connections in imp_conn_list. We must
817 compare not conn's pointers but NIDs, otherwise we can defeat
818 connection throttling.
822 Description: port llog fixes from b1_6 into HEAD
823 Details : Port llog reference couting and some llog cleanups from b1_6
824 (bug 10800) into HEAD, for protect from panic and access to already
825 free llog structures.
829 Description: Detect stride IO mode in read-ahead
830 Details : When a client does stride read, read-ahead should detect that and
831 read-ahead pages according to the detected stride pattern.
835 Description: data checksumming impacts single node performance
836 Details : add support for several checksum algorithm. Currently, only CRC32
837 and Adler-32 are supported. The checksum type can be changed on
838 the fly via /proc/fs/lustre/osc/*/checksum_type.
842 Description: use adler32 for page checksums
843 Details : when available, use the Adler-32 algorithm instead of CRC32 for
848 Description: build for x2 fails
849 Details : fix compile issue on Cray systems.
853 Description: Properly match for duplicate locks
854 Details : Due to different lock order from skiplists code, we need to
855 traverse entire list for now
858 Frequency : only on PPC/SLES10
860 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
861 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
862 be defined. Add a hack in mkfs_lustre.c to work around this compile
867 Description: LASSERT on MDS when client holding flock lock dies
868 Details : ldlm pool logic depends on number of granted locks equal to
869 number of released locks which is not true for flock locks, so
870 just exclude such locks from consideration.
874 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
875 Details : Use fsfilt_sync() for both the whole filesystem sync and
876 individual file sync to eliminate dangerous inode locking
877 with I_LOCK that can lead to a deadlock.
879 --------------------------------------------------------------------------------
881 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
883 * Support for kernels:
884 2.6.5-7.283 (SLES 9),
885 2.6.9-55.EL (RHEL 4),
886 2.6.16.46-0.14 (SLES 10),
887 2.6.18.8 vanilla (kernel.org)
888 * Client support for unpatched kernels:
889 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
890 2.6.16 - 2.6.22 vanilla (kernel.org)
891 * Due to recently discovered recovery problems, we do not recommend
892 using patchless RHEL 4 clients with this or any earlier release.
893 * Recommended e2fsprogs version: 1.39.cfs8
894 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
895 * Starting with this release, the ldiskfs backing filesystem required
896 by Lustre is now in its own package, lustre-ldiskfs. This package
897 should be installed. It is versioned separately from Lustre and
898 may be released separately in future.
902 Description: block reactivating mgc import until all deactivates complete
903 Details : Fix race when failing back MDT/MGS to itself (testing)
906 Frequency : at statup only
908 Description: mds_lov_synchronize race leads to various problems
909 Details : simultaneous MDT->OST connections at startup can cause the
910 sync to abort, leaving the OSC in a bad state.
912 Severity : enhancement
914 Description: add optional extra BUILD_VERSION info
915 Details : add a new environment variable (namely LUSTRE_VERS) which allows
916 to override the lustre version.
919 Frequency : 2.6.18 servers only
921 Description: ll_kern_mount() doesn't release the module reference
922 Details : The ldiskfs module reference count never drops down to 0
923 because ll_kern_mount() doesn't release the module reference.
928 Description: server LBUG when using old ost_num_threads parameter
929 Details : Accept the old ost_num_threads parameter but warn that it
930 is deprecated, and fix an off-by-one error that caused an LBUG.
935 Description: Transient SCSI error results in persistent IO issue
936 Details : iobuf->dr_error is not reinitialized to 0 between two
940 Frequency : sometimes when underlying device returns I/O errors
942 Description: OSTs not going read-only during write failures
943 Details : OSTs are not remounted read-only when the journal commit threads
944 get I/O errors because fsfilt_ext3 calls journal_start/stop()
945 instead of the ext3 wrappers.
949 Description: poor connect scaling with increasing client count
950 Details : Don't run filter_grant_sanity_check for more than 100 exports
951 to improve scaling for large numbers of clients.
954 Frequency : SLES10 only
956 Description: sanity-quota.sh quotacheck failed: rc = -22
957 Details : Quotas cannot be enabled on SLES10.
960 Frequency : liblustre clients only
962 Description: getdirentries does not give error when run on compute nodes
963 Details : getdirentries does not fail when the size specified as an argument
964 is too small to contain at least one entry
966 Severity : enhancement
968 Description: Add LNET router traceability for debug purposes
969 Details : If a checksum failure occurs with a router as part of the
970 IO path, the NID of the last router that forwarded the bulk data
971 is printed so it can be identified.
976 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
977 Details : Due to a race condition, liblustre clients were occasionally
980 Severity : enhancement
982 Description: lfs setstripe use optional parameters instead of postional
985 Severity : enhancement
987 Description: Nanosecond timestamp support for ldiskfs
988 Details : The on-disk ldiskfs filesystem has added support for nanosecond
989 resolution timestamps. There is not yet support for this at
990 the Lustre filesystem level.
993 Frequency : during server recovery
995 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
996 Details : request with rq_no_resend flag not awake l_wait_event if they get a
1000 Frequency : nfs export on patchless client
1002 Description: connectathon hang when test nfs export over patchless client
1003 Details : Disconnected dentry cannot be found with lookup, so we do not need
1004 to unhash it or make it invalid
1007 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1008 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1009 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1014 Description: open req refcounting wrong on reconnect
1015 Details : If reconnect happened between getting open reply from server and
1016 call to mdc_set_replay_data in ll_file_open, we will schedule
1017 replay for unreferenced request that we are about to free.
1018 Subsequent close will crash in variety of ways.
1019 Check that request is still eligible for replay in
1020 mdc_set_replay_data().
1025 Description: disable writes to filesystem when reading health_check file
1026 Details : the default for reading the health_check proc file has changed
1027 to NOT do a journal transaction and write to disk, because this
1028 can cause reads of the /proc file to hang and block HA state
1029 checking on a healthy but otherwise heavily loaded system. It
1030 is possible to return to the previous behaviour during configure
1031 with --enable-health-write.
1033 Severity : enhancement
1035 Description: 64-bit inode version
1036 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1037 to the inode. This will be required for version-based recovery.
1042 Description: MDS fails to start if a duplicate client export is detected
1043 Details : in some rare cases it was possible for a client to connect to
1044 an MDS multiple times. Upon recovery the MDS would detect this
1045 and fail during startup. Handle this more gracefully.
1047 Severity : enhancement
1049 Description: Add -o localflock option to simulate old noflock
1051 Details : This will achieve local-only flock/fcntl locks
1057 Description: log_commit_thread vs filter_destroy race leads to crash
1058 Details : Take import reference before releasing llog record semaphore
1063 Description: Wrong request locking in request set processing
1064 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1065 imp_delayed_list, in this place should be used imp_lock.
1068 Frequency : when reconnection
1070 Description: Grant Leak when osc reconnect to OST
1071 Details : When osc reconnect ost, OST(filter) should check whether it
1072 should grant more space to client by comparing fed_grant and
1073 cl_avail_grant, and return the granted space to client instead
1074 of "new granted" space, because client will call osc_init_grant
1075 to update the client grant space info.
1078 Frequency : when client reconnect to OST
1080 Description: Grant Leak when osc do resend and replay bulk write
1081 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1082 bulk write request, because the grant info while be sync between
1083 OSC and OST when reconnect, and we should ignore the grant info
1084 these of resend/replay write req.
1089 Description: Grant space more than avaiable left space sometimes.
1090 Details : When then OST is about to be full, if two bulk writing from
1091 different clients came to OST. Accord the avaliable space of the
1092 OST, the first req should be permitted, and the second one
1093 should be denied by ENOSPC. But if the seconde arrived before
1094 the first one is commited. The OST might wrongly permit second
1095 writing, which will cause grant space > avaiable space.
1098 Frequency : when client is evicted
1100 Description: Grant might be wrongly erased when osc is evicted by OST
1101 Details : when the import is evicted by server, it will fork another
1102 thread ptlrpc_invalidate_import_thread to invalidate the
1103 import, where the grant will be set to 0. While the original
1104 thread will update the grant it got when connecting. So if
1105 the former happened latter, the grant will be wrongly errased
1106 because of this race.
1111 Description: Checking Stale with correct fid
1112 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1113 to check whether it is stale, because sometimes, we want the
1114 enqueue happened anyway, and op_data.fid2 will not be initialized.
1116 Severity : enhancement
1118 Description: update patchless client
1119 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1122 Frequency : only with 2.4 kernel
1124 Description: random memory corruption
1125 Details : size of struct ll_inode_info is to big for union inode.u and this
1126 can be cause of random memory corruption.
1131 Description: Memory leak in recovery
1132 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1133 It should also check obd_fail before fsfilt_start, otherwise if
1134 fsfilt_start return -EROFS,(failover mds during mds recovery).
1135 then the req will return with repmsg->transno = 0 and rc = EROFS.
1136 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1137 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1138 be freed no matter whether fsfilt_commit success or not.
1141 Frequency : only with huge count clients
1143 Description: Prevents from taking the superblock lock in llap_from_page for
1145 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1146 for llap_from_page call in ll_removepage prevents from taking the
1147 superblock lock for a soon died page.
1152 Description: Not check open intent error before release open handle
1153 Details : in some rare cases, the open intent error is not checked before
1154 release open handle, which may cause
1155 ASSERTION(open_req->rq_transno != 0), because it tries to release
1156 the failed open handle.
1161 Description: Set cat log bitmap only after create log success.
1162 Details : in some rare cases, the cat log bitmap is set too early. and it
1163 should be set only after create log success.
1167 Description: Accessing a block bevice can re-enable I/O when Lustre is
1168 tearing down a device.
1169 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1174 Description: service threads may hog cpus when there are a lot of requests
1175 Details : Insert cond_resched to give other threads a chance to use some CPU
1180 Description: the cat log was not initialized in recovery
1181 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1182 unlink log on mds will not be initialized until mds post
1183 recovery. And also in mds post recovery, the unlink log will
1184 initialization will be done asynchronausly, so there will be race
1185 between add unlink log and unlink log initialization.
1189 Description: brw_stats were being printed incorrectly
1190 Details : brw_stats were being printed as log2 but all of them were not
1191 recorded as log2. Also remove some code duplication arising from
1192 filter_tally_{read,write}.
1196 Frequency : rare, only in recovery.
1197 Description: ASSERTION(req->rq_type != LI_POISON) failed
1198 Details : imp_lock should be held while iterating over imp_sending_list for
1199 prevent destroy request after get timeout in ptlrpc_queue_wait.
1203 Description: replay-single.sh test 52 fails
1204 Details : A lock's skiplist need to be cleanup when it being unlinked
1205 from its resource list.
1209 Description: Short directio read returns full requested size rather than
1211 Details : Direct I/O operations should return actual amount of bytes
1212 transferred rather than requested size.
1214 Severity : enhancement
1216 Description: metadata RPC reduction (e.g. for rm performance)
1217 Details : decrease the amount of synchronous RPC between clients and servers
1218 by canceling conflicing lock before the operation on the client
1219 and packing thier handles into the main operation RPC to server.
1221 Severity : enhancement
1223 Description: Async OSC create to avoid the blocking unnecessarily.
1224 Details : If a OST has no remain object, system will block on the creating
1225 when need to create a new object on this OST. Now, ways use
1226 pre-created objects when available, instead of blocking on an
1227 empty osc while others are not empty. If we must block, we block
1228 for the shortest possible period of time.
1230 Severity : enhancement
1232 Description: refine locking for avoid write wrong info into lov_objid file
1233 Details : fix possible races with add new target and write/update data in
1236 --------------------------------------------------------------------------------
1238 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1243 Frequency : on some architectures
1245 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1246 Details : Uninitialized flags sometimes cause configuration commands to
1250 Frequency : patchless clients only
1252 Description: missing __iget() symbol export
1253 Details : The __iget() symbol export is missing. To avoid the need for
1254 this on patchless clients the deathrow inode reaper is turned
1255 off, and we depend on the VM to clean up old inodes. This
1256 dependency was during via the fix for bug 12181.
1258 --------------------------------------------------------------------------------
1260 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1262 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1263 INTEROPERATE with older versions automatically. In many cases a
1264 special upgrade step is needed. Please read the
1265 user documentation before upgrading any part of a live system.
1266 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
1267 version will not interoperate with 1.6 betas before beta5 (1.5.95).
1268 * WARNING: Lustre configuration and startup changes are required with
1269 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1275 Severity : enhancement
1277 Description: MountConf
1278 Details : Lustre configuration is now managed via mkfs and mount
1279 commands instead of lmc and lconf. New obd types (MGS, MGC)
1280 are added for dynamic configuration management. See
1281 https://mail.clusterfs.com/wikis/lustre/MountConf for
1284 Severity : enhancement
1286 Description: dynamic OST addition
1287 Details : OSTs can now be added to a live filesystem
1289 Severity : enhancement
1291 Description: startup order invariance
1292 Details : MDTs and OSTs can be started in any order. Clients only
1293 require the MDT to complete startup.
1295 Severity : enhancement
1297 Description: parallel, asynchronous orphan cleanup
1298 Details : orphan cleanup is now performed in separate threads for each
1299 OST, allowing parallel non-blocking operation.
1301 Severity : enhancement
1303 Description: optimized stripe assignment
1304 Details : stripe assignments are now made based on ost space available,
1305 ost previous usage, and OSS previous usage, in order to try
1306 to optimize storage space and networking resources.
1308 Severity : enhancement
1310 Description: Permanently set tunables
1311 Details : All writable /proc/fs/lustre tunables can now be permanently
1312 set on a per-server basis, at mkfs time or on a live system.
1314 Severity : enhancement
1316 Description: Lustre message v2
1317 Details : Add lustre message format v2.
1319 Severity : enhancement
1321 Description: client OST exclusion list
1322 Details : Clients can be started with a list of OSTs that should be
1323 declared "inactive" for known non-responsive OSTs.
1327 Description: SPEC SFS validation failure on NFS v2 over lustre.
1328 Details : Changes the blocksize for regular files to be 2x RPC size,
1329 and not depend on stripe size.
1331 Severity : enhancement
1333 Description: Multiple MD RPCs in flight.
1334 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1335 To avoid overly-overloading MDS, introduce a limit on number of
1336 MDS RPCs in flight for a single client and add /proc controls
1337 to adjust this limit.
1339 Severity : enhancement
1341 Description: client read/write statistics
1342 Details : Add client read/write call usage stats for performance
1343 analysis of user processes.
1344 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1345 file access. extents_stats shows chunk size distribution.
1346 extents_stats_per_process show chunk size distribution per
1349 Severity : enhancement
1351 Description: mds statistics
1352 Details : Add detailed mds operations statistics in
1353 /proc/fs/lustre/mds/*/stats.
1357 Description: Failure of copying files with lustre special EAs.
1358 Details : Client side always return success for setxattr call for lustre
1359 special xattr (currently only "trusted.lov").
1364 Description: Refcount LNET uuids
1365 Details : The global LNET uuid list grew linearly with every startup;
1366 refcount repeated list entries instead of always adding to
1369 Severity : enhancement
1371 Description: Dynamic service threads
1372 Details : Within a small range, start extra service threads
1373 automatically when the request queue builds up.
1376 Frequency : mixed-endian client/server environments
1378 Description: mixed-endian crashes
1379 Details : The new msg_v2 system had some failures in mixed-endian
1382 Severity : enhancement
1384 Description: Easy OST removal
1385 Details : OSTs can be permanently deactivated with e.g. 'lctl
1386 conf_param lustre-OST0001.osc.active=0'
1388 Severity : enhancement
1390 Description: MGS proc entries
1391 Details : Added basic proc entries for the MGS showing what filesystems
1394 Severity : enhancement
1396 Description: provide MGS failover
1397 Details : Added config lock reacquisition after MGS server failover.
1399 Severity : enhancement
1401 Description: add Linux 2.4 support
1402 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1406 Description: a large application tries to do I/O to the same resource and dies
1407 in the middle of it.
1408 Details : Check the req->rq_arrival time after the call to
1409 ost_brw_lock_get(), but before we do anything about
1410 processing it & sending the BULK transfer request. This
1411 should help move old stale pending locks off the queue as
1412 quickly as obd_timeout.
1415 Frequency : when an incorrect nid is specified during startup
1417 Description: ptlrpc connect to non-existant node causes kernel crash
1418 Details : LNET can't be re-entered from an event callback, which
1419 happened when we expire a message after the export has been
1420 cleaned up. Instead, hand the zombie cleanup off to another
1423 Severity : enhancement
1425 Description: plain/inodebits lock performance improvement
1426 Details : Grouping plain/inodebits in granted list by their request modes
1427 and bits policy, thus improving the performance of search through
1431 Frequency : only if OST filesystem is corrupted
1433 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1434 Details : for a short time RPCs with bulk IO are in the replay list,
1435 but replay of bulk IOs is unimplemented. If the OST filesystem
1436 is corrupted due to disk cache incoherency and then replay is
1437 started it is possible to trip an assertion. Avoid putting
1438 committed RPCs into the replay list at all to avoid this issue.
1441 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1443 Description: System hang on startup
1444 Details : This bug allowed the liblustre (e.g. catamount) client to
1445 return to the app before handling all startup RPCs. This
1446 could leave the node unresponsive to lustre network traffic
1447 and manifested as a server ptllnd timeout.
1449 Severity : enhancement
1451 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1452 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1453 boolean to control whether to print peer debug info when a
1454 client's RPC times out.
1457 Frequency : only for kernels with patches from Lustre below 1.4.3
1459 Description: Remove old rdonly API
1460 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1463 Frequency : only for devices with external journals
1465 Description: Set external device read-only also
1466 Details : During a commanded failover stop, we set the disk device
1467 read-only while the server shuts down. We now also set any
1468 external journal device read-only at the same time.
1471 Frequency : when upgrading from 1.4 while trying to change parameters
1473 Description: The wrong (new) MDC name was used when setting parameters for
1474 upgraded MDT's. Also allows changing of OSC (and MDC)
1475 parameters if --writeconf is specified at tunefs upgrade time.
1478 Frequency : when setting specific ost indicies
1480 Description: QOS code breaks on skipped indicies
1481 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1482 created with --index need not be sequential.
1484 Severity : enhancement
1486 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1487 Details : The uninit_groups feature works in conjunction with the kernel
1488 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1489 up the pass1 processing of e2fsck. This is a read-only feature
1490 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1491 mount filesystems that have had this feature enabled.
1493 Severity : enhancement
1495 Description: Improve multi-block allocation algorithm to avoid fragmentation
1496 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1497 allocation locality and avoid filesystem fragmentation.
1499 ------------------------------------------------------------------------------
1501 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1503 * Support for kernels:
1504 2.4.21-47.0.1.EL (RHEL 3)
1505 2.6.5-7.283 (SLES 9)
1506 2.6.9-42.0.10.EL (RHEL 4)
1507 2.6.12.6 vanilla (kernel.org)
1508 2.6.16.27-0.9 (SLES 10)
1509 * Recommended e2fsprogs version: 1.39.cfs5
1511 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1515 Frequency : occasional, depends on client load and configuration
1516 Bugzilla : 12181, 12203
1517 Description: data loss for recently-modified files
1519 Details : In some cases it is possible that recently written or created
1520 files may not be written to disk in a timely manner (this should
1521 normally be within 30s unless client IO load is very high).
1522 The problem appears as zero-length files or files that are a
1523 multiple of 1MB in size after a client crash or client eviction
1524 that are missing data at the end of the file.
1526 This problem is more likely to be hit on clients where files are
1527 repeatedly created and unlinked in the same directory, clients
1528 have a large amount of RAM, have many CPUs, the filesystem has
1529 many OSTs, the clients are rebooted frequently, and/or the files
1530 are not accessed by other nodes after being written.
1532 The presence of the problem can be detected by looking at
1533 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1534 smaller than the second (nr_unused) then dirty files will not
1535 be flushed automatically to disk. "sync; sleep 10" should be
1536 run several times on the node before unmounting it to update
1537 Lustre (this is also safe to run on nodes without this problem).
1539 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1540 that can cause this same problem, so customers using that kernel
1541 also need to update the kernel in addition to Lustre. In order
1542 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1544 It is normal that files written just before a client crash (less
1545 than 30s) may not yet have been flushed to disk, even for local
1549 Frequency : frequent on thin XT3 nodes
1551 Description: UUID collision on thin XT3 Linux nodes
1552 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1553 randomly, since we relied on an insufficiently-seeded PRNG.
1558 Description: OSS hangs after "All ost request buffers busy"
1559 Details : A deadlock between quota and journal operations caused OSS
1560 hangs after printing "All ost request buffers busy."
1563 Frequency : always on liblustre builds
1565 Description: Cleanup compiler warnings on liblustre
1568 Frequency : always on liblustre builds on XT3
1570 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1574 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1575 > 32000 subdirectories
1576 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1577 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1578 subdirectory count crosses 32000. This will aid e2fsck to
1579 correctly handle more than 32000 subdirectories.
1582 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1584 Description: System hang on startup
1585 Details : This bug allowed the liblustre (e.g. catamount) client to
1586 return to the app before handling all startup RPCs. This
1587 could leave the node unresponsive to lustre network traffic
1588 and manifested as a server ptllnd timeout.
1590 Severity : enhancement
1592 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1593 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1594 boolean to control whether to print peer debug info when a
1595 client's RPC times out.
1600 Description: make O_SYNC working on 2.6 kernels
1601 Details : 2.6 kernels use different method for mark pages for write,
1602 so need add a code to lustre for O_SYNC work.
1607 Description: Failure to close file and release space on NFS
1608 Details : Put inode details into lock acquired in ll_intent_file_open.
1609 Use mdc_intent_lock in ll_intent_open to properly
1610 detect all kind of errors unhandled by mdc_enqueue.
1615 Description: proc file read during shutdown sometimes raced obd removal,
1617 Details : Add lock to prevent obd access after proc file removal.
1620 Frequency : Only for files larger than 4GB on 32-bit clients.
1622 Description: improperly doing page alignment of locks
1623 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1624 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1629 Description: RPCs being resent when they shouldn't be
1630 Details : Some RPCs that should not be resent are being resent. This
1631 can cause inconsistencies in the RPC state machine. Do not
1632 resend such requests.
1635 Frequency : rare, only with NFS export
1637 Description: Crash on NFS re-export node
1638 Details : under very unusual load conditions an assertion is hit in
1639 ll_intent_file_open()
1642 Frequency : only if OST filesystem is corrupted
1644 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1645 Details : for a short time RPCs with bulk IO are in the replay list,
1646 but replay of bulk IOs is unimplemented. If the OST filesystem
1647 is corrupted due to disk cache incoherency and then replay is
1648 started it is possible to trip an assertion. Avoid putting
1649 committed RPCs into the replay list at all to avoid this issue.
1654 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1655 Details : Large single O_DIRECT read and write calls can fail to allocate
1656 a sufficiently large buffer to process the request. In case of
1657 allocation failure the allocation is retried with a smaller
1658 buffer and broken into smaller requests.
1660 Severity : enhancement
1662 Description: Add -o localflock option to simulate old noflock behaviour.
1663 Details : This will achieve local-only flock/fcntl locks coherentness.
1668 Description: versioning check is incomplete
1669 Details : Checking the version difference of client vs. server, report
1670 error if the gap is too big.
1675 Description: add support PG_writeback bit
1676 Details : add support for PG_writeback bit for Lustre, for more carefull
1677 work with page cache in 2.6 kernel. This also fix some deadlocks
1678 and remove hack for work O_SYNC with 2.6 kernel.
1680 Severity : enhancement
1682 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1683 Details : The uninit_groups feature works in conjunction with the kernel
1684 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1685 up the pass1 processing of e2fsck. This is a read-only feature
1686 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1687 mount filesystems that have had this feature enabled.
1689 Severity : enhancement
1691 Description: Improve multi-block allocation algorithm to avoid fragmentation
1692 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1693 allocation locality and avoid filesystem fragmentation.
1695 ------------------------------------------------------------------------------
1697 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1699 * Support for kernels:
1700 2.6.9-42.0.3.EL (RHEL 4)
1701 2.6.5-7.276 (SLES 9)
1702 2.4.21-47.0.1.EL (RHEL 3)
1703 2.6.12.6 vanilla (kernel.org)
1704 2.6.16.21-0.8 (SLES10)
1705 * Recommended e2fsprogs version: 1.39.cfs2-0
1707 * The backwards-compatible /proc/sys/portals symlink has been removed
1708 in this release. Before upgrading, please ensure that you change
1709 any configuration scripts or /etc/sysctl.conf files that access
1710 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1711 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1712 in advance of the upgrade on any system running Lustre 1.4.6 or
1713 newer, since /proc/sys/lnet was added in that version.
1714 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1718 Frequency : only when quota is used
1720 Description: avoid scanning export list for quota master
1721 Details : Change the algorithms to avoid scanning export list in order
1722 to improve the efficiency.
1725 Frequency : MDS failover only, very rarely
1727 Description: "went back in time" messages on mds failover
1728 Details : The greatest transno may be lost when the current operation
1729 finishes with an error (transno==0) and the client's last_rcvd
1730 record is over-written. Save the greatest transno in the
1731 mds_last_transno for this case.
1734 Frequency : always for specific kernels and striping counts
1736 Description: client may get "Matching packet too big" without ACL support
1737 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1738 when trying to access files in certain configurations. The
1739 clients should in fact be denied when mounting because they do
1740 not understand ACLs.
1743 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1745 Description: many clients connecting with IO in progress causes connect timeouts
1746 Details : Avoid synchronous journal commits to avoid delays caused by many
1747 clients connecting/disconnecting when bulk IO is in progress.
1748 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1749 OST_IO_PORTAL to avoid delays behind potentially many pending
1753 Frequency : occasionally with multiple writers to a single file
1755 Description: shared writes to file may result in wrong size reported by stat()
1756 Details : Allow growing of kms when extent lock is cancelled
1759 Frequency : always with random mmap IO to multi-striped file
1761 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1762 Details : Only if the hole is at the end of OST object so that kms is too
1763 small. Fix is to increase kms accordingly in ll_nopage.
1766 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1768 Description: writes to a missing object would leak memory on the OST
1769 Details : If there is an inconsistency between the MDS and OST filesystems,
1770 such that the MDS references an object that doesn't exist, writes
1771 to that object will leak memory due to incorrect cleanup in the
1772 error handling path, eventually running out of memory on the OST.
1777 Description: Creating too long symlink causes lustre errors
1778 Details : Check symlink and name lengths before sending requests to MDS.
1781 Frequency : only if flock is enabled (not on by default)
1783 Description: posix locks not released on fd closure on 2.6.9+
1784 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1785 kernels, this caused such locks not to be released on fd close and
1786 then assertions on fs unmount about still used locks.
1789 Frequency : MDS failover only, very rarely
1791 Description: clients may get ASSERTION(granted_lock != NULL)
1792 Details : When request was taking a long time, and a client was resending
1793 a getattr by name lock request. The were multiple lock requests
1794 with the same client lock handle and
1795 mds_getattr_name->fixup_handle_for_resent_request found one of the
1796 lock handles but later failed with ASSERTION(granted_lock != NULL).
1801 Description: handle->h_buffer_credits > 0, assertion failure
1802 Details : h_buffer_credits is zero after truncate, causing assertion
1803 failure. This patch extends the transaction or creates a new
1807 Frequency : NFS re-export or patchless client
1808 Bugzilla : 11179, 10796
1809 Description: Crash on NFS re-export node (__d_move)
1810 Details : We do not want to hash the dentry if we don't have a lock.
1811 But if this dentry is later used in d_move, we'd hit uninitialised
1812 list head d_hash, so we just do this to init d_hash field but
1813 leave dentry unhashed.
1816 Frequency : NFS re-export or patchless client
1818 Description: NFS exports has problem with symbolic link
1819 Details : lustre client didn't properly install dentry when re-exported
1820 to NFS or running patchless client.
1823 Frequency : NFS re-export or patchless client
1825 Description: Various nfs/patchless fixes.
1826 Details : fixes reuse disconected alias for lookup process - this fixes
1827 warning "find_exported_dentry: npd != pd",
1828 fix permission error with open files at nfs.
1829 fix apply umask when do revalidate.
1832 Frequency : occasional
1834 Description: Crash on NFS re-export node
1835 Details : calling clear_page() on the wrong pointer triggered oops in
1836 generic_mapping_read().
1839 Frequency : rarely, using O_DIRECT IO
1841 Description: unaligned directio crashes client with LASSERT
1842 Details : check for unaligned buffers before trying any requests.
1845 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1847 Description: stale data returned from RAID cache
1848 Details : If only a small amount of IO is done to the RAID device before
1849 reading it again it is possible to get stale data from the RAID
1850 cache instead of reading it from disk.
1853 Frequency : always for sles10 kernel
1855 Description: sles10 support
1856 Details : ll_follow_link: compile fixes and using of nd_set_link
1857 under newer kernels.
1860 Frequency : depends on arch, kernel and compiler version, always on sles10
1863 Description: recursive or deep enough symlinks cause stack overflow
1864 Details : getting rid of large stack-allocated variable in
1868 Frequency : depends on hardware
1870 Description: lustre write performance loss in the SLES10 kernel
1871 Details : the performance loss is caused by using of write barriers in the
1872 ext3 code. The SLES10 kernel turns barrier support on by
1873 default. The fix is to undo that change for ldiskfs.
1875 ------------------------------------------------------------------------------
1877 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1879 * Support for kernels:
1880 2.6.9-42.0.3EL (RHEL 4)
1881 2.6.5-7.276 (SLES 9)
1882 2.4.21-47.0.1.EL (RHEL 3)
1883 2.6.12.6 vanilla (kernel.org)
1887 Frequency : quota enabled and large files being deleted
1889 Description: releasing more than 4GB of quota at once hangs OST
1890 Details : If a user deletes more than 4GB of files on a single OST it
1891 will cause the OST to spin in an infinite loop. Release
1892 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1897 Description: statfs data retrieved from /proc may be stale or zero
1898 Details : When reading per-device statfs data from /proc, in the
1899 {kbytes,files}_{total,free,avail} files, it may appear
1900 as zero or be out of date.
1903 Frequency : systems with MD RAID1 external journal devices
1905 Description: lconf's call to blkid is confused by RAID1 journal devices
1906 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1907 of returning all block devices that match the journal UUID.
1910 Frequency : always, for aggregate stripe size over 4GB
1912 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1913 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1914 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1917 Frequency : always if "lfs find" used on a local file/directory
1919 Description: "lfs find" segfaults if used on a local file/directory
1920 Details : The case where a directory component was not specified wasn't
1921 handled correctly. Handle this properly.
1924 Frequency : always on ppc64
1926 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1927 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1928 perform endianness conversion.
1931 Frequency : rarely (truncate to non-zero file size after write under load)
1932 Bugzilla : 10730, 10687
1933 Description: Files padded with zeros to next 4K multiple
1934 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1935 it is possible that files that are truncated to a non-zero size
1936 immediately after being written are filled with zero bytes beyond
1937 the truncated size. No file data is lost.
1939 Severity : enhancement
1941 Description: Allow recovery/failover for liblustre clients.
1942 Details : liblustre clients were unaware of failover configurations until
1945 Severity : enhancement
1947 Description: user file locks should fail when not mounting with flock option
1948 Details : Set up an error-returning stub in ll_file_operations.lock field
1949 to prevent incorrect behaviour when client is mounted without
1950 flock option. Also, set up properly f_op->flock field for
1954 Frequency : always on ia64
1956 Description: "lfs df" loops on printing out MDS statfs information
1957 Details : The obd_ioctl_data was not initialized and in some systems
1958 this caused a failure during the ioctl that did not return
1959 an error. Initialize the struct and return an error on failure.
1962 Frequency : SLES 9 only
1964 Description: Error of copying files with lustre special EAs as root
1965 Details : Client side always return success for setxattr call for lustre
1966 special xattr (currently only "trusted.lov").
1969 Frequency : rarely on clusters with both ia64+i386 clients
1971 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1972 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1973 clients are doing concurrent writes to the same file it is
1974 possible that the ia64 clients may LASSERT because the OST
1975 extent locks are not PAGE_SIZE aligned. Ensure that grown
1976 locks are always aligned on the request boundary.
1979 Frequency : specific use, occasional
1981 Description: Overwriting in use executable truncates on-disk binary image
1982 Details : If one node attempts to overwrite an executable in use by
1983 another node, we now correctly return ETXTBSY instead of
1984 truncating the file.
1986 Severity : enhancement
1988 Description: Async OSC create to avoid the blocking unnecessarily.
1989 Details : If a OST has no remain object, system will block on the creating
1990 when need to create a new object on this OST. Now, ways use
1991 pre-created objects when available, instead of blocking on an
1992 empty osc while others are not empty. If we must block, we block
1993 for the shortest possible period of time.
1998 Description: chmod on Lustre root is propagated to other clients
1999 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2000 invalid by the follow_mount time.
2003 Frequency : liblustre clients only
2005 Description: Race in 'instant cancel' lock handling could lead to such locks
2006 never to be granted in case of SMP MDS
2007 Details : Do not destroy not yet granted but cbpending locks in
2011 Frequency : replay/resend of open
2013 Description: non null lock assetion failure in mds_intent_policy
2014 Details : Trying to replay/resend lockless open requests resulted in
2015 mds_open() returning 0 with no lock. Now it sets a flag if
2016 a lock is going to be returned.
2018 Severity : enhancement
2020 Description: Checksum enhancements
2021 Details : New checksum enhancements allow for resending RPCs that failed
2024 Severity : enhancement
2026 Description: Tunables on number of dirty pages in cacche
2027 Details : Allow to set limit on number of dirty pages cached.
2032 Description: client crash on unmount - lock still has references
2033 Details : In some error handling cases it was possible to leak a lock
2034 reference on a client while accessing a file. This was not
2035 harmful to the client during operation, but would cause the
2036 client to crash when the filesystem is unmounted.
2039 Frequency : specific case, rare
2041 Description: ETXTBSY on mds though file not in use by client
2042 Details : ETXTBSY is no longer incorrectly returned when attempting to
2043 chmod or chown a directory that the user previously tried to
2044 execute or a currently-executing binary.
2047 Frequency : extremely rare except on liblustre-based clients
2049 Description: Lustre space not freed when files are deleted
2050 Details : Clean up open-unlinked files after client eviction. Previously
2051 the unlink was skipped and the files remained as orphans.
2056 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2057 Details : In some cases it was possible to send a blocking callback to
2058 a client doing a glimpse, even though that client didn't get
2059 a lock granted. When the glimpse lock is cancelled on the OST
2060 the freed lock is left on the waiting list and corrupted the list.
2063 Frequency : all core dumps
2065 Description: Broke core dumps to lustre
2066 Details : Negative dentry may be unhashed if parent does not have UPDATE
2067 lock, but some callers, e.g. do_coredump, expect dentry to be
2068 hashed after successful create, hash it in ll_create_it.
2070 ------------------------------------------------------------------------------
2072 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2074 * Support for kernels:
2075 2.6.9-42.0.2.EL (RHEL 4)
2076 2.6.5-7.276 (SLES 9)
2077 2.4.21-40.EL (RHEL 3)
2078 2.6.12.6 vanilla (kernel.org)
2082 Frequency : always on RHEL 3
2084 Description: Number of open files grows over time
2085 Details : The number of open files grows over time, whether or not
2086 Lustre is started. This was due to a filp leak introduced
2087 by one of our kernel patches.
2089 ------------------------------------------------------------------------------
2091 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2093 * Support for kernels:
2094 2.6.9-42.EL (RHEL 4)
2095 2.6.5-7.276 (SLES 9)
2096 2.4.21-40.EL (RHEL 3)
2097 2.6.12.6 vanilla (kernel.org)
2102 Bugzilla : 5719, 9635, 9792, 9684
2103 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2104 Details : If a server is under heavy load and cannot reply to new
2105 connection requests before the client resends the (re)connect,
2106 the connection handling code can behave badly if two service
2107 threads are concurrently handing separate (re)connections from
2108 the same client. Add better locking to the connection handling
2109 code, and ensure that only a single connection will be processed
2110 for a given client UUID, even if the lock is dropped.
2112 Severity : enhancement
2114 Description: add TCP zero-copy support to kernel
2115 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2116 sends if the hardware supports scatter-gather and checksumming.
2117 This allows socklnd to do client-write and server-read more
2118 efficiently and reduce CPU utilization from skbuf copying.
2121 Frequency : only if NFS exporting from client
2123 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2124 Details : If setting attributes on a file created under NFS that had
2125 never been opened it would be possible to oops the client
2126 if the file had no objects.
2129 Frequency : always for liblustre
2131 Description: liblustre client does MDS+OSTs setattr RPC for each write
2132 Details : When doing a write from a liblustre client, the client
2133 incorrectly issued an RPC to the MDS and each OST the file was
2134 striped over in order to update the timestamps. When writing
2135 with small chunks and many clients this could overwhelm the MDS
2136 with RPCs. In all cases it would slow down the write because
2137 these RPCs are unnecessary.
2139 Severity : enhancement
2141 Description: allow number of MDS service threads to be changed at module load
2142 Details : It is now possible to change the number of MDS service threads
2143 running. Adding "options mds mds_num_threads={N}" to the MDS's
2144 /etc/modprobe.conf will set the number of threads for the next
2145 time Lustre is restarted (assuming the "mds" module is also
2146 reloaded at that time). The default number of threads will
2147 stay the same, 32 for most systems.
2152 Description: OST crash if filesystem is unformatted or corrupt
2153 Details : If an OST is started on a device that has never been formatted
2154 or if the filesystem is corrupt and cannot even mount then the
2155 error handling cleanup routines would dereference a NULL pointer.
2160 Description: NULL pointer deref in llap_from_page.
2161 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2162 incorrect) mapping if the page was truncated/reclaimed while it was
2163 searched for. Check for this condition and skip such pages when
2164 doing readahead. Introduce extra check to llap_from_page() to
2165 verify page->mapping->host is non-NULL (so page is not anonymous).
2168 Frequency : Sometimes when using sys_sendfile
2170 Description: "page not covered by a lock" warnings from ll_readpage
2171 Details : sendfile called ll_readpage without right page locks present.
2172 Now we introduced ll_file_sendfile that does necessary locking
2173 around call to generic_file_sendfile() much like we do in
2177 Frequency : with certain MDS communication failures at client mount time
2179 Description: NULL pointer deref after failed client mount
2180 Details : a client connection request may delayed by the network layer
2181 and not be sent until after the PTLRPC layer has timed out the
2182 request. If the client fails the mount immediately it will try
2183 to clean up before the network times out the request. Add a
2184 reference from the request import to the obd device and delay
2185 the cleanup until the network drops the request.
2188 Frequency : occasionally during client (re)connect
2190 Description: assertion failure during client (re)connect
2191 Details : processing a client connection request may be delayed by the
2192 client or server longer than the client connect timeout. This
2193 causes the client to resend the connection request. If the
2194 original connection request is replied in this interval, the
2195 client may trip an assertion failure in ptlrpc_connect_interpret()
2196 which thought it would be the only running connect process.
2199 Frequency : only with obd_echo servers and clients that are rebooted
2201 Description: kernel BUG accessing uninitialized data structure
2202 Details : When running an obd_echo server it did not start the ping_evictor
2203 thread, and when a client was evicted an uninitialized data
2204 structure was accessed. Start the ping_evictor in the RPC
2205 service startup instead of the OBD startup.
2207 Severity : enhancement
2208 Bugzilla : 10193 (patchless)
2209 Description: Remove dependency on various unexported kernel interfaces.
2210 Details : No longer need reparent_to_init, exit_mm, exit_files,
2211 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2214 Frequency : rare (only users of deprecated and unsupported LDAP config)
2216 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2217 Details : LDAP apparently contains 'lustreName' attributes instead of
2218 'name'. A simple remapping of the name is sufficient.
2221 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2223 Description: waiting_locks_callback trips kernel BUG if client is evicted
2224 Details : Running with the dump_on_timeout debug flag turned on makes
2225 it possible that the waiting_locks_callback() can try to dump
2226 the Lustre kernel debug logs from an interrupt handler. Defer
2227 this log dumping to the expired_lock_main() thread.
2229 Severity : enhancement
2231 Description: Support NFS exporting on 2.6 kernels.
2232 Details : Implement non-rawops metadata methods for NFS server to use without
2233 changing NFS server code.
2236 Frequency : very rare (synthetic metadata workload only)
2238 Description: two racing renames might cause an MDS thread to deadlock
2239 Details : Running the "racer" program may cause one MDS thread to rename
2240 a file from being the source of a rename to being the target of
2241 a rename at exactly the same time that another thread is doing
2242 so, and the second thread has already enqueued these locks after
2243 doing a lookup of the target and is trying to relock them in
2244 order. Ensure that we don't try to re-lock the same resource.
2247 Frequency : only very large systems with liblustre clients
2249 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2250 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2251 all OSTs in parallel. This allows the eviction of stale liblustre
2252 clients to proceed much faster than if they were done in series,
2253 and also offers similar improvements for other set_info RPCs.
2258 Description: excessive CPU usage during initial read phase on client
2259 Details : During the initial read phase on a client, it would agressively
2260 retry readahead on the file, consuming too much CPU and impacting
2261 performance (since 1.4.5.8). Improve the readahead algorithm
2262 to avoid this, and also improve some other common cases (read
2263 of small files in particular, where "small" is files smaller than
2264 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2269 Description: MDS crash when receiving packet with unknown intent.
2270 Details : Do not LBUG in unknown intent case, just return -EFAULT
2272 Severity : enhancement
2273 Bugzilla : 9293, 9385
2274 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2275 Details : Do not serialize getattr (non-intent version) and statfs.
2278 Frequency : occasional, when OST network is overloaded/intermittent
2280 Description: client evicted by OST after bulk IO timeout
2281 Details : If a client sends a bulk IO request (read or write) the OST
2282 may evict the client if it is unresposive to its data GET/PUT
2283 request. This is incorrect if the network is overloaded (takes
2284 too long to transfer the RPC data) or dropped the OST GET/PUT
2285 request. There is no need to evict the client at all, since
2286 the pinger and/or lock callbacks will handle this, and the
2287 client can restart the bulk request.
2290 Frequency : Always when mmapping file with no objects
2292 Description: client crashes when mmapping file with no objects
2293 Details : Check that we actually have objects in a file before doing any
2294 operations on objects in ll_vm_open, ll_vm_close and
2300 Description: Request leak when working with deleted CWD
2301 Details : Introduce advanced request refcount tracking for requests
2302 referenced from lustre intent.
2304 Severity : Enhancement
2306 Description: Cache open file handles on client.
2307 Details : MDS now will return special lock along with openhandle, if
2308 requested and client is allowed to hold openhandle, even if unused,
2309 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2310 closing files for every read/write openration.
2312 Severity : Enhancement
2314 Description: Cache open negative dentries on client when possible.
2315 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2316 negative dentries on lock revocation.
2321 Description: Remounting a client read-only wasn't possible with a zconf mount
2322 Details : It wasn't possible to remount a client read-only with llmount.
2324 Severity : enhancement
2325 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2326 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2327 MPICH is included to add Lustre-specific ADIO interfaces.
2328 This is based closely on the UFS ADIO layer and only differs
2329 in file creation, in order to allow the OST striping to be set.
2330 This is user-contributed code and not supported by CFS.
2335 Description: extended inode attributes (immutable, append-only) work improperly
2336 when 2.4 and 2.6 kernels are used on client/server or vice versa
2337 Details : Introduce kernel-independent values for these flags.
2339 Severity : enhancement
2342 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2343 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2344 granularity. Now, Fractional MB granularity is be supported,
2345 this is very useful for low memory system.
2347 Severity : enhancement
2349 Description: Getattr by fid
2350 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2351 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2352 have locks and avoids getattr rpc requests for it on every stat.
2355 Frequency : Always, for filesystems larger than 2TB
2357 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2358 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2359 filesystems larger than 2TB to be mounted with mballoc enabled.
2362 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2364 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2365 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2366 filesystems larger than 2TB is mounted then the kernel will
2367 silently corrupt the start of the filesystem. CONFIG_LBD is
2368 enabled for all CFS-supported kernels, but the possibility of
2369 this happening with a modified kernel config exists.
2371 Severity : enhancement
2373 Description: add client O_DIRECT support for 2.6 kernels
2374 Details : It is now possible to do O_DIRECT reads and writes to files
2375 in the Lustre client mountpoint on 2.6 kernel clients.
2377 Severity : enhancement
2379 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2380 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2381 parallel, not waiting for response from every OST before sending
2382 a rpc to the next OST.
2387 Description: setattr vs write race when updating file timestamps
2388 Details : Client processes that update a file timestamp into the past
2389 right after writing to the file (e.g. tar) it is possible that
2390 the updated file modification time can be reset to the current
2391 time due to a race between processing the setattr and write RPC.
2393 Severity : enhancement
2395 Description: Bring 'lfs find' closer in line with regular Linux find.
2396 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2397 -print0 options and obtains all the needed info through the lustre
2400 Severity : enhancement
2402 Description: support up to 1024 configured devices on one node
2403 Details : change obd_dev array from statically allocated to dynamically
2404 allocated structs as they are first used to reduce memory usage
2409 Description: Flush dirty partially truncated pages during truncate
2410 Details : Immediatelly flush partially truncated pages in filter_setattr,
2411 this way we completely avoid having any pages in page cache on OST
2412 and can retire ugly workarounds during writes to flush such pages.
2417 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2418 Details : protect inode from truncation within vfs_unlink() context
2419 just take a reference before calling vfs_unlink() and release it
2420 when parent's i_sem is free.
2425 Description: last_id value checked outside lock on OST caused LASSERT failure
2426 Details : If there were multiple MDS->OST object precreate requests in
2427 flight, it was possible that the OST's last object id was checked
2428 outside a lock and incorrectly tripped an assertion. Move checks
2429 inside locks, and discard old precreate requests.
2432 Frequency : always, if extents are used on OSTs
2434 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2435 Details : OSTs using the extents format would not zero the high 16 bits of
2436 the index physical block number. This is not a problem for any
2437 OST filesystems smaller than 16TB, and no kernels support ext3
2438 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2439 new/modified files) and can be fixed for existing filesystems
2440 with e2fsprogs-1.39-cfs1.
2445 Description: import connection selection may be incorrect if timer wraps
2446 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2447 import connections to be ignored if the 32-bit jiffies counter
2448 wraps. Use a 64-bit jiffies counter.
2451 Frequency : very large clusters immediately after boot
2453 Description: LNET request buffers exhausted under heavy short-term load
2454 Details : If a large number of client requests are generated on a service
2455 that has previously never seen so many requests it is possible
2456 that the request buffer growth cannot keep up with the spike in
2457 demand. Instead of dropping incoming requests, they are held in
2458 the LND until the RPC service can accept more requests.
2461 Frequency : Sometimes during replay
2463 Description: Assertion failure in ll_local_open after replay.
2464 Details : If replay happened on an open request reply before we were able
2465 to set replay handler, reply will become not swabbed tripping the
2466 assertion in ll_local_open. Now we set the handler right after
2467 recognising of open request
2470 Frequency : very rare
2472 Description: kernel reports "badness in vsnprintf"
2473 Details : Reading from the "recovery_status" /proc file in small chunks
2474 may cause a negative length in lprocfs_obd_rd_recovery_status()
2475 call to vsnprintf() (which is otherwise harmless). Exit early
2476 if there is no more space in the output buffer.
2478 Severity : enhancement
2480 Description: clear OBD RPC statistics by writing to them
2481 Details : It is now possible to clear the OBD RPC statistics by writing
2482 to the "stats" file.
2487 Description: Client mtime is not the same on different clients after utimes
2488 Details : In some cases, the client was using the utimes() syscall on
2489 a file cached on another node. The clients now validate the
2490 ctime from the MDS + OSTs to determine which one is right.
2495 Description: Inability to activate failout mode
2496 Details : lconf script incorrectly assumed that in pythong string's numeric
2497 value is used in comparisons.
2500 Frequency : always with multiple stripes per file
2502 Description: Inefficient object allocation for mutli-stripe files
2503 Details : When selecting which OSTs to stripe files over, for files with
2504 a stripe count that divides evenly into the number of OSTs,
2505 the MDS is always picking the same starting OST for each file.
2506 Return the OST selection heuristic to the original design.
2511 Description: mount failures may take full timeout to return an error
2512 Details : Under some heavy load conditions it is possible that a
2513 failed mount can wait for the full obd_timeout interval,
2514 possibly several minutes, before reporting an error.
2515 Instead return an error as soon as the status is known.
2517 Frequency : quota enabled and large files being deleted
2519 Description: releasing more than 4GB of quota at once hangs OST
2520 Details : If a user deletes more than 4GB of files on a single OST it
2521 will cause the OST to spin in an infinite loop. Release
2522 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2527 Description: statfs data retrieved from /proc may be stale or zero
2528 Details : When reading per-device statfs data from /proc, in the
2529 {kbytes,files}_{total,free,avail} files, it may appear
2530 as zero or be out of date.
2533 Frequency : systems with MD RAID1 external journal devices
2535 Description: lconf's call to blkid is confused by RAID1 journal devices
2536 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2537 of returning all block devices that match the journal UUID.
2540 Frequency : always, for aggregate stripe size over 4GB
2542 Description: assertion fails when trying to use 4GB stripe size
2543 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2544 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2547 Frequency : always on ppc64
2549 Description: the first write on an ext3 filesystem with mballoc got stuck
2550 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2551 perform endianness conversion.
2553 ------------------------------------------------------------------------------
2555 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2557 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2558 INTEROPERATE with older versions automatically. Please read the
2559 user documentation before upgrading any part of a live system.
2560 * WARNING: Lustre networking configuration changes are required with
2561 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2564 * Support for kernels:
2565 2.6.9-22.0.2.EL (RHEL 4)
2566 2.6.5-7.244 (SLES 9)
2567 2.6.12.6 vanilla (kernel.org)
2570 Severity : enhancement
2571 Bugzilla : 7981/8208
2572 Description: Introduced Lustre Networking (LNET)
2573 Details : LNET is new networking infrastructure for Lustre, it includes
2574 a reorganized network configuration mode (see the user
2575 documentation for full details) as well as support for routing
2576 between different network fabrics. Lustre Networking Devices
2577 (LNDs) for the supported network fabrics have also been
2578 created for this new infrastructure.
2580 Severity : enhancement
2581 Description: Introduced Access control lists
2582 Details : clients can set ACLs on files and directories in order to have
2583 more fine-grained permissions than the standard Unix UGO+RWX.
2584 The MDS must be started with the "-o acl" mount option.
2586 Severity : enhancement
2587 Description: Introduced filesystem quotas
2588 Details : Administrators may now establish per-user quotas on the
2591 Severity : enhancement
2593 Description: Configuration change for the XT3
2594 The PTLLND is now used to run Lustre over Portals on the XT3
2595 The configure option(s) --with-cray-portals are no longer used.
2596 Rather --with-portals=<path-to-portals-includes> is used to
2597 enable building on the XT3. In addition to enable XT3 specific
2598 features the option --enable-cray-xt3 must be used.
2603 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2604 Details : A race between allocating a new llcd and re-getting the llcd_lock
2605 allowed another thread to grab newly-allocated llcd.
2607 Severity : enhancement
2609 Description: 2.6 OST async journal commit and locking fix to improve performance
2610 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2611 2.6 kernels are now async as they already were in 2.4 kernels so
2612 that they can commit concurrently with the network bulk transfer.
2613 For block-allocated files the filter allocation semaphore is held
2614 to avoid filesystem fragmentation during allocation. BKL lock
2615 removed for 2.6 xattr operations where it is no longer needed.
2620 Description: lconf incorrectly determined whether two IP networks could talk
2621 Details : In some more complicated routing and multiple-network
2622 configurations, lconf will avoid trying to make a network
2623 connection to a disjoint part of the IP space. It was doing the
2624 math incorrectly for one set of cases.
2629 Description: Fix for potential infinite loop processing records in an llog.
2630 Details : If an llog record is corrupted/zeroed, it is possible to loop
2631 forever in llog_process(). Validate the llog record length
2632 and skip the remainder of the block on error.
2635 Frequency : occasional (liblustre only)
2637 Description: liblustre could not open files whose last component is a symlink
2638 Details : sysio_path_walk() would incorrectly pass the open intent to
2639 intermediate path components.
2642 Frequency : rare (liblustre only with non-standard tuning)
2643 Bugzilla : 7201 (7350)
2644 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2645 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2646 after it has been released, only do this for referenced locks
2648 Severity : enhancement
2650 Description: specifying an (invalid) directory default stripe_size of -1
2651 would reset the directory default striping
2652 Details : stripe_size -1 was used internally to signal directory stripe
2653 removal, now use "all default" to signal dir stripe removal
2654 as a directory striping of "all default" is not useful
2657 Frequency : common for large clusters running liblustre clients
2659 Description: doing an ls when liblustre clients are running is slow
2660 Details : sending a glimpse AST to a liblustre client waits for every AST
2661 to time out, as liblustre clients will not respond. Since they
2662 cannot cache data we refresh the OST lock LVB from disk instead.
2664 Severity : enhancement
2666 Description: doing an ls at the same time as file IO can be slow
2667 Details : enqueue and other "small" requests can be blocked behind many
2668 large IO requests. Create a new OST IO portal for non-IO
2669 requests so they can be processed faster.
2672 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2674 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2675 created file with mode 000
2676 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2677 allow an ftruncate() as a non-root user to a file with mode 000.
2678 Lustre now respects this flag to disable mode checks when
2679 truncating a file owned by the user
2682 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2685 Description: Revoking locks from clients that went dead or catatonic might take
2687 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2688 cancellation of such locks instant on servers without waiting for
2689 any reply from clients. Clients drops these locks when cancel
2690 notification from server is received without replying.
2693 Frequency : liblustre-only, when liblustre client dies or becomes busy
2695 Description: Doing ls on Linux clients can take a long time with active
2697 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2698 granting such locks to it in the first place if possible. Locks
2699 are taken by proxy on the OST during the read or write and
2700 dropped immediately afterward. Add connect flags handling, do
2701 not grant locks to liblustre clients for glimpse ASTs.
2703 Severity : enhancement
2705 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2706 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2707 sequential reads and behaves badly with applications doing random
2708 reads. Improve it by reading ahead at least read region, and
2709 avoiding excessive large RPC for small reads.
2711 Severity : enhancement
2713 Description: Creating more than 1000 files for a single job may cause a load
2714 imbalance on the OSTs if there are also a large number of OSTs.
2715 Details : qos_prep_create() uses an OST index reseed value that is an
2716 even multiple of the number of available OSTs so that if the
2717 reseed happens in the middle of the object allocation it will
2718 still utilize the OSTs as uniformly as possible.
2723 Description: OST or MDS may oops in ping_evictor_main()
2724 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2725 but doesn't restart at beginning of obd_exports_timed list
2728 Severity : enhancement
2730 Description: improve by-nid export eviction on the MDS and OST
2731 Details : allow multiple exports with the same NID to be evicted at one
2732 time without re-searching the exports list.
2735 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2737 Description: MDS may oops in groups_free()
2738 Details : in rare race conditions a newly allocated group_info struct is
2739 freed again, and this can be NULL. The 2.4 compatibility code
2740 for groups_free() checked for a NULL pointer, but 2.6 did not.
2743 Frequency : common for liblustre clients doing little filesystem IO
2744 Bugzilla : 9352, 7313
2745 Description: server may evict liblustre clients accessing contended locks
2746 Details : if a client is granted a lock or receives a completion AST
2747 with a blocking AST already set it would not reply to the AST
2748 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2751 Frequency : lfs setstripe, only systems with more than 160 OSTs
2753 Description: unable to set striping with a starting offset beyond OST 160
2754 Details : llapi_create_file() incorrectly limited the starting stripe
2755 index to the maximum single-file stripe count.
2758 Frequency : LDAP users only
2760 Description: lconf did not handle in-kernel recovery with LDAP properly
2761 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2763 Severity : enhancement
2765 Description: bind OST threads to NUMA nodes to improve performance
2766 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2767 node and do their allocations there to localize memory access
2769 Severity : enhancement
2771 Description: llmount can determine client NID directly from Myrinet (GM)
2772 Details : the client NID code from gmnalnid was moved directly into
2773 llmount, removing the need to use this or specifying the
2774 client NID explicitly when mounting GM clients with zeroconf
2777 Frequency : if client is started with down MDS
2779 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2780 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2781 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2782 ptlrpc_import_delayed_req() because we didn't check if the
2783 request was interrupted, and we also didn't break out of the
2784 event loop if there was no timeout
2789 Description: data loss during non-page-aligned writes to a single file from
2790 both multiple nodes and multiple threads on one node at same time
2791 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2792 inconsistency led to false short-reads, that were cached and later
2793 used by ->prepare_write() to fill in partially written page,
2794 leading to data loss.
2797 Frequency : always, if lconf --abort_recovery used
2799 Description: lconf --abort_recovery fails with 'Operation not supported'
2800 Details : lconf was attempting to abort recovery on the MDT device and not
2803 Severity : enhancement
2805 Description: remove cleanup logs
2806 Details : replace lconf-generated cleanup logs with lustre internal
2807 cleanup routines. Eliminates the need for client-cleanup and
2810 Severity : enhancement
2812 Description: add support for EAs (user and system) on lustre filesystems
2813 Details : it is now possible to store extended attributes in the Lustre
2814 client filesystem, and with the user_xattr mount option it
2815 is possible to allow users to store EAs on their files also
2817 Severity : enhancement
2819 Description: Add possibility (config option) to show minimal available OST free
2821 Details : When compiled with --enable-mindf configure option, statfs(2)
2822 (and so, df) will return least minimal free space available from
2823 all OSTs as amount of free space on FS, instead of summary of
2824 free spaces of all OSTs.
2826 Severity : enhancement
2828 Description: do not expand extent locks acquired on OST-side
2829 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2830 by server: they are not cached anyway.
2833 Frequency : when mmap is used/binaries executed from Lustre
2835 Description: Unmmap pages before throwing them away from read cache.
2836 Details : llap_shrink cache now attempts to unmap pages before discarding
2837 them (if unmapping failed - do not discard). SLES9 kernel has
2838 extra checks that trigger if this unmapping is not done first.
2843 Description: lconf didn't resolve symlinks before checking to see whether a
2844 given mountpoint was already in use
2847 Frequency : when migrating failover services
2848 Bugzilla : 6395, 9514
2849 Description: When migrating a subset of services from a node (e.g. failback
2850 from a failover service node) the remaining services would
2851 time out and evict clients.
2852 Details : lconf --force (implied by --failover) sets the global obd_timeout
2853 to 5 seconds in order to quickly disconnect, but this caused
2854 other RPCs to time out too quickly. Do not change the global
2855 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2857 Severity : enhancement
2858 Frequency : if MDS is started with down OST
2859 Bugzilla : 9439,5706
2860 Description: Allow startup/shutdown of an MDS without depending on the
2861 availability of the OSTs.
2862 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2863 Add appropriate locking and lov-osc refcounts for safe
2864 cleaning. Add osc abort_inflight calls in case the
2865 synchronize never started.
2868 Frequency : occasional (Cray XT3 only)
2870 Description: root not authorized to access files in CRAY_PORTALS environment
2871 Details : The client process capabilities were not honoured on the MDS in
2872 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2873 been accessed by an authorized user then root was able to access
2874 the file on the local client also. The root user capabilities
2875 are now allowed on the MDS, as this environment has secure UID.
2878 Frequency : occasional
2880 Description: ldiskfs "too long searching" message happens too often
2881 Details : A debugging message (otherwise harmless) prints too often on
2882 the OST console. This has been reduced to only happen when
2883 there are fragmentation problems on the filesystem.
2888 Description: Division by zero in statfs when all OSCs are inactive
2889 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2890 lov_statfs divides by value returned by lov_get_stripecnt().
2894 Bugzilla : 9489, 3273
2895 Description: First write from each client to each OST was only 4kB in size,
2896 to initialize client writeback cache, which caused sub-optimal
2897 RPCs and poor layout on disk for the first writen file.
2898 Details : Clients now request an initial cache grant at (re)connect time
2899 and so that they can start streaming writes to the cache right
2900 away and always do full-sized RPCs if there is enough data.
2901 If the OST is rebooted the client also re-establishes its grant
2902 so that client cached writes will be honoured under the grant.
2907 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2908 Details : Now I/O RPCs go to different portal number and (presumably) fast
2909 lock requests (and glimses) and other RPCs get their own service
2910 threads pool that should be able to service those RPCs
2913 Severity : enhancement
2915 Description: Ability to exchange lustre version between client and servers and
2916 issue warnings at client side if client is too old. Also for
2917 liblustre clients there is ability to refuse connection of too old
2919 Details : New 'version' field is added to connect data structure that is
2920 filled with version info. That info is later checked by server and
2924 Frequency : rare, liblustre only.
2925 Bugzilla : 9296, 9581
2926 Description: Two simultaneous writes from liblustre at offset within same page
2927 might proceed at the same time overwriting eachother with stale
2929 Details : I/O lock withing llu_file_prwv was released too early, before data
2930 actually was hitting the wire. Extended lock-holding time until
2931 server acknowledges receiving data.
2934 Frequency : extremely rare. Never observed in practice.
2936 Description: avoid generating lustre_handle cookie of 0.
2937 Details : class_handle_hash() generates handle cookies by incrementing
2938 global counter, and can hit 0 occasionaly (this is unlikely, but
2939 not impossible, because initial value of cookie counter is
2940 selected randonly). Value of 0 is used as a sentinel meaning
2941 "unassigned handle" --- avoid it. Also coalesce two critical
2942 sections in this function into one.
2944 Severity : enhancement
2946 Description: allow liblustre clients to delegate truncate locking to OST
2947 Details : To avoid overhead of locking, liblustre client instructs OST to
2948 take extent lock in ost_punch() on client's behalf. New connection
2949 flag is added to handle backward compatibility.
2951 Severity : enhancement
2952 Bugzilla : 4928, 7341, 9758
2953 Description: allow number of OST service threads to be specified
2954 Details : a module parameter allows the number of OST service threads
2955 to be specified via "options ost ost_num_threads={N}" in the
2956 OSS's /etc/modules.conf or /etc/modprobe.conf.
2960 Bugzilla : 6146, 9635, 9895
2961 Description: servers crash with bad pointer in target_handle_connect()
2962 Details : In rare cases when a client is reconnecting it was possible that
2963 the connection request was the last reference for that export.
2964 We would temporarily drop the export reference and get a new
2965 one, but this may have been the last reference and the export
2966 was just destroyed. Get new reference before dropping old one.
2968 Severity : enhancement
2969 Frequency : if client is started with failover MDS
2971 Description: Allow multiple MDS hostnames in the mount command
2972 Details : Try to read the configuration from all specified MDS
2973 hostnames during a client mount in case the "primary"
2976 Severity : enhancement
2978 Description: Stop sending data to evicted clients as soon as possible.
2979 Details : Check if the client we are about to send or are sending data to
2980 was evicted already. (Check is done every second of waiting,
2981 for which l_wait_event interface was extended to allow checking
2982 of exit condition at specified intervals).
2985 Frequency : rare, normally only when NFS exporting is done from client
2987 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2989 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2992 Frequency : always, if srand() is called before liblustre initialization
2994 Description: Liblustre uses system PRNG disturbing its usage by user application
2995 Details : Introduce internal to lustre fast and high-quality PRNG for
2996 lustre usage and make liblustre and some other places in generic
2997 lustre code to use it.
2999 Severity : enhancement
3000 Bugzilla : 9477, 9557, 9870
3001 Description: Verify that the MDS configuration logs are updated when xml is
3002 Details : Check if the .xml configuration logs are newer than the config
3003 logs stored on the MDS and report an error if this is the case.
3004 Request --write-conf, or allow starting with --old_conf.
3006 Severity : enhancement
3008 Description: Handle symlinks in the path when checking if Lustre is mounted.
3009 Details : Resolve intermediate symlinks when checking if a client has
3010 mounted a filesystem to avoid duplicate client mounts.
3015 Description: lconf can hit an error exception but still return success.
3016 Details : The lconf command catches the Command error exception at the top
3017 level script context and will exit with the associated exit
3018 status, but doesn't ensure that this exit status is non-zero.
3023 Description: failure of ptlrpc thread startup can cause oops
3024 Details : Starting a ptlrpc service thread can fail if there are a large
3025 number of threads or the server memory is very fragmented.
3026 Handle this without oopsing.
3029 Frequency : always, only if liblustre and non-default acceptor port was used
3031 Description: liblustre cannot connect to servers with non-default acceptor port
3032 Details : tcpnal_set_default_params() was not called and was therefore
3033 ignoring the environment varaible TCPNAL_PORT, as well as other
3034 TCPNAL_ environment variables
3039 Description: two objects could be created on the same OST for a single file
3040 Details : If an OST is down, in some cases it was possible to create two
3041 objects on a single OST for a single file. No problems other
3042 than potential performance impact and spurious error messages.
3046 Bugzilla : 5681, 9562
3047 Description: Client may oops in ll_unhash_aliases
3048 Details : Client dcache may become inconsistent in race condition.
3049 In some cases "getcwd" can fail if the current directory is
3055 Description: Inode refcounting problems in NFS export code
3056 Details : link_raw functions used to call d_instantiate without obtaining
3057 extra inode reference first.
3061 Bugzilla : 9942, 9903
3062 Description: Referencing freed requests leading to crash, memleaks with NFS.
3063 Details : We used to require that call to ll_revalidate_it was always
3064 followed by ll_lookup_it. Also with revalidate_special() it is
3065 possible to call ll_revalidate_it() twice for the same dentry
3066 even if first occurence returned success. This fix changes semantic
3067 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3068 reference on a request referred from the intent.
3069 ll_intent_release() then releases such a request.
3072 Frequency : rare, normally benchmark loads only
3074 Description: unlinked inodes were kept in memory on the client
3075 Details : If a client is repeatedly creating and unlinking files it
3076 can accumulate a lot of stale inodes in the inode slab cache.
3077 If there is no other client load running this can cause the
3078 client node to run out of memory. Instead flush old inodes
3079 from client cache that have the same inode number as a new inode.
3082 Frequency : SLES9 2.6.5 kernel and long filenames only
3083 Bugzilla : 9969, 10379
3084 Description: utime reports stale NFS file handle
3085 Details : SLES9 uses out-of-dentry names in some cases, which confused
3086 the lustre dentry revalidation. Change it to always use the
3090 Frequency : rare, unless heavy write-truncate concurrency is continuous
3091 Bugzilla : 4180, 6984, 7171, 9963, 9331
3092 Description: OST becomes very slow and/or deadlocked during object unlink
3093 Details : filter_destroy() was holding onto the parent directory lock
3094 while truncating+unlinking objects. For very large objects this
3095 may block other threads for a long time and slow overall OST
3096 responsiveness. It may also be possible to get a lock ordering
3097 deadlock in this case, or run out of journal credits because of
3098 the combined truncate+unlink. Solution is to do object truncate
3099 first in one transaction without parent lock, and then do the
3100 final unlink in a new transaction with the parent lock. This
3101 reduces the lock hold time dramatically.
3104 Frequency : rare, 2.4 kernels only
3106 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3107 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3108 before unmounting their filesystem in order to avoid deadlock.
3109 The kernel_locked() function in 2.4 kernels only checks whether
3110 the kernel lock is held, not whether it is this process that is
3111 holding it as 2.6 kernels do.
3116 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3117 Details : The client ptlrpc code may be trying to reconnect to a down
3118 server before a previous connection attempt has timed out.
3119 Increase the reconnect interval to be longer than the connection
3120 timeout interval to avoid sending duplicate connections to
3124 Frequency : echo_client brw_test command
3126 Description: fix echo_client to work with OST preallocated code
3127 Details : OST preallocation code (5137) didn't take echo_client IO path
3128 into account: echo_client calls filter methods outside of any
3129 OST thread and, hence, there is no per-thread preallocated
3130 pages and buffers to use. Solution: hijack pga pages for IO. As
3131 a byproduct, this avoids unnecessary data copying.
3135 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3136 Description: Client can oops in mdc_commit_close() after open replay
3137 Details : It was possible for the MDS to return an open request with no
3138 transaction number in mds_finish_transno() if the client was
3139 evicted, but without actually returning an error. Clients
3140 would later try to replay that open and may trip an assertion
3141 Simplify the client close codepath, and always return an error
3142 from the MDS in case the open is not successful.
3145 Frequency : rare, 2.6 OSTs only
3147 Description: OST may deadlock under high load on fragmented files
3148 Details : If there was a heavy load and highly-fragmented OST filesystems
3149 it was possible to have all the OST threads deadlock waiting on
3150 allocation of biovecs, because the biovecs were not released
3151 until the entire RPC IO was completed. Instead, release biovecs
3152 as soon as they are complete to ensure forward IO progress.
3154 Severity : enhancement
3156 Description: Support for specifying external journal device at mount
3157 Details : If an OST or MDS device is formatted with an external journal
3158 device, this device major/minor is stored in the ext3 superblock
3159 and may not be valid for failover. Allow detecting and
3160 specifying the external journal at mount time.
3165 Description: Mounting an MDS with pending unlinked files may cause oops
3166 Details : target_finish_recovery() calls mds_postrecov() which returned
3167 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3168 considers this an error and immediately begins cleaning up the
3169 lov, just after starting the mds_lov process
3171 Severity : enhancement
3173 Description: Implement 'lfs df' to report actual free space on per-OST basis
3174 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3175 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3176 usage of objects; '-h' to report in human readable format.
3178 ------------------------------------------------------------------------------
3180 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
3187 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3188 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3189 at startup during orphan unlinking. Instead, skip buddy update
3190 before setup, as it will be regenerated after recovery is complete.
3195 Description: If an OST is inactive, its locks might reference stale inodes.
3196 Details : lov_change_cbdata() must iterate over all namespaces, even if
3197 they are inactive to clear inode references from the lock.
3199 Severity : enhancement
3200 Frequency : occasional, if non-standard max_dirty_mb used
3202 Description: Client will block write RPCs if not enough grant
3203 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3204 then the client will block writes while waiting for another RPC
3205 to complete instead of consuming its dirty limit. With change
3206 we get improved performance when max_dirty_mb is small.
3208 Severity : enhancement
3209 Bugzilla : 3389, 6253
3210 Description: Add support for supplementary groups on the MDS.
3211 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3212 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3213 lookups for user supplementary groups into a cache.
3217 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3218 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3219 for liblustre O_EXCL is enforced by MDS.
3223 Description: readdir never returns NULL in liblustre.
3224 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3225 dentry in struct dirent.
3229 Frequency : liblustre only, depends on application IO pattern
3230 Description: liblustre clients evicted if not contacting servers
3231 Details : Don't put liblustre clients into the ping_evictor list, so
3232 they will not be evicted by the pinger ever.
3234 Severity : enhancement
3236 Description: Add ability to evict clients by NID from MDS.
3237 Details : By echoing "nid:$NID" string into
3238 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3239 $NID would be instantly evicted from this MDS and from all active
3240 OSTs connected to it.
3244 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3245 Details : lookup_it_finish() used to query file size from OSTs that was not
3250 Description: service threads change working directory to that of init
3251 Details : Starting lustre service threads may pin the working directory
3252 of the parent thread, making that filesystem busy. Threads
3253 now change to the working directory of init to avoid this.
3257 Frequency : during shutdown only
3258 Description: shutdown with a failed MDS or OST can cause unmount to hang
3259 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3264 Frequency : 2.6 only
3265 Description: chmod/chown may include an extra supplementary group
3266 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3267 supplementary group and if none is specified this is used.
3270 Bugzilla : 5479 (6816)
3272 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3273 Details : If lookup is in progress on a file that is unlinked we might try
3274 to revalidate the inode and fail in revalidate after lookup is
3275 complete and ll_file_open() enqueues the open again but
3276 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3279 Frequency : always, if lconf --abort_recovery used
3281 Description: lconf --abort_recovery fails with 'Operation not supported'
3282 Details : lconf was attempting to abort recovery on the MDT device and not
3285 ------------------------------------------------------------------------------
3287 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3292 Frequency : rare (only unsupported configurations with a node running as an
3294 Bugzilla : 6514, 5137
3295 Description: Mounting a Lustre file system on a node running as an OST could
3297 Details : OSTs now preallocates memory needed to write out data at
3298 startup, instead of when needed, to avoid having to
3299 allocate memory in possibly low memory situations.
3300 Specifically, if the file system is mounted on on OST,
3301 memory pressure could force it to try to write out data,
3302 which it needed to allocate memory to do. Due to the low
3303 memory, it would be unable to do so and the node would
3304 become unresponsive.
3306 Severity : enhancement
3308 Description: Addition of lconf --service command line option
3309 Details : lconf now accepts a '--service <arg>' option, which is
3310 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3312 Severity : enhancement
3314 Description: Failover mode is now the default for OSTs.
3315 Details : By default, OSTs will now run in failover mode. To return to
3316 the old behaviour, add '--failout' to the lmc line for OSTs.
3318 Severity : enhancement
3320 Description: Health checks are now provided for MDS and OSTs
3321 Details : Additional detailed health check information on MSD and OSTs
3322 is now provided through the procfs health_check value.
3325 Frequency : occasional, depends on IO load
3327 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3328 after numerous create/delete cycles
3329 Details : The ext3 inode allocation policy would not allocate new inodes
3330 very well on the OSTs because there are no new directories
3331 being created. Instead we look for groups with free space if
3332 the parent directories are nearly full.
3337 Description: Network or server problems during mount may cause partially
3338 mounted clients instead of returning an error.
3339 Details : The config llog parsing code may overwrite the error return
3340 code during mount error handling, returning success instead
3346 Description: MDS can fail to allocate large reply buffers
3347 Details : After long uptimes the MDS can fail to allocate large reply
3348 buffers (e.g. zconf client mount config records) due to memory
3349 fragmentation or consumption by the buffer cache. Preallocate
3350 some large reply buffers so that these replies can be sent even
3351 under memory pressure.
3355 Frequency : rare (liblustre)
3356 Description: fsx running with liblustre complained that using truncate() to
3357 extend the file doesn't work. This patch corrects that issue.
3358 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3359 ATTR_SIZE and lsm use in llu_setattr_raw.
3363 Frequency : rare, only 2.6 kernels
3364 Description: Unusual file access patterns on the MDS may result in inode
3365 data being lost in very rare circumstances.
3366 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3367 code in the 2.6 kernel caused the inode and/or EA data not to
3368 be read from disk, causing single-file corruption.
3372 Frequency : rare, only 2.6 filesystems using extents
3373 Description: Heavy concurrent write and delete load may cause data corruption.
3374 Details : It was possible under high-load situations to have an extent
3375 metadata block in the block device cache from a just-unlinked
3376 file overwrite a newly-allocated data block. We now unmap any
3377 metadata buffers that alias just-allocated data blocks.
3381 Frequency : filesystems with default stripe_count larger than 77
3382 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3383 Details : lconf specifies an inode size of 4096 bytes when the default
3384 stripe_count is larger than 77. This conflicts with the default
3385 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3386 this case to avoid pinning too much memory for large EAs.
3388 ------------------------------------------------------------------------------
3390 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3395 Frequency : rare (extremely heavy IO load with hundreds of clients)
3397 Description: Client is evicted, gets IO error writing to file
3398 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3399 caused clients to be evicted for AST timeouts. The fixes in
3400 bug 5192 mean we no longer need to have such short AST timeouts
3401 so ldlm_timeout has been increased.
3404 Frequency : occasional during --force or --failover shutdown under load
3405 Bugzilla : 5949, 4834
3406 Description: Server oops/LBUG if stopped with --force or --failover under load
3407 Details : a collection of import/export refcount and cleanup ordering
3408 issues fixed for safer force cleanup
3411 Frequency : only filesystems larger than 120 OSTs
3412 Bugzilla : 5990, 6223
3413 Description: lfs getstripe would oops on a very large filesystem
3414 Details : lov_getconfig used kfree on vmalloc'd memory
3417 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3418 Bugzilla : 6242, 6243
3419 Description: reading from files that had been truncated to a non-zero size
3420 but never opened returned no data
3421 Details : ll_file_read() reads zeros from no-object files to EOF
3426 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3427 Details : There's little harm in aborting MDS/OSS recovery and letting it
3428 try again, so I removed the LASSERT and return an error instead.
3430 Severity : enhancement
3432 Description: New debugging infrastructure for tracking down data corruption
3433 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3434 (b) cover more of the client-side code path, and (c) try to narrow
3435 down where problems occurred
3439 Bugzilla : 3819, 4364, 4397, 6313
3440 Description: Racing close and eviction MDS could cause assertion in mds_close
3441 Details : It was possible to get multiple mfd references during close and
3442 client eviction, leading to one thread referencing a freed mfd.
3444 Severity: : enhancement
3445 Bugzilla : 3262, 6359
3446 Description: Attempts to reconnect to servers are now more aggressive.
3447 Details : This builds on the enhanced upcall-less recovery that was added
3448 in 1.4.2. When trying to reconnect to servers, clients will
3449 now try each server in the failover group every 10 seconds. By
3450 default, clients would previously try one server every 25 seconds.
3455 Description: After recovery, certain operations trigger a failed
3456 assertion on a client.
3457 Details : Failing over an mds, using lconf -d --failover, while a
3458 client was doing a readdir() call would cause the client to
3459 LBUG after recovery completed and the readdir() was resent.
3461 Severity : enhancement
3463 Description: Default groups are now added by lconf
3464 Details : You can now run lconf --group <servicename> without having to
3465 manually add groups with lmc.
3468 Frequency : occasional
3470 Description: Nodes with an elan id of 0 trigger a failed assertion
3473 Frequency : always when accessing e.g. tty/console device nodes
3475 Description: tty and some other devices nodes cannot be used on lustre
3476 Details : file's private_data field is used by device data and lustre
3477 values in there got lost. New field was added to struct file to
3478 store fs-specific private data.
3481 Frequency : when exporting Lustre via NFS
3483 Description: NFSD failed occasionally when looking up a path component
3484 Details : NFSD is looking up ".." which was broken in ext3 directories
3485 that had grown large enough to become hashed.
3488 Frequency : Clusters with multiple interfaces not on the same subnet
3490 Description: Nodes will repeatedly try to reconnect to an interface which it
3491 cannot reach and report an error to the log.
3492 Details : Extra peer list entries will be created by lconf with some peers
3493 unreachable. lconf now validates the peer before adding it.
3496 Frequency : Only if a default stripe is set on the filesystem root.
3498 Description: Setting a default stripe on the filesystem root prevented the
3499 filesystem from being remounted.
3500 Details : The client was sending extra request flags in the root getattr
3501 request and did not allocate a reply buffer for the dir EA.
3504 Frequency : occasional, higher if lots of files are accessed by one client
3505 Bugzilla : 6159, 6097
3506 Description: Client trips assertion regarding lsm mismatch/magic
3507 Details : While revalidating inodes the VFS looks up inodes with ifind()
3508 and in rare cases can find an inode that is being freed.
3509 The ll_test_inode() code will free the lsm during ifind()
3510 when it finds an existing inode and then the VFS later attaches
3511 this free lsm to a new inode.
3515 Bugzilla : 6422, 7030
3516 Description: MDS deadlock between mkdir and client eviction
3517 Details : Creating a new file via mkdir or mknod (starting a transaction
3518 and getting the ns lock) can deadlock with client eviction
3519 (gets ns lock and trying to finish a synchronous transaction).
3522 Frequency : occasional
3523 Description: While starting a server, the fsfilt_ext3 module could not be
3525 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3526 kernels. Previously, lconf would still use the ext3 name
3527 when trying to load modules. Now, it will correctly use
3528 ext3 on 2.4 and ldiskfs on 2.6.
3530 Severity : enhancement
3531 Description: The default stripe count has been changed to 1
3532 Details : The interpretation of the default stripe count (0, to lfs
3533 or lmc) has been changed to mean striping across a single
3534 OST, rather than all available. For general usage we have
3535 found a stripe count of 1 or 2 works best.
3537 Severity : enhancement
3538 Description: Add support for compiling against Cray portals.
3539 Details : Conditional compiling for some areas that are different
3543 Frequency : occasional
3544 Bugzilla : 6409, 6834
3545 Description: Creating files with an explicit stripe count may lead to
3546 a failed assertion on the MDS
3547 Details : If some OSTs are full or unavailable, creating files may
3548 trigger a failed assertion on the MDS. Now, Lustre will
3549 try to use other servers or return an error to the
3553 Frequency : occasional
3555 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3556 caused lock timeout during readahead (since 1.4.2).
3557 Details : Processes doing readahead might match a lock that hasn't been
3558 granted yet if there are overlapping and conflicting lock
3559 requests. The readahead process waits on ungranted lock
3560 (original lock is CBPENDING), while OST waits for that process
3561 to cancel CBPENDING read lock and eventually evicts client.
3563 Severity : enhancement
3565 Description: Initial enabling of flock support for clients
3566 Details : Implements fcntl advisory locking and file status functions.
3567 This feature is provided as an optional mount flag (default
3568 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3569 locking are implemented yet, and those that are are not guaranteed
3570 to be completely correct in production environments.
3571 mount -t lustre -o [flock|noflock] ...
3574 Frequency : occasional
3576 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3577 assertion in the ext3 JBD (journaling) layer.
3578 Details : The b_committed_data struct is protected by the big kernel lock
3579 in 2.4 kernels, serializing journal_commit_transaction() and
3580 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3581 access to this struct. In 2.6 kernels there is finer grained
3582 locking to improve SMP performance of the JBD layer.
3586 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3591 Description: merge and cleanup kernel patches.
3594 -----------------------------------------------------------------------------
3596 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3598 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3599 versions of Lustre. Please update all servers and clients to
3600 version 1.4.2 or later at the same time. You must also run
3601 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3602 to update the configuration logs.
3604 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3605 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3606 - swab llog records of type '0' so we get proper header size/idx (5861)
3607 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3608 - fix rename of one directory over another leaking an inode (5953)
3609 - avoid SetPageDirty on 2.6 (5981)
3610 - don't re-add just-being-destroyed locks to the waiting list (5653)
3611 - when creating new directories, inherit the parent's custom
3612 striping settings if present parent (3048)
3613 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3614 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3615 - don't hold client locks on temporary worklist from l_lru (5666)
3616 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3617 - automatically evict dead clients (5921)
3618 - Update file size properly in create+truncate+fstat case (6196)
3619 - Do not unhash mountpoint dentries, do not allow removal of
3621 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3622 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3623 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3624 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3625 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3626 - Correctly handle reads of files with no objects (6243)
3627 - lctl recover will also mark a device active if deactivate used (5933)
3629 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3630 - allow --write-conf on an MDS with different nettype than client (5619)
3631 - don't write config llogs to MDS for mounts not from that MDS (5617)
3632 - lconf should create multiple TCP connections from a client (5201)
3633 - init scripts are now turned off by default; run chkconfig --on
3634 lustre and chkconfig --on lustrefs to use them
3635 - upcalls are no longer needed for clients to recover to failover
3637 - add --abort-recovery option to lconf to abort recovery on device
3639 - add support for an arbitrary number of OSTs (3026)
3640 - Quota support protocol changes.
3641 - forward compatibility changes to wire structs (6007)
3642 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3643 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3644 - improved "lustre status" script
3645 - initialize blocksize for non-regular files (6062)
3646 - added --disable-server and --disable-client configure options (5782)
3647 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3648 - Vanilla 2.4.29 support
3649 - increase maximum number of obd devices to 520 (6242)
3650 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3651 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3652 SLES 9 2.6.5 kernels
3653 - limit stripes per file to 160 (the maximum EA size) (6093)
3655 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3658 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3659 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3660 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3661 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3662 - small ext3 extents cleanups and fixes (5733)
3663 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3664 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3665 - avoid needless client->OST connect, fix handle mismatch (5317)
3666 - fix DLM error path that led to out-of-sync client, long delays (5779)
3667 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3668 - fix several locking issues related to i_size (5492,5624,5654,5672)
3669 - don't move pending lock onto export if it is already evicted (5683)
3670 - fix kernel oops when creating .foo in unlinked directory (5548)
3671 - fix deadlock in obdfilter statistics vs. object create (5811)
3672 - use time_{before,after} to avoid timer jiffies wrap (5882)
3673 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3674 - Do not leak request if server was not able to process it (5154)
3675 - If mds_open unable to find parent dir, make that negative lookup(5154)
3676 - don't create new directories with extent-mapping (5909, 5936)
3678 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3679 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3680 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3681 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3682 - support for RHEL4 kernel on i686 (5773)
3683 - provide error messages when incompatible logs are encountered (5898)
3685 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3686 * version 1.4.0.10 (1.4.1 release candidate 1)
3688 - don't keep a lock reference when lock is not granted (4238)
3689 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3690 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3691 - drop import inflight refcount on signal_completed_replay error (5255)
3692 - unlock page after async write error during send (3677)
3693 - handle missing objects in filter_preprw_read properly (5265)
3694 - no transno return for symlink open, don't save no-trasno open (3440)
3695 - don't try to complete elan receive that already failed (4012)
3696 - free RPC server reply state on error (5406)
3697 - clean up thread from ptlrpc_start_thread() on error (5160)
3698 - readahead could read extra page into cache that wasn't ejected (5388)
3699 - prevent races in class_attach/setup/cleanup/detach (5260)
3700 - don't dereference de->d_inode after l_dput of de (5458)
3701 - use "int" for stripe value returned from lock_to_stripe (5544)
3702 - mballoc allocation and error-checking fixes in 2.6 (5504)
3703 - block device patches to fix I/O request sizes in 2.6 (5482)
3704 - look up hostnames for IB nals (5602)
3705 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3706 - don't start multiple acceptors for the same port (5277)
3707 - fix incorrect LASSERT in mds_getattr_name (5635)
3708 - export a proc file for general "ping" checking (5628)
3709 - fix "lfs check" to not block when the MDS is down (5628)
3711 - service request history (4965)
3712 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3713 - create an "evict_client" /proc entry on OSTs, like the MDS has
3714 - fix mount usage message, return errors per mount(8) (5168)
3715 - change grep [] to grep "[]" in tests so they work in more UMLs
3716 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3717 - remove ancient LOV_MAGIC_V0 EA support (5047)
3718 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3719 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3720 - allow readahead allocations to fail when low on memory (5383)
3721 - mmap locking landed again, after considerable improvement (2828)
3722 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3724 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3727 - send OST transaction number in read/write reply to free req (4966)
3728 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3729 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3730 - the watchdog thread now runs as interruptible (5246)
3731 - flock/lockf fixes (but it's still disabled, pending 5135)
3732 - don't use EXT3 constants in llite code (5094)
3733 - memory shortage at startup could cause assertion (5176)
3735 - reorganization of lov code
3736 - single portals codebase
3738 - add extents/mballoc support (5025)
3739 - direct I/O reads in the obdfilter (4048)
3740 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3742 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3745 - send OST transaction number in read/write reply to free req (4966)
3746 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3747 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3748 - the watchdog thread now runs as interruptible (5246)
3749 - handle missing objects in filter_preprw_read properly (5265)
3750 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3751 - drop import inflight refcount on signal_completed_replay error (5255)
3752 - unlock page after async write error during send (3677)
3753 - return original error code on reconstructed replies (3761)
3754 - no transno return for symlink open, don't save no-trasno open (3440)
3756 - add pid to ldlm debugging output (4922)
3757 - bump the watchdog timeouts -- we can't handle 30sec yet
3758 - extra debugging for orphan dentry/inode bug (5259)
3760 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3763 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3764 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3765 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3766 - fix NULL dereference in /proc/sys/portals/routes (4827)
3767 - allow failed mdc_close() operations to be interrupted (4561)
3768 - stop precreate on OST before MDS would time out on it (4778)
3769 - don't send partial-page writes before EOF from client (4410)
3770 - discard client grant for sub-page writes on large-page clients (4520)
3771 - don't free dentries not owned by NFS code, check generation (4806)
3772 - fix lsm leak if mds_create_objects() fails (4801)
3773 - limit debug_daemon file size, always print CERROR messages (4789)
3774 - use transno after validating reply (3892)
3775 - process timed out requests if import state changes (3754)
3776 - update mtime on OST during writes, return in glimpse (4829)
3777 - add mkfsoptions to LDAP (4679)
3778 - use ->max_readahead method instead of zapping global ra (5039)
3779 - don't interrupt __l_wait_event() during strace
3781 - add software watchdogs to catch hung threads quickly (4941)
3782 - make lustrefs init script start after nfs is mounted
3783 - fix CWARN/ERROR duplication (4930)
3784 - return async write errors to application if possible (2248)
3785 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3786 - print NAL number in %x format (4645)
3787 - update barely-supported suse-2.4.21-171 series (4842)
3788 - support for sles 9 %post scripts
3789 - support for building 2.6 kernel-source packages
3790 - support for sles km_* packages
3792 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3795 - ignore -ENOENT errors in osc_destroy (3639)
3796 - notify osc create thread that OSC is being cleaned up (4600)
3797 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3798 - reconstruct ost_handle() like mds_handle() (4657)
3799 - create a new thread to do import eviction to avoid deadlock (3969)
3800 - let lconf resolve symlinked-to devices (4629)
3801 - don't unlink "objects" from directory with default EA (4554)
3802 - hold socknal file ref over connect in case target is down (4394)
3803 - allow more than 32000 subdirectories in a single directory (3244)
3804 - fix blocks count for O_DIRECT writes (3751)
3805 - OST returns ENOSPC from object create when no space left (4539)
3806 - don't send truncate RPC if file size isn't changing (4410)
3807 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3808 - bind to privileged port in socknal and tcpnal (3689)
3810 - rate limit CERROR/CWARN console message to avoid overload (4519)
3811 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3812 - basic mmap support (3918)
3813 - kernel patch series update from b1_4 (4711)
3815 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3818 - avoid crash during MDS cleanup with OST shut down (2775)
3819 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3820 - don't use bad inodes on the MDS (3744)
3821 - dynamic object preallocation to improve recovery speed (4236)
3822 - don't hold spinlock over lock dumping or change debug flags (4401)
3823 - don't zero obd_dev when it is force cleaned (3651)
3824 - print grants to console if they go negative (4431)
3825 - "lctl deactivate" will stop automatic recovery attempts (3406)
3826 - look for existing locks in ldlm_handle_enqueue() (3764)
3827 - don't resolve lock handle twice in recovery avoiding race (4401)
3828 - revalidate should check working dir is a directory (4134)
3830 - don't always mark "slow" obdfilter messages as errors (4418)
3832 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3835 - don't close LustreDB during write_conf until it is done (3860)
3836 - fix typo in lconf for_each_profile (3821)
3837 - allow dumping logs from multiple threads at one time (3820)
3838 - don't allow multiple threads in OSC recovery (3812)
3839 - fix debug_size parameters (3864)
3840 - fix mds_postrecov to initialize import for llog ctxt (3121)
3841 - replace config semaphore with spinlock (3306)
3842 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3843 - don't allow enqueue to complete on a destroyed export (3822)
3844 - down write_lock before checking llog header bitmap (3825)
3845 - recover from lock replay timeout (3764)
3846 - up llog sem before sending rpc (3652)
3847 - reduce ns lock hold times when setting kms (3267)
3848 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3849 - fix NULL deref and obd_dev leak on setup error (3312)
3850 - replace some LBUG about llog ops with error handling (3841)
3851 - don't match INVALID dentries from d_lookup and spin (3784)
3852 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3853 - fix invalid assertion in ptlrpc_set_wait (3880)
3855 - add libwrap support for the TCP acceptor (3996)
3856 - add /proc/sys/portals/routes for non-root route listing (3994)
3857 - allow setting MDS UUID in .xml (2580)
3858 - print the stack of a process that LBUGs (4228)
3860 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3863 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3864 - make sure to unset replay flag from failed open requests (3440)
3865 - if default stripe count is 0, use OST count for inode size (3636)
3866 - update parent mtime/ctime on client for create/unlink (2611)
3867 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3868 - free recovery state on server during a forced cleanup (3571)
3869 - unregister_reply for resent reqs (3063)
3870 - loop back devices mounting and status check on 2.6 (3563)
3871 - fix resource-creation race that can provoke i_size == 0 (3513)
3872 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3873 - more debugging for page-accounting assertion (3746)
3874 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3875 - avoid deadlock after precreation failure (3758)
3876 - fix race and lock order deadlock in orphan handling (3450, 3750)
3877 - add validity checks when grabbing inodes from l_ast_data (3599)
3879 - add /proc/.../recovery_status to obdfilter (3428)
3880 - lightweight CDEBUG infrastructure, debug daemon (3668)
3881 - change default OSC RPC parameters to be better on small clusters
3882 - turn off OST read cache for files smaller than 32MB
3883 - install man pages and include them in rpms (3100)
3884 - add new init script for (un)mounting lustre filesystems (2593)
3885 - run chkconfig in %post for init scripts (3701)
3886 - drop scimac NAL (unmaintained)
3888 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3891 - clean kiobufs before and after use (3485)
3892 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3893 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3894 - change init script to fail more clearly if not run as root (1528)
3895 - allow clients to reconnect during replay (1742)
3896 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3897 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3898 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3899 - fix cond_resched (3554)
3900 - only evict unfinished clients after recovery (3515)
3901 - allow bulk resend, prevent data loss (3570)
3902 - dynamic ptlrpc request buffer allocation (2102)
3903 - don't allow unlinking open directory if it isn't empty (2904)
3904 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3905 - remove extraneous obd dereference causing LASSERT failure (3334)
3906 - don't use get_cycles() when creating temp. files on the mds (3156)
3907 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3908 - handle EEXIST for set-stripe, set proper directory name (3336)
3910 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3911 - fix ksocknal_fmb_callback() error messages (2918)
3913 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3916 - don't copy lvb into (possibly NULL) reply on error (2983)
3917 - don't deref dentry after dput, don't free lvb on error (2922)
3918 - use the kms to determine writeback rpc length (2947)
3919 - increment oti_logcookies when osc is inactive (2948)
3920 - update client's i_blocks count via lvb messages (2543)
3921 - handle intent open/close of special files properly (1557)
3922 - mount MDS with errors=remount-ro, like obdfilter (2009)
3923 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3924 - don't use cancelling-locks' kms values (2947)
3925 - use highest lock extent for kms, not last one (2925)
3926 - don't dereference ERR_PTR() dentry in error handling path (3107)
3927 - fix thread race in portals_debug_dumplog() (3122)
3928 - create lprocfs device entries at setup instead of at attach (1519)
3929 - common AST error handler, don't evict client on completion race (3145)
3930 - zero nameidata in detach_mnt in 2.6 (3118)
3931 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3932 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3933 - initialize RPC timeout timer earlier for 2.6 (3219)
3934 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3935 - print nal/nid for unknown nid (3258)
3936 - additional checks for oscc recovery before doing precreate (3284)
3937 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3938 - don't crash in mdc_close for bad permissions on open (3285)
3939 - zero i_rdev for non-device files (3147)
3940 - clear page->private before handing to FS, better assertion (3119)
3941 - tune the read pipeline (3236)
3942 - fix incorrect decref of invalidated dentry (2350)
3943 - provide read-ahead stats and refine rpc in flight stats (3328)
3944 - don't hold journal transaction open across create RPC (3313)
3945 - update atime on MDS at close time (3265)
3946 - close LDAP connection when recovering to avoid server load (3315)
3947 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3948 - don't leak open file on MDS after open resend (3325)
3949 - serialize filter_precreate and filter_destroy_precreated (3329)
3950 - loop device shouldn't call sync_dev() for nul device (3092)
3951 - clear page cache after eviction (2766)
3952 - resynchronize MDS->OST in background (2824)
3953 - refuse to mount the same filesystem twice on same mountpoint (3394)
3954 - allow llmount to create routes for mounting behind routers (3320)
3955 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3956 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3957 - fix rare problem with rename on htree directories (3417)
3959 - allow default OST striping configuration per directory (1414)
3960 - fix compilation for qswnal for 2.6 kernels (3125)
3961 - increase maximum number of MDS request buffers for large systems
3962 - change liblustreapi to be useful for external progs like lfsck (3098)
3963 - increase local configuration timeout for slow disks (3353)
3964 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3966 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3969 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3970 - don't overwrite extent policy data in reply if lock was blocked (2901)
3971 - drop filter export grants atomically with removal from device (2663)
3972 - del obd_self_export from work_list in class_disconnect_exports (2908)
3973 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3974 - swab reply message in mdc_close, other PPC fixes (2464)
3975 - fix destroying of named logs (2325)
3976 - overwrite old logs when running lconf --write_conf (2264)
3977 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3978 - fix race in target_handle_connect (2898)
3979 - mds_reint_create() should take same inode create lock (2926)
3980 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3981 - don't close files for self_export to avoid uninitialized obd (2936)
3982 - allow MDS with the same name as client node (2939)
3983 - hold dentry reference for closed log files for unlink (2325)
3984 - reserve space for all logs during transactions (2059)
3985 - don't evict page beyond end of stripe extent (2925)
3986 - don't oops on a deleted current working directory (2399)
3987 - handle hard links to targets without a parent properly (2517)
3988 - don't dereference NULL lock when racing during eviction (2867)
3989 - don't grow lock extents when lots of conflicting locks (2919)
3991 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3994 - account for cache space usage on clients to avoid data loss (974)
3995 - lfsck support in lustre kernel code (2349)
3996 - reduce journal credits needed for BRW writes (2370)
3997 - orphan handling to avoid losing space on client/server crashes
3998 - ptlrpcd can be blocked, stopping ALL progress (2477)
3999 - use lock value blocks to assist in proper KMS, faster stat (1021)
4000 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4001 - recovery for initial connections (2355)
4002 - fixes for mds_cleanup_orphans (1934)
4003 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4004 - block all file creations until orphan recovery completes (1901)
4005 - client remove rq_connection from request struct (2423)
4006 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4007 - recovery timer race (2670)
4008 - mdc_close recovey bug (2532)
4009 - ptlrpc cleanup bug (2710)
4010 - mds timeout on local locks (2588)
4011 - namespace lock held during RPCs (2431)
4012 - handle interrupted sync write properly (2503)
4013 - don't try to handle a message that hasn't been replied to (2699)
4014 - client assert failure during cleanup after abort recovery (2701)
4015 - leak mdc device after failed mount (2712)
4016 - ptlrpc_check_set allows timedout requests to complete (2714)
4017 - wait for inflight reqs when ptlrpcd finishes (2710)
4018 - make sure unregistered services are removed from the srv_list
4019 - reset bulk XID's when resending them (caught by 1138 test)
4020 - unregister_bulk after timeout
4021 - fix lconf error (2694)
4022 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4023 - readahead locks pages, leaves pending causing memory pressure (2673)
4024 - increase OST request buffers to 4096 on large machines (2729)
4025 - fix up permission of existing directories in simple_mkdir (2661)
4026 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4027 - don't assign transno to errored transactions (2742)
4028 - don't delete objects on OST if given a bogus objid from MDS (2751)
4029 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4030 - if rq_no_resend, then timeout request after recovery (2432)
4031 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4032 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4033 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4034 - MDS recovery shouldn't delete objects that it has given out (2730)
4035 - if enqueue arrives after completion, don't clobber LVB (2819)
4036 - don't unlock pages twice when trigger_group_io returns error (2814)
4037 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4038 - don't write pages to disk if there was an error (1450)
4039 - don't ping imports that have recovery disabled (2676)
4040 - take buffered bytes into account when balancing socknal conn (2817)
4041 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4042 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4043 - remove little-endian swabbing of llog records (1987)
4044 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4045 - retry reposting request buffers if they fail (1191)
4046 - grow extent at grant time to avoid granting a revoked lock (2809)
4047 - lock revoke doesn't evict page if covered by a second lock (2765)
4048 - disable VM readahead to avoid reading outside lock extents (2805)
4050 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4051 - updated kernel patches for hp-2.4.20 kernel (2681)
4053 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4056 - fix truncated write corruption (2366)
4057 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4059 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4060 - ensure len(uuid) < 37 in lmc (1171)
4061 - fix ia64 OOPS in llog_test (2255)
4062 - zero end of page at obdfilter for partial page writes (2648)
4063 - don't leave stale dentries around after renames (bug 2428)
4064 - fix timeouts when evicting a client with a single lock held (2642)
4065 - set deadline for the initial HELLO message to drain (2634)
4066 - print out dotted-quad IP addresses in the socknal (2302)
4068 - additional debugging for MDS client eviction problem (2443)
4069 - fix mkfsoptions support for osts (2603, 2604)
4071 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4074 - add series for the vanilla 2.6.0 kernel
4075 - add series for the vanilla 2.4.24 kernel
4076 - add series for a cray x86/64 UL kernel drop
4077 - fix xattr patches for the vanilla 2.4.19 series
4079 - generate true UUIDs in lmc (1171)
4080 - have portals stack dumping break in UML (2466)
4081 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4082 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4083 - wake callback waiting threads on client eviction (2460)
4084 - Add --ptldebug and --subsystem to lmc (1719)
4085 - update assertion to allow safe interrupt allocation
4086 - set rq_no_resend for cancel requests (2432)
4087 - recalculate ptlrpcd timeout after resend (2494)
4088 - call vfs_rmdir when removing pending directories (2368)
4089 - fix renaming a file to itself (2429)
4090 - lmc creates a default one-stripe lov (2454)
4091 - expand procfs space to handle large clusters (2326)
4092 - increase UML stack to avoid overflow
4093 - update lconf's list of debug and subsystem masks
4094 - fix lfs find --obd (2510)
4095 - /proc tunable for disabling filter read caching (2591)
4096 - stop rpm packages from altering slapd.conf (2301)
4097 - disable nagle in the socknal under 0conf (2578)
4098 - choose mds inode size based on stripe count (2572)
4099 - fix kernel-source rpm problems (2516)
4101 - add --disable-doc to avoid pdf generation (2421)
4102 - update documentation, tests, type-os, comments
4103 - avoid format warnings on ia64
4104 - remove the TOE NAL
4105 - tiny code cleanups by removing unused fields
4107 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4110 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4111 - remove the most blatant lies from BUILDING file (2371)
4112 - change default debug level to reasonable production setting
4113 - reduce client side cache size to reduce cache flush time
4114 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4115 - make TCP zerocopy and pinger support enabled by default (2476)
4116 - sync writes completed after process exits caused crashes (2319)
4117 - maintain correct mount count on the MDS (2356)
4118 - backout 1557, because 2316 wasn't really fixed
4119 - better file I/O statistics gathering in /proc
4120 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4121 - another kernel patch to fix zero-copy TCP function export
4122 - don't take duplicate lock when processing re-sent getattr (2420)
4123 - lctl uses obd_self_export instead of creating new conn (2353)
4124 - MDS/OST recovery case which requires object creation asserted (2425)
4125 - move lfs from /usr/sbin to /usr/bin in packages
4126 - fix race between mds_client_add and mds_client_free (2417)
4127 - use kmalloc instead of slabs in portals (2430)
4128 - don't create duplicate records when a failover MDS is present (2442)
4129 - remove unnecessary mount age check (2332)
4130 - don't remove directory inodes from locks prematurely (2451)
4131 - don't break if MDS service name is the same as hostname (2103)
4132 - fix races in client write RPC generation when cache full (2482)
4134 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4137 - remove now-unused request->rq_obd (278)
4138 - if an allocation fails, print out how much memory we've used (1933)
4139 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4140 - add the "configurable stack size" patch to most series files (1256)
4141 - ability to write large log records, for 100+ OST configs (2306)
4142 - fix NULL deref when filter_prep fails (2314)
4143 - fix operator precedence error in filter_sync
4144 - dynamic allocation of socknal TX descriptors (2315)
4145 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4146 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4147 - ensure that CWARN messages in Portals always get to the syslog
4148 - __init/__exit are not for prototype decls (ldlm_init/exit)
4149 - x86-64 compile warning fixes
4150 - fix gateway LMC keyword conflict (2318)
4151 - fix MDS lock inversions in getattr/reint paths (1844)
4152 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4153 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4154 - DLM race condition prevented some lock evictions (2328)
4155 - ENOMEM detection and retry on socknal sends (2230)
4156 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4157 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4158 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4159 - better error messages when a client is rejected during recovery (1505)
4160 - avoid cancelling locks which were never granted, after failure (2330)
4161 - fix i_sem/journal inversion in mds_client_add (2333)
4162 - fix truncate/getattr lock cycle deadlock (2334)
4163 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4164 - fix two rare exit paths which could leak an l_lock() ref (2321)
4165 - fencepost error in MDS/OST orphan recovery (2226)
4166 - make log record alignment 8 bytes (1988)
4167 - lstripe now fails when requested offset > ost_count (2237)
4168 - ensure that all kernel series have a complete list.h (1607)
4169 - fix crashes in special-file operations (2316)
4170 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4172 - allow configurable automake binary, for testing new versions
4173 - small update to the lfs documentation
4175 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4177 * fix negative export reference count in fsfilt_sync (2312)
4179 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4180 * release candidate 0.9.1
4182 - orphans are moved into the PENDING directory for possible recovery
4183 - replayed opens now open by fid for orphan/rename safety (1042)
4184 - last close of an orphan inode generates a transno (683)
4185 - chdir() and mount() now pin the directory entry (1020)
4186 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4187 - discard very old requests without processing them (1502)
4188 - remove some common, well-understood CERRORs (1505)
4189 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4190 - clear "grant" flags in OST replies until OST grant code lands (1644)
4191 - fix read performance by not clobbering i_blksize on client (1598)
4192 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4193 - make LRU size a /proc tunable, clears locks when reduced (707)
4194 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4195 - server locks take references on exports now (1558)
4196 - build fixes for 2.4.20-rh trees (1663)
4197 - return an error from lov_create if all OSCs are inactive (1751)
4198 - fix import levels when a reconnect happens without a timeout (1597)
4199 - exit early from mds_open if we get a lookup error (1749)
4200 - partial page read at EOF wouldn't wait for disk before sending (1642)
4201 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4202 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4203 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4204 - write/truncate lock inversion (1639)
4205 - Don't auto-load obdclass, portals modules during cleanup (1495)
4206 - fix timestamps from jumping to "now" (1763)
4207 - extra journal assertions (1648)
4208 - add an extra multiunlink test (1771)
4209 - fix read_record/write_record API (1776)
4210 - fix leak of offset_extent, possible incorrect i_size later (1772)
4211 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4212 - Debugging for the kqswnal_get_idle_tx problems (1820)
4213 - Allow recovery to be attempted multiple times (1536)
4214 - Write out MDS last_rcvd file after it is first created (1600)
4215 - Fix tx_descriptor leak in failed transmit situations (1827)
4216 - ext3 journaling fixes for assertion failure after IO error (1871)
4217 - class_export_put() on freed export after completion AST error (1896)
4218 - Fix revalidate looping in VFS (1322)
4219 - Don't access a freed export during MDS_REINT timeout (1521)
4220 - Add open-unlink recovery support on the MDS (1673,1764)
4221 - Return an error if no MDS data was read from last_rcvd (1946)
4222 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4223 - Don't setup a disk device that doesn't match exported UUID (317)
4224 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4225 - avoid committing NULL handle in force close
4226 - local.sh is now a one-stripe LOV configuration
4227 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4228 - fix ext3 htree duplicate directory entry corruption (1516)
4229 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4230 - update to new LOV EA format (2097)
4231 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4233 2003-06-15 Phil Schwan <phil@clusterfs.com>
4236 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4237 - per-import/export recovery handling (958, 931, 959)
4238 - multiple last-rcvd slots, for serving multiple FSes (949)
4239 - connections are again shared between multiple imp/exports (963, 964)
4240 - "umount -f" would hang if any requests needed to be sent (393, 978)
4241 - avoid pinning large req buffer by copying for queued messages (989)
4242 - add "uuid" to "lctl device" command to help upcalls (991)
4243 - "open" RPCs with transnos would confuse recovery counters (1037)
4244 - do proper endian conversion of all wire messages (288, 340, 891)
4245 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4246 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4247 - fix ptlrpc_connection leak in target_handle_connect (1174)
4248 - fix import refcounting bug in OST and MDS cleanup (1134)
4249 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4250 - fix very unlikely obd_types race condition (501)
4251 - remove osc_open hack for echo_client (1187)
4252 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4253 - a failure in read_inode2 leads to deadlock (1139)
4254 - cancel ack-locks as soon as transaction is committed (1072)
4255 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4256 - make sure to commitrw after any preprw to avoid deadlock (1162)
4257 - failing to execute a file in a lustre FS would lock inode (1203)
4258 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4259 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4260 - obd_uuid2tgt would walk off the end of the list (1255)
4261 - on IA64 the osc would give portals incorrect bulk size (1258)
4262 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4263 - fix lock inversion caused by new llite matching code (1282)
4264 - limit the number of dirty pages on a client to 10MB (1286)
4265 - timed out locks were not being corrected cancelled (1289)
4266 - fix O_DIRECT above 4GB on IA-32 (1292)
4267 * major user-visible changes
4268 - fail out/fail over policy now controlled by the upcall (993)
4270 - add OBD_PING to check server availability and failure (954)
4271 - lustre messages are now sent in sending host order (288, 340, 891)
4272 - add eadatalen to MDS getattr reply (340)
4273 - OST read replies may contain second buffer, with per-page status (593)
4275 2003-03-11 Phil Schwan <phil@clusterfs.com>
4278 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4279 - failed open()s could cause deadlock; fixed (867, 869)
4280 - stop cancelling OST locks when files are closed (481)
4281 - overlapping XID spaces caused network corruption (851, 853)
4282 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4283 - setattr_raw added, to do single-RPC, server-side setattrs
4284 - lmc/lconf syntax change for OST UUIDs
4285 - fix crashy race condition between ptlrpc_free_req and osc_close
4286 - don't use request in mdc_enqueue if we hit a timeout (889)
4287 - don't set the inode i_size for regular files from the MDS (896)
4288 - handle out of order completion AST (842)
4289 - don't LBUG if a lock request times out after receiving AST (913)
4290 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4291 - if a bad lock AST arrives, send an error instead of dropping entirely
4292 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4293 - fix leak in bulk IO when only partially completed (899, 900, 926)
4294 - fix O_DIRECT for ia64 (55)
4295 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4296 - C-z after timeout could hang a process forever; fixed (977)
4298 - client-side I/O cache (678, 924, 929, 941, 970)
4300 - READPAGE and SETATTRs which don't take server-side locks get
4303 2003-02-11 Phil Schwan <phil@clusterfs.com>
4306 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4307 - Fix fsfilt_extN_readpage() to read a full page of directory
4308 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4309 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4310 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4311 - Most intent operations were removed, in favour of a new RPC mode
4312 that does a single RPC to the server and bypasses most of the VFS
4313 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4314 - Aggressively cancel local locks on DLM servers
4315 - mds_reint_unlink sends EA to the client if it's the last nlink.
4316 client uses that EA to unlink OST objects.
4317 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4318 - recursive symlinks were fixed (439)
4319 - fixed NULL deref in DEBUG_REQ
4320 - filter_update_lastobjid no longer calls sync, which annoyed extN
4321 - fixed multi-client small-writes to a single file problem (445)
4322 - fixed mtime updates during file writes (607)
4323 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4324 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4325 - fixed Linux OST size reporting problem (444, 656)
4326 - OST now updates object mtime with writes or setattr (607, 619)
4327 - client verifies file size before zeroing page past EOF (445)
4328 - OST now writes last allocated objid to disk with allocation (108)
4329 - LOV on echo now works (409)
4331 - mds_reint_unlink sends a new buffer, with the EA included. this
4332 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4333 set if a regular file was being unlinked, and it was the last link
4334 - use PtlGet from the target for bulk writes (315)
4335 - OST now updates object mtime with writes or setattr (607, 619)
4336 - LDLM now has a grant-time callback to revalidate locked items, if
4338 - Many MDS operations were reorganized to combat race conditions
4340 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4341 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4343 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4346 - Fully reactivate OST imports after reconnection (512, others)
4347 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4348 - More graceful error handling for truncating on dead OST (515)
4349 - Don't error out unless we're actually accessing dead stripes (474)
4350 - Fix garbage sizes when stripes are missing (410)
4351 - LRU counters were broken, causing constant lock purge (433, 432)
4352 - garbage on read from stripes with failed OSTs (441)
4353 - mark OSCs as active before reconnecting during recovery (438)
4354 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4355 - lfind did not preserve OST order in output (443)
4356 - symlinks cause hung clients, incorrect data (439)
4357 - stop dereferencing request after dropping refcount (457)
4358 - don't LASSERT(spin_is_locked) on non-SMP (455)
4359 - fixes for many rename() bugs
4360 - fstat didn't correctly synchronize attributes (399)
4361 - server must handle lock cancellation during blocking AST prep (487)
4362 - bulk descriptors were free()d too soon (511)
4363 - fix paths in lconf, which would load incorrect modules (451, 507)
4364 - fix confusing lconf 'host not found' error message (386)
4365 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4366 - fix race condition in mdc_blocking_ast() for inode access (526)
4367 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4368 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4369 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4370 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4371 - fix lstripe in several ways (kernel side) (527)
4372 - fix request leak in ldlm_cli_enqueue (262)
4373 - incorrect OSC was marked inactive after OST failure
4374 - call mds_fs_cleanup before unmounting filesystem (524)
4375 - fix races between taking ns_lock and ldlm_lock_change_resource
4376 - fix races updating LOV export open file list
4377 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4378 - fix recovery NULL deref in ldlm_cli_cancel_unused
4379 - fix some DLM races by using new hash table for lock handles (419)
4380 - permit the client to specify desired inodes, at replay
4381 - duplicate requests when we queue them for replay reintegration
4382 - fix last_rcvd offset calculation
4383 - sync after each recovered transaction, so we always make progress
4384 - never, not always, ERESTART requests without transnos
4385 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4387 - skip replay if the MDS says that the client is already connected
4388 - don't check for a recovery-enabled export to match lctl's UUID
4389 - don't INC_USE_COUNT for phantom exports
4390 - don't crash when cleaning up phantom exports (567)
4391 - don't double-finish or set replay data for errored mdc_open requests
4392 - abort requests when they time out, so we don't get old replies
4393 - send/receive replies for AST messages again
4394 - if the client says that it doesn't have the lock, cancel it on the
4396 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4397 mark it as destroyed, it will all work out when decref is called
4398 - fix module use counts (22, 581)
4400 - ASTs now expect a reply (server cancels lock on error reply)
4402 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4405 - fix many simultaneous client startup (392)
4406 - fix dentry->d_it clobbering
4407 - credentials weren't being shipped for readdir/getattr operations
4408 - remove invalid assertions triggered during some concurrent MD
4410 - proper Lustre versions added (336, 389)
4411 - fix memory leak for create error case (398)
4412 - fix LOV locking bug that would get cli/srv out of sync
4413 - fix echo client over LOV (409)
4414 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4415 - fix double-O_EXCL intent crash (424)
4416 - avoid sending multiple lock CANCELs (352)
4418 - MDS can do multi-client recovery (modulo bugs in new code)
4420 - many updates, edits, cleanups
4422 2002-11-18 Phil Schwan <phil@clusterfs.com>
4425 - fix null d_it dereference (346)
4426 - fix full OST/dbench hang (333)
4427 - fix permission problem with file removal (286)
4428 - fix removal of OSCs from LOV when they fail
4429 - fix NULL deref during bulk timeout (214)
4430 - fix problems related to multiple filesystems on one MDS (241)
4431 - fixed serious subtle metadata locking bugs
4432 - free locks on clients when inodes are removed due to memory
4434 - fix inode pointer in lock data (285)
4435 - partial support for multiple MDS on a single host (241)
4436 - data locks weren't cancelled at clear_inode time (290, 311)
4437 - intent locks could lead to unbounded lock growth (205)
4438 - added a maximum lock count, an LRU list, and a flusher
4439 - fix multiple rename (365)
4440 - properly abstracted the echo client
4441 - OSC locked 1 byte too many; fixed
4442 - rewrote brw callback code:
4443 - fixed recovery bugs related to LOVs (306)
4444 - fixed too-many-pages-in-one-write crash (191)
4445 - fixed (again) crash in sync_io_timeout (214)
4446 - probably fixed callback-related race (385)
4448 - Add capability to MDS protocol
4449 - LDLM cancellations and callbacks on different portals
4451 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4454 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4455 - defer open object destruction to close time (601981, 138)
4456 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4457 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4458 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4459 - don't call obd_setattr in ll_file_release for destroyed objects
4460 * protocol change to lustre_msg: move |version| and add |flags|
4461 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4462 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4463 * added replay of create, unlink, link and rename operations during
4464 MDS failover; recovery should be much more robust now
4465 * remove failed OSCs from LOVs (only lov_create uses this so far)
4466 * the lustre-HOWTO was brought (more) up to date (582544)
4468 2002-10-23 Phil Schwan <phil@clusterfs.com>
4471 - in-use dentries weren't being reused properly (617851)
4472 - prevent multiple LDLM setup (599178)
4473 - fix LOV size calculations for truncate (617853)
4474 - fix client handling of MDS intent errors (POSIX)
4475 - fix permission bug in lovstripe.c test (624321)
4476 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4477 - truncate past end of file could corrupt data
4478 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4479 - a race in recovery could return ETIMEDOUT to apps (623947)
4480 - building outside the source directory was fixed
4481 * the lustre-HOWTO was brought (more) up to date (582544)
4482 * major progress was made on recovery functionality
4484 2002-10-10 Phil Schwan <phil@clusterfs.com>
4487 - recovery deadlock fix
4488 - rm -rf causes LBUG fix (617817)
4489 - file open by multiple tasks fix (618962)
4490 - directory permissions bugs (602707 and 620007)
4491 - journal_stop fixed with locking (611313)
4492 - O_APPEND failures resolved (618273, perhaps 614459)
4493 - lconf PATH fix (619770)
4494 - IA64 build fix (621450)
4495 - RPC buffer sizes scale with amount of memory
4497 2002-10-01 Phil Schwan <phil@clusterfs.com>
4500 - locks would be cancelled without throwing away data pages,
4501 resulting in inconsistent data (605627)
4502 - inode attributes were not always being refreshed (605627, 612449)
4503 - lconf now continues to cleanup after lctl reports an error
4504 - MDS now enforces user permissions (602707)
4505 - lprocfs cleanup fixed, but not yet enabled (614157)
4506 - fixed infinite server hang, should a client not respond to an AST
4507 - avoid going into recovery if user calls readlink() with a buffer
4508 that's too small (613941)
4509 - AST RPCs no longer require replies (614867) -- this may be changed
4510 - don't crash server if client sends an IOV that's too big (611336)
4511 - fixed lock conversion deadlock (611892)
4512 - fixed the following of symlinks (614622)
4513 * recovery: the server can remove locks from a client that dies, other
4514 clients can make progress
4515 * more extN patch fixes
4516 * compile-time configurable ptlrpc buffer allocations
4518 - collaborative read cache document
4519 - Lustre Lite Performance CDR document-in-progress
4521 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4524 - fix typo in patch-2.4.18
4526 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4529 - clear ptlrpc request each time in handle_incoming_request()
4530 - unlink of files now destroys the object on the OST
4532 2002-09-19 Peter Braam <braam@clusterfs.com>
4534 * add hard link support
4535 * change obdfile creation method
4536 * kernel patch changed
4538 2002-09-19 Peter Braam <braam@clusterfs.com>
4541 - stack overflow bug in extN fixed
4543 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4545 * documentation updates
4546 - add man pages for config tools
4547 - update tests/README to describe testing with new config tools
4548 - finish metadata API descriptions
4549 * bug fixes and cleanups
4550 - statfs workaround for 16TB limit
4551 - LOV stripe allocation improved, can stripe on subset of OSTs
4552 - LOV file size/IO offset was wrong for files > 4GB in size
4553 - object EA data was being dropped, caused files to be unreadable
4554 - memory overflow with non-LOV OST caused memory corruption
4555 - fixed regression tests to work with new config tools, obdfilter
4556 - fixed bug when directory size became larger than 1 block
4557 - fixed bug (for single client case) when PWD was deleted
4558 - invalidate local directory pages when doing intent-based ops
4559 - avoid LDLM oops when lock callback contained bad data
4561 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4563 * documentation updates
4564 * bug fixes and cleanups
4565 - configuration tools
4568 - 64-bit compile warnings
4569 - 64-bit internal statfs data
4571 * test_brw on persistent OST devices
4573 * lprocfs (disabled)
4575 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4577 * documentation updates
4578 * bug fixes and cleanups
4579 * configuration tools
4581 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4584 * many small fixes to 0_5_4
4585 * io/network handling
4586 * thinkos in MDS operations
4588 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4591 * crucial basic fixes to 0.5.3
4592 * IOR, Iozone work over Elan
4595 2002-08-07 Phil Schwan <phil@clusterfs.com>
4596 * version 0_5_3, our first alpha
4597 * we use the new Portals iovs
4598 * documentation updates
4599 * bug fixes and cleanups
4600 * small changes in the DLM wire protocol
4602 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4603 * version 0_5_1 with some initial stability,
4604 * locking on MD and file I/O.
4605 * documentation updates
4606 * several bug fixes since 0.5.0
4607 * small changes in wire protocol
4609 2002-07-18 Phil Schwan <phil@clusterfs.com>
4611 * delivered as Lustre Light Alpha
4612 * fixed a crash after handling invalid MDS requests
4613 * fixed directory pages for architectures with non-4k pages sizes
4615 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4616 * release version v0_4_4
4617 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4619 * Fixes a number of interruption problems with OST operations.
4620 * Update documentation for portals header changes
4621 * Move all wire protocol structs/defines to lustre_idl.h
4622 * Fixes symlink length bug.
4623 * Add tcpdump to repository.
4625 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4626 * release version v0_4_3
4627 * Fixes statfs for inodes on extN.
4628 * Fixes bug in runtests which would delete /etc/hosts.
4629 * Use 64-bit object IDs wherever possible (not into VFS though)
4630 Remove ost_get_info, which is unused by lustre, and out of date.
4632 2002-07-03 Peter Braam <braam@clusterfs.com>
4633 * release version v0_4_2 Fixes a lookup error (type not passed)
4634 * move forward to head of Portals
4635 * move forward to latest Lustre kernel
4637 2002-06-25 Peter Braam <braam@clusterfs.com>
4638 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob