1 tbd Sun Microsystems, Inc.
5 2.6.16.54-0.2.3 (SLES 10),
6 2.6.18-53.1.4.el5 (RHEL 5).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.40.4-cfs1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
17 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
18 Details : If an OST is down the MDS will hang indefinitely in
19 obd_statfs_rqset() waiting for the statfs data. While for
20 MDS QOS usage of statfs, it should not stuck in waiting.
22 Severity : enhancement
24 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
26 Severity : enhancement
28 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
32 Description: make mgs_setparam() handle fsname containing dash
33 Details : fsname containing a dash does not work with lctl conf_param
35 Severity : enhancement
37 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
39 Severity : enhancement
41 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
45 Frequency : rare, at shutdown
46 Description: access already free / zero obd_namespace.
47 Details : if client_disconnect_export was called without force flag set,
48 and exist connect request in flight, this can produce access to
49 NULL pointer (or already free pointer) when connect_interpret
50 store ocd flags in obd_namespace.
54 Frequency : only at startup
55 Description: not alloc memory with spinlock held.
56 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
59 Severity : enhancement
61 Description: make lustre randomly fail allocating memory
62 Details : Make lustre randomly failed allocating memory for testing purpose.
64 Severity : enhancement
66 Description: lost problems with lov objid file
67 Details : Fixes some scability and access to not inited memory problems
68 in work with lov objdid file.
73 Description: lfs find does not continue on file error
74 Details : Continue other files processing when a file/dir is absent.
78 Description: Inconsistent usage of lustre_pack_reply()
79 Details : Standardize the usage of lustre_pack_reply() such that it
80 always generate a CERROR on failure.
83 Frequency : occasional
85 Description: MDS hang or stay in waiting lock
86 Details : If client receive lock with CBPENDING flag ldlm need send lock
87 cancel as separate rpc, to avoid situation when cancel request
88 can't processed due all i/o threads stay in wait lock.
92 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
96 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
100 Description: Update to SLES9 kernel-2.6.5-7.287.3.
104 Description: Update to RHEL4 latest kernel.
106 Severity : enhancement
108 Description: Build SLES10 patchless client fails
109 Details : The configure was broken by run ./configure with
110 --with-linux-obj=.... argument for patchless client. When the
111 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
112 can't search header adequately. Use absolute path such as
113 -I($LINUX)/include instead.
117 Description: interrupt oig_wait produce painc on resend.
118 Details : brw_redo_request can be used for resend requests from ptlrpcd and
119 private set, and this produce situation when rq_ptlrpcd_data not
120 copyed to new allocated request and triggered LBUG on assert
121 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
122 ptlrpcd set if request is changed and can be safety changed to use
127 Description: LASSERT_{REQ,REP}SWAB macros are buggy
128 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
129 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
130 these macros and replace them with their difinitions instead.
134 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
135 Details : Update which_patch & target file for SLES10 latest kernel.
138 Description: Remove client patches from SLES 10 kernel.
139 Details : This causes SLES 10 clients to behave as patchless clients
140 even on a Lustre-patched (server) kernel.
142 Severity : enhancement
144 Description: self-adjustable client's lru lists
145 Details : use adaptive algorithm for managing client cached locks lru
146 lists according to current server load, other client's work
147 pattern, memory activities, etc. Both, server and client
148 side namespaces provide number of proc tunables for controlling
151 Severity : enhancement
153 Description: light-weight GSS support
154 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
155 and reduce performance overhead.
157 Severity : enhancement
159 Description: Linux keyring support
160 Details : Support using service of Linux keyring for Lustre GSS internal
161 context refresh/cache mechanism.
165 Description: Fix errors in lfs documentation
166 Details : Fixes man pages
170 Description: don't use GFP_* in generic Lustre code.
171 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
175 Description: obdclass is limited by single OBD_ALLOC(idarray)
176 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
180 Description: warnings with build patchless client with vanila 2.6.19 and up
181 Details : change old ctl_table style and replace ctl_table/ctl_table_header
182 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
186 Frequency : Only for SLES
187 Description: sanity_quota fail test_1
188 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
189 file, which makes TSTID[2] inunique.
193 Frequency : Only for SLES9
194 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
195 Details : Update target/ChangeLog/which_patch .
199 Frequency : Only for test_10 in sanity-quota.sh
200 Description: a bug in quota test code
201 Details : checking a test flag in a wrong place
205 Frequency : Only for test_13 in sanity-quota.sh
206 Description: a bug in quota test script
207 Details : shouldn't check overfull allocated quota
211 Description: jbd statistics
212 Details : Port older jbd statistics patch for sles10
214 Severity : enhancement
216 Description: Add printing inode info into message about error in writepage.
221 Description: reply_lock_interpret crash due to race with it and lock cancel.
222 Details : Do not replay locks that are being cancelled. Do not reference
223 locks by their address during replay, just by their handle.
227 Frequency : When flocks are used.
228 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
229 Details : Flock locks might destroy just granted lock if it could be merged
230 with another existing flock, this is done in completion handler,
231 so teach ldlm_cli_enquque_fini that this is a valid case for
237 Description: Oops in read and write path when failing to allocate lock.
238 Details : Check if lock allocation failed and return error back.
242 Description: lstripe command fails for valid OST index
243 Details : The stripe offset is compared to lov->desc.ld_tgt_count
244 instead of lov->desc.ld_active_tgt_count.
248 Description: sanity.sh failed test 103
249 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
254 Description: df doesn't work properly if diskfs blocksize != 4K
255 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
259 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
260 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
264 Description: obd_health_check_timeout too short
265 Details : set obd_health_check_timeout as 1.5x of obd_timeout
269 Description: llapi_file_create() does not allow some changes
270 Details : add llapi_file_open() that allows specifying the mode and
271 open flags, and also returns an open file handle.
275 Description: merge and cleanup kernel patches.
276 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
280 Description: Add journal checksum support.(Kernel part)
281 Details : The journal checksum feature adds two new flags i.e
282 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
283 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
284 indicates that the commit block contains the checksum for
285 the blocks described by the descriptor blocks. Now commit
286 record can be sent to disk without waiting for descriptor
287 blocks to be written to disk. This behavior is controlled
288 using JBD2_FEATURE_ASYNC_COMMIT flag.
292 Description: OSS needs mutliple precreate threads
293 Details : Add ability to start more than one create thread per OSS.
297 Description: Sanity.sh test_65k failed: setstripe should have succeeded
298 Details : execute lfs setstripe on client
302 Description: mds_obd_create error creating tmp object
303 Details : When the user sets quota on root, llog will be affected and can't
304 create files and write files.
307 Frequency : Always on ia64 patchless client, and possibly others.
309 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
310 Details : This allows the patchless client to be loaded on architectures
315 Description: RedHat Update kernel for RHEL5
316 Details : Add the kernel config file for RHEL5.
320 Description: RedHat Update kernel for RHEL5
321 Details : Modify the kernel config file more closer RHEL5.
325 Description: Build failure against Centos5 (RHEL5)
326 Details : Define PAGE_SIZE when it isn't present.
330 Description: test_7 on sanity-quota.sh can't be used on separate machines
331 Details : add facet to handle it
335 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
336 Details : with NFS, the anon dentry's parent was set to itself in
337 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
338 corresponding dentry other than use rec->ur_name.
340 Severity : enhancement
342 Description: lfs setstripe enhancement
343 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
347 Description: enable data checksumming by default
348 Details : enable checksum by default, allow --disable-checksum
349 configure option and "-o nochecksum" mount option. Checksums
350 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
354 Description: lustre support for RHEL5
355 Details : Add support for RHEL5.
359 Description: Client eviction due to failover config
360 Details : after a connection loss, the lustre client should attempt to
361 reconnect to the last active server first before trying the
362 other potential connections.
366 Description: when mds and osts use different quota unit(32bit and 64bit),
367 quota will be released repeatly.
368 Details : void sending multiple quota reqs to mds, which will keep the
369 status between the reqs.
373 Description: osts not allocated evenly to files
374 Details : change the condition to increase offset_idx
377 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
379 Description: Data corruption for OSTs that are formatted larger than 2TB
381 Details : When generating the bio request for lustre file writes the
382 sector number would overflow a temporary variable before being
383 used for the IO. The data reads correctly from Lustre (which
384 will overflow in a similar manner) but other file data or
385 filesystem metadata may be corrupted in some cases.
389 Description: Tune the kernel for good SCSI performance.
390 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
391 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
396 Description: rewrite ext2-derived code in obdclass/uuid.c
397 Details : rewrite inherited code (uuid parsing code from ext2 utils)
398 from scratch preserving functionality.
402 Description: "lfs find -obd UUID" prints directories
403 Details : "lfs find -obd UUID" will return all directory names instead
404 of just file names. It is incorrect because the directories
405 do not reside on the OSTs.
409 Description: use i_size_read and i_size_write in 2.6 port
410 Details : replace inode->i_size access with i_size_read/write()
413 Frequency : when removing large files
415 Description: scheduling issue during removal of large Lustre files
416 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
417 It causes scheduling issues when removing large files (17TB in the
421 Frequency : only with liblustre clients on XT3
423 Description: evictions taking too long
424 Details : allow llrd to evict clients directly on OSTs
427 Frequency : only on ppc
429 Description: /proc/fs/lustre/devices broken on ppc
430 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
431 We should make sure the type of 'index' is loff_t and then cast
432 explicitly as needed below. Do not assign an explicitly cast
436 Frequency : only for rhel5
438 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
439 Details : Modify the target file & which_kernel.
441 Severity : enhancement
443 Description: omit set fsid for export NFS
444 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
445 as FS_REQUIRES_DEV for avoid problems with generate fsid.
449 Frequency : Always, for kernels after 2.6.16
450 Description: Fix warning idr_remove called for id=.. which is not allocated.
451 Details : Last kernels save old s_dev before kill super and not allow
452 to restore from callback - restore it before call kill_anon_super.
456 Description: buffer overruns could theoretically occur
457 Details : llapi_semantic_traverse() modifies the "path" argument by
458 appending values to the end of the origin string, and a
459 overrun may occur. Adding buffer overrun check in liblustreapi.
463 Description: Fix error on 'ls .' at the top of the Lustre mount.
464 Details : Don't revalidate dentry if it is a root dentry.
468 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
469 Details : Modify vm-tunables-rhel4.patch.
473 Description: Kernel config for 2.6.18-vanilla.
474 Details : Modify targets/2.6-vanilla.target.in.
475 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
476 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
477 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
478 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
482 Description: improve handling recoverable errors
483 Details : if request processig with error which can be recoverable on server
484 request should be resend, otherwise page released from cache and
489 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
490 Details : Modify target file & which_patch.
491 A flaw was found in the IA32 system call emulation provided
492 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
493 value could be stored in the %RAX register, which could trigger an
494 out-of-bounds system call table access. An untrusted local user
495 could exploit this flaw to run code in the kernel
496 (ie a root privilege escalation). (CVE-2007-4573).
500 Description: O_DIRECT bypasses client statistics.
501 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
502 read_ahead_stats were updated. copied the stats section from
503 osc_send_oap_rpc() into async_internal().
507 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
511 Description: change order of libsysio includes
512 Details : '#include sysio.h' should always come before '#include xtio.h'
516 Description: handle "serious error: objid * already exists" more gracefully
517 Details : If LAST_ID value on disk is smaller than the objects existing in
518 the O/0/d* directories, it indicates disk corruption and causes an
519 LBUG(). If the object is 0-length, then we should use the existing
520 object. This will help to avoid a full fsck in most cases.
522 Severity : enhancement
524 Description: adapt the lustre_config script to support the upgrade case
525 Details : Add "-u" option for lustre_config script to support upgrading 1.4
526 server targets to 1.6 in parallel.
530 Description: To avoid grant space > avaible space when the disk is almost
531 full. Without this patch you might see the error "grant XXXX >
532 available" or some LBUG about grant, when the disk is almost
534 Details : In filter_check_grant, for non_grant cache write, we should
535 check the left space by if (*left > ungranted + bytes), instead
536 of (*left > ungranted), because only we are sure the left space
537 is enough for another "bytes", then the ungrant space should be
538 increase. In client, we should update cl_avail_grant only there
539 is OBD_MD_FLGRANT in the reply.
543 Description: Update RHEL 4 kernel to fix local root privilege escalation.
544 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
545 described in CVE-2007-4573. This problem could allow untrusted
546 local users to gain root access.
549 Frequency : when using O_DIRECT and quotas
551 Description: Incorrect file ownership on O_DIRECT output files
552 Details : block usage reported by 'lfs quota' does not take into account
553 files that have been written with O_DIRECT.
558 Description: touch file failed when fs is not full
559 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
560 otherwise we can get ENOSP while fs is not full.
564 Description: parallel lock callbacks
565 Details : Instead of sending blocking and completion callbacks as separated
566 requests, adding them to a set and sending in parallel.
569 Frequency : only for Cray XT3
570 Bugzilla : 12829/13455
571 Description: Changing primary group doesn't change the group lustre assigns to
573 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
574 overridden with the primary group provided by the group upcall,
575 whereas the supplied fsgid can be trusted if it is in the list of
576 supplementary groups returned by the group upcall.
578 Severity : enhancement
580 Description: Allow masking D_WARNING, D_ERROR messages from console
581 Details : Console messages can now be disabled via lnet.printk.
585 Description: User code with malformed file open parameter crashes client node
586 Details : Before packing join_file req, all the related reference should be
587 checked carefully in case some malformed flags cause fake
588 join_file req on client.
590 --------------------------------------------------------------------------------
592 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
594 * Support for kernels:
595 2.6.5-7.283 (SLES 9),
596 2.6.9-55.EL (RHEL 4),
597 2.6.16.46-0.14 (SLES 10),
598 2.6.18.8 vanilla (kernel.org)
599 * Client support for unpatched kernels:
600 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
601 2.6.16 - 2.6.22 vanilla (kernel.org)
602 * Due to recently discovered recovery problems, we do not recommend
603 using patchless RHEL 4 clients with this or any earlier release.
604 * Recommended e2fsprogs version: 1.39.cfs8
605 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
606 * Starting with this release, the ldiskfs backing filesystem required
607 by Lustre is now in its own package, lustre-ldiskfs. This package
608 should be installed. It is versioned separately from Lustre and
609 may be released separately in future.
613 Description: block reactivating mgc import until all deactivates complete
614 Details : Fix race when failing back MDT/MGS to itself (testing)
617 Frequency : at statup only
619 Description: mds_lov_synchronize race leads to various problems
620 Details : simultaneous MDT->OST connections at startup can cause the
621 sync to abort, leaving the OSC in a bad state.
623 Severity : enhancement
625 Description: add optional extra BUILD_VERSION info
626 Details : add a new environment variable (namely LUSTRE_VERS) which allows
627 to override the lustre version.
630 Frequency : 2.6.18 servers only
632 Description: ll_kern_mount() doesn't release the module reference
633 Details : The ldiskfs module reference count never drops down to 0
634 because ll_kern_mount() doesn't release the module reference.
639 Description: server LBUG when using old ost_num_threads parameter
640 Details : Accept the old ost_num_threads parameter but warn that it
641 is deprecated, and fix an off-by-one error that caused an LBUG.
646 Description: Transient SCSI error results in persistent IO issue
647 Details : iobuf->dr_error is not reinitialized to 0 between two
651 Frequency : sometimes when underlying device returns I/O errors
653 Description: OSTs not going read-only during write failures
654 Details : OSTs are not remounted read-only when the journal commit threads
655 get I/O errors because fsfilt_ext3 calls journal_start/stop()
656 instead of the ext3 wrappers.
660 Description: poor connect scaling with increasing client count
661 Details : Don't run filter_grant_sanity_check for more than 100 exports
662 to improve scaling for large numbers of clients.
665 Frequency : SLES10 only
667 Description: sanity-quota.sh quotacheck failed: rc = -22
668 Details : Quotas cannot be enabled on SLES10.
671 Frequency : liblustre clients only
673 Description: getdirentries does not give error when run on compute nodes
674 Details : getdirentries does not fail when the size specified as an argument
675 is too small to contain at least one entry
677 Severity : enhancement
679 Description: Add LNET router traceability for debug purposes
680 Details : If a checksum failure occurs with a router as part of the
681 IO path, the NID of the last router that forwarded the bulk data
682 is printed so it can be identified.
687 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
688 Details : Due to a race condition, liblustre clients were occasionally
691 Severity : enhancement
693 Description: lfs setstripe use optional parameters instead of postional
696 Severity : enhancement
698 Description: Nanosecond timestamp support for ldiskfs
699 Details : The on-disk ldiskfs filesystem has added support for nanosecond
700 resolution timestamps. There is not yet support for this at
701 the Lustre filesystem level.
704 Frequency : during server recovery
706 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
707 Details : request with rq_no_resend flag not awake l_wait_event if they get a
711 Frequency : nfs export on patchless client
713 Description: connectathon hang when test nfs export over patchless client
714 Details : Disconnected dentry cannot be found with lookup, so we do not need
715 to unhash it or make it invalid
718 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
719 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
720 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
725 Description: open req refcounting wrong on reconnect
726 Details : If reconnect happened between getting open reply from server and
727 call to mdc_set_replay_data in ll_file_open, we will schedule
728 replay for unreferenced request that we are about to free.
729 Subsequent close will crash in variety of ways.
730 Check that request is still eligible for replay in
731 mdc_set_replay_data().
736 Description: disable writes to filesystem when reading health_check file
737 Details : the default for reading the health_check proc file has changed
738 to NOT do a journal transaction and write to disk, because this
739 can cause reads of the /proc file to hang and block HA state
740 checking on a healthy but otherwise heavily loaded system. It
741 is possible to return to the previous behaviour during configure
742 with --enable-health-write.
744 Severity : enhancement
746 Description: 64-bit inode version
747 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
748 to the inode. This will be required for version-based recovery.
753 Description: MDS fails to start if a duplicate client export is detected
754 Details : in some rare cases it was possible for a client to connect to
755 an MDS multiple times. Upon recovery the MDS would detect this
756 and fail during startup. Handle this more gracefully.
758 Severity : enhancement
760 Description: Add -o localflock option to simulate old noflock
762 Details : This will achieve local-only flock/fcntl locks
768 Description: log_commit_thread vs filter_destroy race leads to crash
769 Details : Take import reference before releasing llog record semaphore
774 Description: Wrong request locking in request set processing
775 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
776 imp_delayed_list, in this place should be used imp_lock.
779 Frequency : when reconnection
781 Description: Grant Leak when osc reconnect to OST
782 Details : When osc reconnect ost, OST(filter) should check whether it
783 should grant more space to client by comparing fed_grant and
784 cl_avail_grant, and return the granted space to client instead
785 of "new granted" space, because client will call osc_init_grant
786 to update the client grant space info.
789 Frequency : when client reconnect to OST
791 Description: Grant Leak when osc do resend and replay bulk write
792 Details : When osc reconnect to OST, OST(filter)should clear grant info of
793 bulk write request, because the grant info while be sync between
794 OSC and OST when reconnect, and we should ignore the grant info
795 these of resend/replay write req.
800 Description: Grant space more than avaiable left space sometimes.
801 Details : When then OST is about to be full, if two bulk writing from
802 different clients came to OST. Accord the avaliable space of the
803 OST, the first req should be permitted, and the second one
804 should be denied by ENOSPC. But if the seconde arrived before
805 the first one is commited. The OST might wrongly permit second
806 writing, which will cause grant space > avaiable space.
809 Frequency : when client is evicted
811 Description: Grant might be wrongly erased when osc is evicted by OST
812 Details : when the import is evicted by server, it will fork another
813 thread ptlrpc_invalidate_import_thread to invalidate the
814 import, where the grant will be set to 0. While the original
815 thread will update the grant it got when connecting. So if
816 the former happened latter, the grant will be wrongly errased
817 because of this race.
822 Description: Checking Stale with correct fid
823 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
824 to check whether it is stale, because sometimes, we want the
825 enqueue happened anyway, and op_data.fid2 will not be initialized.
827 Severity : enhancement
829 Description: update patchless client
830 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
833 Frequency : only with 2.4 kernel
835 Description: random memory corruption
836 Details : size of struct ll_inode_info is to big for union inode.u and this
837 can be cause of random memory corruption.
842 Description: Memory leak in recovery
843 Details : Lov_mds_md was not free in an error handler in mds_create_object.
844 It should also check obd_fail before fsfilt_start, otherwise if
845 fsfilt_start return -EROFS,(failover mds during mds recovery).
846 then the req will return with repmsg->transno = 0 and rc = EROFS.
847 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
848 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
849 be freed no matter whether fsfilt_commit success or not.
852 Frequency : only with huge count clients
854 Description: Prevents from taking the superblock lock in llap_from_page for
856 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
857 for llap_from_page call in ll_removepage prevents from taking the
858 superblock lock for a soon died page.
863 Description: Not check open intent error before release open handle
864 Details : in some rare cases, the open intent error is not checked before
865 release open handle, which may cause
866 ASSERTION(open_req->rq_transno != 0), because it tries to release
867 the failed open handle.
872 Description: Set cat log bitmap only after create log success.
873 Details : in some rare cases, the cat log bitmap is set too early. and it
874 should be set only after create log success.
878 Description: Accessing a block bevice can re-enable I/O when Lustre is
879 tearing down a device.
880 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
885 Description: service threads may hog cpus when there are a lot of requests
886 Details : Insert cond_resched to give other threads a chance to use some CPU
891 Description: the cat log was not initialized in recovery
892 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
893 unlink log on mds will not be initialized until mds post
894 recovery. And also in mds post recovery, the unlink log will
895 initialization will be done asynchronausly, so there will be race
896 between add unlink log and unlink log initialization.
900 Description: brw_stats were being printed incorrectly
901 Details : brw_stats were being printed as log2 but all of them were not
902 recorded as log2. Also remove some code duplication arising from
903 filter_tally_{read,write}.
907 Frequency : rare, only in recovery.
908 Description: ASSERTION(req->rq_type != LI_POISON) failed
909 Details : imp_lock should be held while iterating over imp_sending_list for
910 prevent destroy request after get timeout in ptlrpc_queue_wait.
914 Description: replay-single.sh test 52 fails
915 Details : A lock's skiplist need to be cleanup when it being unlinked
916 from its resource list.
920 Description: Short directio read returns full requested size rather than
922 Details : Direct I/O operations should return actual amount of bytes
923 transferred rather than requested size.
925 Severity : enhancement
927 Description: metadata RPC reduction (e.g. for rm performance)
928 Details : decrease the amount of synchronous RPC between clients and servers
929 by canceling conflicing lock before the operation on the client
930 and packing thier handles into the main operation RPC to server.
932 Severity : enhancement
934 Description: Async OSC create to avoid the blocking unnecessarily.
935 Details : If a OST has no remain object, system will block on the creating
936 when need to create a new object on this OST. Now, ways use
937 pre-created objects when available, instead of blocking on an
938 empty osc while others are not empty. If we must block, we block
939 for the shortest possible period of time.
941 Severity : enhancement
943 Description: refine locking for avoid write wrong info into lov_objid file
944 Details : fix possible races with add new target and write/update data in
947 --------------------------------------------------------------------------------
949 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
954 Frequency : on some architectures
956 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
957 Details : Uninitialized flags sometimes cause configuration commands to
961 Frequency : patchless clients only
963 Description: missing __iget() symbol export
964 Details : The __iget() symbol export is missing. To avoid the need for
965 this on patchless clients the deathrow inode reaper is turned
966 off, and we depend on the VM to clean up old inodes. This
967 dependency was during via the fix for bug 12181.
969 --------------------------------------------------------------------------------
971 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
973 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
974 INTEROPERATE with older versions automatically. In many cases a
975 special upgrade step is needed. Please read the
976 user documentation before upgrading any part of a live system.
977 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
978 version will not interoperate with 1.6 betas before beta5 (1.5.95).
979 * WARNING: Lustre configuration and startup changes are required with
980 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
986 Severity : enhancement
988 Description: MountConf
989 Details : Lustre configuration is now managed via mkfs and mount
990 commands instead of lmc and lconf. New obd types (MGS, MGC)
991 are added for dynamic configuration management. See
992 https://mail.clusterfs.com/wikis/lustre/MountConf for
995 Severity : enhancement
997 Description: dynamic OST addition
998 Details : OSTs can now be added to a live filesystem
1000 Severity : enhancement
1002 Description: startup order invariance
1003 Details : MDTs and OSTs can be started in any order. Clients only
1004 require the MDT to complete startup.
1006 Severity : enhancement
1008 Description: parallel, asynchronous orphan cleanup
1009 Details : orphan cleanup is now performed in separate threads for each
1010 OST, allowing parallel non-blocking operation.
1012 Severity : enhancement
1014 Description: optimized stripe assignment
1015 Details : stripe assignments are now made based on ost space available,
1016 ost previous usage, and OSS previous usage, in order to try
1017 to optimize storage space and networking resources.
1019 Severity : enhancement
1021 Description: Permanently set tunables
1022 Details : All writable /proc/fs/lustre tunables can now be permanently
1023 set on a per-server basis, at mkfs time or on a live system.
1025 Severity : enhancement
1027 Description: Lustre message v2
1028 Details : Add lustre message format v2.
1030 Severity : enhancement
1032 Description: client OST exclusion list
1033 Details : Clients can be started with a list of OSTs that should be
1034 declared "inactive" for known non-responsive OSTs.
1038 Description: SPEC SFS validation failure on NFS v2 over lustre.
1039 Details : Changes the blocksize for regular files to be 2x RPC size,
1040 and not depend on stripe size.
1042 Severity : enhancement
1044 Description: Multiple MD RPCs in flight.
1045 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1046 To avoid overly-overloading MDS, introduce a limit on number of
1047 MDS RPCs in flight for a single client and add /proc controls
1048 to adjust this limit.
1050 Severity : enhancement
1052 Description: client read/write statistics
1053 Details : Add client read/write call usage stats for performance
1054 analysis of user processes.
1055 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1056 file access. extents_stats shows chunk size distribution.
1057 extents_stats_per_process show chunk size distribution per
1060 Severity : enhancement
1062 Description: mds statistics
1063 Details : Add detailed mds operations statistics in
1064 /proc/fs/lustre/mds/*/stats.
1068 Description: Failure of copying files with lustre special EAs.
1069 Details : Client side always return success for setxattr call for lustre
1070 special xattr (currently only "trusted.lov").
1075 Description: Refcount LNET uuids
1076 Details : The global LNET uuid list grew linearly with every startup;
1077 refcount repeated list entries instead of always adding to
1080 Severity : enhancement
1082 Description: Dynamic service threads
1083 Details : Within a small range, start extra service threads
1084 automatically when the request queue builds up.
1087 Frequency : mixed-endian client/server environments
1089 Description: mixed-endian crashes
1090 Details : The new msg_v2 system had some failures in mixed-endian
1093 Severity : enhancement
1095 Description: Easy OST removal
1096 Details : OSTs can be permanently deactivated with e.g. 'lctl
1097 conf_param lustre-OST0001.osc.active=0'
1099 Severity : enhancement
1101 Description: MGS proc entries
1102 Details : Added basic proc entries for the MGS showing what filesystems
1105 Severity : enhancement
1107 Description: provide MGS failover
1108 Details : Added config lock reacquisition after MGS server failover.
1110 Severity : enhancement
1112 Description: add Linux 2.4 support
1113 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1117 Description: a large application tries to do I/O to the same resource and dies
1118 in the middle of it.
1119 Details : Check the req->rq_arrival time after the call to
1120 ost_brw_lock_get(), but before we do anything about
1121 processing it & sending the BULK transfer request. This
1122 should help move old stale pending locks off the queue as
1123 quickly as obd_timeout.
1126 Frequency : when an incorrect nid is specified during startup
1128 Description: ptlrpc connect to non-existant node causes kernel crash
1129 Details : LNET can't be re-entered from an event callback, which
1130 happened when we expire a message after the export has been
1131 cleaned up. Instead, hand the zombie cleanup off to another
1134 Severity : enhancement
1136 Description: plain/inodebits lock performance improvement
1137 Details : Grouping plain/inodebits in granted list by their request modes
1138 and bits policy, thus improving the performance of search through
1142 Frequency : only if OST filesystem is corrupted
1144 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1145 Details : for a short time RPCs with bulk IO are in the replay list,
1146 but replay of bulk IOs is unimplemented. If the OST filesystem
1147 is corrupted due to disk cache incoherency and then replay is
1148 started it is possible to trip an assertion. Avoid putting
1149 committed RPCs into the replay list at all to avoid this issue.
1152 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1154 Description: System hang on startup
1155 Details : This bug allowed the liblustre (e.g. catamount) client to
1156 return to the app before handling all startup RPCs. This
1157 could leave the node unresponsive to lustre network traffic
1158 and manifested as a server ptllnd timeout.
1160 Severity : enhancement
1162 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1163 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1164 boolean to control whether to print peer debug info when a
1165 client's RPC times out.
1168 Frequency : only for kernels with patches from Lustre below 1.4.3
1170 Description: Remove old rdonly API
1171 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1174 Frequency : only for devices with external journals
1176 Description: Set external device read-only also
1177 Details : During a commanded failover stop, we set the disk device
1178 read-only while the server shuts down. We now also set any
1179 external journal device read-only at the same time.
1182 Frequency : when upgrading from 1.4 while trying to change parameters
1184 Description: The wrong (new) MDC name was used when setting parameters for
1185 upgraded MDT's. Also allows changing of OSC (and MDC)
1186 parameters if --writeconf is specified at tunefs upgrade time.
1189 Frequency : when setting specific ost indicies
1191 Description: QOS code breaks on skipped indicies
1192 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1193 created with --index need not be sequential.
1195 Severity : enhancement
1197 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1198 Details : The uninit_groups feature works in conjunction with the kernel
1199 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1200 up the pass1 processing of e2fsck. This is a read-only feature
1201 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1202 mount filesystems that have had this feature enabled.
1204 Severity : enhancement
1206 Description: Improve multi-block allocation algorithm to avoid fragmentation
1207 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1208 allocation locality and avoid filesystem fragmentation.
1210 ------------------------------------------------------------------------------
1212 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1214 * Support for kernels:
1215 2.4.21-47.0.1.EL (RHEL 3)
1216 2.6.5-7.283 (SLES 9)
1217 2.6.9-42.0.10.EL (RHEL 4)
1218 2.6.12.6 vanilla (kernel.org)
1219 2.6.16.27-0.9 (SLES 10)
1220 * Recommended e2fsprogs version: 1.39.cfs5
1222 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1226 Frequency : occasional, depends on client load and configuration
1227 Bugzilla : 12181, 12203
1228 Description: data loss for recently-modified files
1230 Details : In some cases it is possible that recently written or created
1231 files may not be written to disk in a timely manner (this should
1232 normally be within 30s unless client IO load is very high).
1233 The problem appears as zero-length files or files that are a
1234 multiple of 1MB in size after a client crash or client eviction
1235 that are missing data at the end of the file.
1237 This problem is more likely to be hit on clients where files are
1238 repeatedly created and unlinked in the same directory, clients
1239 have a large amount of RAM, have many CPUs, the filesystem has
1240 many OSTs, the clients are rebooted frequently, and/or the files
1241 are not accessed by other nodes after being written.
1243 The presence of the problem can be detected by looking at
1244 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1245 smaller than the second (nr_unused) then dirty files will not
1246 be flushed automatically to disk. "sync; sleep 10" should be
1247 run several times on the node before unmounting it to update
1248 Lustre (this is also safe to run on nodes without this problem).
1250 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1251 that can cause this same problem, so customers using that kernel
1252 also need to update the kernel in addition to Lustre. In order
1253 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1255 It is normal that files written just before a client crash (less
1256 than 30s) may not yet have been flushed to disk, even for local
1260 Frequency : frequent on thin XT3 nodes
1262 Description: UUID collision on thin XT3 Linux nodes
1263 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1264 randomly, since we relied on an insufficiently-seeded PRNG.
1269 Description: OSS hangs after "All ost request buffers busy"
1270 Details : A deadlock between quota and journal operations caused OSS
1271 hangs after printing "All ost request buffers busy."
1274 Frequency : always on liblustre builds
1276 Description: Cleanup compiler warnings on liblustre
1279 Frequency : always on liblustre builds on XT3
1281 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1285 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1286 > 32000 subdirectories
1287 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1288 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1289 subdirectory count crosses 32000. This will aid e2fsck to
1290 correctly handle more than 32000 subdirectories.
1293 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1295 Description: System hang on startup
1296 Details : This bug allowed the liblustre (e.g. catamount) client to
1297 return to the app before handling all startup RPCs. This
1298 could leave the node unresponsive to lustre network traffic
1299 and manifested as a server ptllnd timeout.
1301 Severity : enhancement
1303 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1304 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1305 boolean to control whether to print peer debug info when a
1306 client's RPC times out.
1311 Description: make O_SYNC working on 2.6 kernels
1312 Details : 2.6 kernels use different method for mark pages for write,
1313 so need add a code to lustre for O_SYNC work.
1318 Description: Failure to close file and release space on NFS
1319 Details : Put inode details into lock acquired in ll_intent_file_open.
1320 Use mdc_intent_lock in ll_intent_open to properly
1321 detect all kind of errors unhandled by mdc_enqueue.
1326 Description: proc file read during shutdown sometimes raced obd removal,
1328 Details : Add lock to prevent obd access after proc file removal.
1331 Frequency : Only for files larger than 4GB on 32-bit clients.
1333 Description: improperly doing page alignment of locks
1334 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1335 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1340 Description: RPCs being resent when they shouldn't be
1341 Details : Some RPCs that should not be resent are being resent. This
1342 can cause inconsistencies in the RPC state machine. Do not
1343 resend such requests.
1346 Frequency : rare, only with NFS export
1348 Description: Crash on NFS re-export node
1349 Details : under very unusual load conditions an assertion is hit in
1350 ll_intent_file_open()
1353 Frequency : only if OST filesystem is corrupted
1355 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1356 Details : for a short time RPCs with bulk IO are in the replay list,
1357 but replay of bulk IOs is unimplemented. If the OST filesystem
1358 is corrupted due to disk cache incoherency and then replay is
1359 started it is possible to trip an assertion. Avoid putting
1360 committed RPCs into the replay list at all to avoid this issue.
1365 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1366 Details : Large single O_DIRECT read and write calls can fail to allocate
1367 a sufficiently large buffer to process the request. In case of
1368 allocation failure the allocation is retried with a smaller
1369 buffer and broken into smaller requests.
1371 Severity : enhancement
1373 Description: Add -o localflock option to simulate old noflock behaviour.
1374 Details : This will achieve local-only flock/fcntl locks coherentness.
1379 Description: versioning check is incomplete
1380 Details : Checking the version difference of client vs. server, report
1381 error if the gap is too big.
1386 Description: add support PG_writeback bit
1387 Details : add support for PG_writeback bit for Lustre, for more carefull
1388 work with page cache in 2.6 kernel. This also fix some deadlocks
1389 and remove hack for work O_SYNC with 2.6 kernel.
1391 Severity : enhancement
1393 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1394 Details : The uninit_groups feature works in conjunction with the kernel
1395 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1396 up the pass1 processing of e2fsck. This is a read-only feature
1397 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1398 mount filesystems that have had this feature enabled.
1400 Severity : enhancement
1402 Description: Improve multi-block allocation algorithm to avoid fragmentation
1403 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1404 allocation locality and avoid filesystem fragmentation.
1406 ------------------------------------------------------------------------------
1408 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1410 * Support for kernels:
1411 2.6.9-42.0.3.EL (RHEL 4)
1412 2.6.5-7.276 (SLES 9)
1413 2.4.21-47.0.1.EL (RHEL 3)
1414 2.6.12.6 vanilla (kernel.org)
1415 2.6.16.21-0.8 (SLES10)
1416 * Recommended e2fsprogs version: 1.39.cfs2-0
1418 * The backwards-compatible /proc/sys/portals symlink has been removed
1419 in this release. Before upgrading, please ensure that you change
1420 any configuration scripts or /etc/sysctl.conf files that access
1421 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1422 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1423 in advance of the upgrade on any system running Lustre 1.4.6 or
1424 newer, since /proc/sys/lnet was added in that version.
1425 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1429 Frequency : only when quota is used
1431 Description: avoid scanning export list for quota master
1432 Details : Change the algorithms to avoid scanning export list in order
1433 to improve the efficiency.
1436 Frequency : MDS failover only, very rarely
1438 Description: "went back in time" messages on mds failover
1439 Details : The greatest transno may be lost when the current operation
1440 finishes with an error (transno==0) and the client's last_rcvd
1441 record is over-written. Save the greatest transno in the
1442 mds_last_transno for this case.
1445 Frequency : always for specific kernels and striping counts
1447 Description: client may get "Matching packet too big" without ACL support
1448 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1449 when trying to access files in certain configurations. The
1450 clients should in fact be denied when mounting because they do
1451 not understand ACLs.
1454 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1456 Description: many clients connecting with IO in progress causes connect timeouts
1457 Details : Avoid synchronous journal commits to avoid delays caused by many
1458 clients connecting/disconnecting when bulk IO is in progress.
1459 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1460 OST_IO_PORTAL to avoid delays behind potentially many pending
1464 Frequency : occasionally with multiple writers to a single file
1466 Description: shared writes to file may result in wrong size reported by stat()
1467 Details : Allow growing of kms when extent lock is cancelled
1470 Frequency : always with random mmap IO to multi-striped file
1472 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1473 Details : Only if the hole is at the end of OST object so that kms is too
1474 small. Fix is to increase kms accordingly in ll_nopage.
1477 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1479 Description: writes to a missing object would leak memory on the OST
1480 Details : If there is an inconsistency between the MDS and OST filesystems,
1481 such that the MDS references an object that doesn't exist, writes
1482 to that object will leak memory due to incorrect cleanup in the
1483 error handling path, eventually running out of memory on the OST.
1488 Description: Creating too long symlink causes lustre errors
1489 Details : Check symlink and name lengths before sending requests to MDS.
1492 Frequency : only if flock is enabled (not on by default)
1494 Description: posix locks not released on fd closure on 2.6.9+
1495 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1496 kernels, this caused such locks not to be released on fd close and
1497 then assertions on fs unmount about still used locks.
1500 Frequency : MDS failover only, very rarely
1502 Description: clients may get ASSERTION(granted_lock != NULL)
1503 Details : When request was taking a long time, and a client was resending
1504 a getattr by name lock request. The were multiple lock requests
1505 with the same client lock handle and
1506 mds_getattr_name->fixup_handle_for_resent_request found one of the
1507 lock handles but later failed with ASSERTION(granted_lock != NULL).
1512 Description: handle->h_buffer_credits > 0, assertion failure
1513 Details : h_buffer_credits is zero after truncate, causing assertion
1514 failure. This patch extends the transaction or creates a new
1518 Frequency : NFS re-export or patchless client
1519 Bugzilla : 11179, 10796
1520 Description: Crash on NFS re-export node (__d_move)
1521 Details : We do not want to hash the dentry if we don't have a lock.
1522 But if this dentry is later used in d_move, we'd hit uninitialised
1523 list head d_hash, so we just do this to init d_hash field but
1524 leave dentry unhashed.
1527 Frequency : NFS re-export or patchless client
1529 Description: NFS exports has problem with symbolic link
1530 Details : lustre client didn't properly install dentry when re-exported
1531 to NFS or running patchless client.
1534 Frequency : NFS re-export or patchless client
1536 Description: Various nfs/patchless fixes.
1537 Details : fixes reuse disconected alias for lookup process - this fixes
1538 warning "find_exported_dentry: npd != pd",
1539 fix permission error with open files at nfs.
1540 fix apply umask when do revalidate.
1543 Frequency : occasional
1545 Description: Crash on NFS re-export node
1546 Details : calling clear_page() on the wrong pointer triggered oops in
1547 generic_mapping_read().
1550 Frequency : rarely, using O_DIRECT IO
1552 Description: unaligned directio crashes client with LASSERT
1553 Details : check for unaligned buffers before trying any requests.
1556 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1558 Description: stale data returned from RAID cache
1559 Details : If only a small amount of IO is done to the RAID device before
1560 reading it again it is possible to get stale data from the RAID
1561 cache instead of reading it from disk.
1564 Frequency : always for sles10 kernel
1566 Description: sles10 support
1567 Details : ll_follow_link: compile fixes and using of nd_set_link
1568 under newer kernels.
1571 Frequency : depends on arch, kernel and compiler version, always on sles10
1574 Description: recursive or deep enough symlinks cause stack overflow
1575 Details : getting rid of large stack-allocated variable in
1579 Frequency : depends on hardware
1581 Description: lustre write performance loss in the SLES10 kernel
1582 Details : the performance loss is caused by using of write barriers in the
1583 ext3 code. The SLES10 kernel turns barrier support on by
1584 default. The fix is to undo that change for ldiskfs.
1586 ------------------------------------------------------------------------------
1588 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1590 * Support for kernels:
1591 2.6.9-42.0.3EL (RHEL 4)
1592 2.6.5-7.276 (SLES 9)
1593 2.4.21-47.0.1.EL (RHEL 3)
1594 2.6.12.6 vanilla (kernel.org)
1598 Frequency : quota enabled and large files being deleted
1600 Description: releasing more than 4GB of quota at once hangs OST
1601 Details : If a user deletes more than 4GB of files on a single OST it
1602 will cause the OST to spin in an infinite loop. Release
1603 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1608 Description: statfs data retrieved from /proc may be stale or zero
1609 Details : When reading per-device statfs data from /proc, in the
1610 {kbytes,files}_{total,free,avail} files, it may appear
1611 as zero or be out of date.
1614 Frequency : systems with MD RAID1 external journal devices
1616 Description: lconf's call to blkid is confused by RAID1 journal devices
1617 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1618 of returning all block devices that match the journal UUID.
1621 Frequency : always, for aggregate stripe size over 4GB
1623 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1624 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1625 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1628 Frequency : always if "lfs find" used on a local file/directory
1630 Description: "lfs find" segfaults if used on a local file/directory
1631 Details : The case where a directory component was not specified wasn't
1632 handled correctly. Handle this properly.
1635 Frequency : always on ppc64
1637 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1638 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1639 perform endianness conversion.
1642 Frequency : rarely (truncate to non-zero file size after write under load)
1643 Bugzilla : 10730, 10687
1644 Description: Files padded with zeros to next 4K multiple
1645 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1646 it is possible that files that are truncated to a non-zero size
1647 immediately after being written are filled with zero bytes beyond
1648 the truncated size. No file data is lost.
1650 Severity : enhancement
1652 Description: Allow recovery/failover for liblustre clients.
1653 Details : liblustre clients were unaware of failover configurations until
1656 Severity : enhancement
1658 Description: user file locks should fail when not mounting with flock option
1659 Details : Set up an error-returning stub in ll_file_operations.lock field
1660 to prevent incorrect behaviour when client is mounted without
1661 flock option. Also, set up properly f_op->flock field for
1665 Frequency : always on ia64
1667 Description: "lfs df" loops on printing out MDS statfs information
1668 Details : The obd_ioctl_data was not initialized and in some systems
1669 this caused a failure during the ioctl that did not return
1670 an error. Initialize the struct and return an error on failure.
1673 Frequency : SLES 9 only
1675 Description: Error of copying files with lustre special EAs as root
1676 Details : Client side always return success for setxattr call for lustre
1677 special xattr (currently only "trusted.lov").
1680 Frequency : rarely on clusters with both ia64+i386 clients
1682 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1683 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1684 clients are doing concurrent writes to the same file it is
1685 possible that the ia64 clients may LASSERT because the OST
1686 extent locks are not PAGE_SIZE aligned. Ensure that grown
1687 locks are always aligned on the request boundary.
1690 Frequency : specific use, occasional
1692 Description: Overwriting in use executable truncates on-disk binary image
1693 Details : If one node attempts to overwrite an executable in use by
1694 another node, we now correctly return ETXTBSY instead of
1695 truncating the file.
1697 Severity : enhancement
1699 Description: Async OSC create to avoid the blocking unnecessarily.
1700 Details : If a OST has no remain object, system will block on the creating
1701 when need to create a new object on this OST. Now, ways use
1702 pre-created objects when available, instead of blocking on an
1703 empty osc while others are not empty. If we must block, we block
1704 for the shortest possible period of time.
1709 Description: chmod on Lustre root is propagated to other clients
1710 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1711 invalid by the follow_mount time.
1714 Frequency : liblustre clients only
1716 Description: Race in 'instant cancel' lock handling could lead to such locks
1717 never to be granted in case of SMP MDS
1718 Details : Do not destroy not yet granted but cbpending locks in
1722 Frequency : replay/resend of open
1724 Description: non null lock assetion failure in mds_intent_policy
1725 Details : Trying to replay/resend lockless open requests resulted in
1726 mds_open() returning 0 with no lock. Now it sets a flag if
1727 a lock is going to be returned.
1729 Severity : enhancement
1731 Description: Checksum enhancements
1732 Details : New checksum enhancements allow for resending RPCs that failed
1735 Severity : enhancement
1737 Description: Tunables on number of dirty pages in cacche
1738 Details : Allow to set limit on number of dirty pages cached.
1743 Description: client crash on unmount - lock still has references
1744 Details : In some error handling cases it was possible to leak a lock
1745 reference on a client while accessing a file. This was not
1746 harmful to the client during operation, but would cause the
1747 client to crash when the filesystem is unmounted.
1750 Frequency : specific case, rare
1752 Description: ETXTBSY on mds though file not in use by client
1753 Details : ETXTBSY is no longer incorrectly returned when attempting to
1754 chmod or chown a directory that the user previously tried to
1755 execute or a currently-executing binary.
1758 Frequency : extremely rare except on liblustre-based clients
1760 Description: Lustre space not freed when files are deleted
1761 Details : Clean up open-unlinked files after client eviction. Previously
1762 the unlink was skipped and the files remained as orphans.
1767 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1768 Details : In some cases it was possible to send a blocking callback to
1769 a client doing a glimpse, even though that client didn't get
1770 a lock granted. When the glimpse lock is cancelled on the OST
1771 the freed lock is left on the waiting list and corrupted the list.
1774 Frequency : all core dumps
1776 Description: Broke core dumps to lustre
1777 Details : Negative dentry may be unhashed if parent does not have UPDATE
1778 lock, but some callers, e.g. do_coredump, expect dentry to be
1779 hashed after successful create, hash it in ll_create_it.
1781 ------------------------------------------------------------------------------
1783 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1785 * Support for kernels:
1786 2.6.9-42.0.2.EL (RHEL 4)
1787 2.6.5-7.276 (SLES 9)
1788 2.4.21-40.EL (RHEL 3)
1789 2.6.12.6 vanilla (kernel.org)
1793 Frequency : always on RHEL 3
1795 Description: Number of open files grows over time
1796 Details : The number of open files grows over time, whether or not
1797 Lustre is started. This was due to a filp leak introduced
1798 by one of our kernel patches.
1800 ------------------------------------------------------------------------------
1802 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1804 * Support for kernels:
1805 2.6.9-42.EL (RHEL 4)
1806 2.6.5-7.276 (SLES 9)
1807 2.4.21-40.EL (RHEL 3)
1808 2.6.12.6 vanilla (kernel.org)
1813 Bugzilla : 5719, 9635, 9792, 9684
1814 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1815 Details : If a server is under heavy load and cannot reply to new
1816 connection requests before the client resends the (re)connect,
1817 the connection handling code can behave badly if two service
1818 threads are concurrently handing separate (re)connections from
1819 the same client. Add better locking to the connection handling
1820 code, and ensure that only a single connection will be processed
1821 for a given client UUID, even if the lock is dropped.
1823 Severity : enhancement
1825 Description: add TCP zero-copy support to kernel
1826 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1827 sends if the hardware supports scatter-gather and checksumming.
1828 This allows socklnd to do client-write and server-read more
1829 efficiently and reduce CPU utilization from skbuf copying.
1832 Frequency : only if NFS exporting from client
1834 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1835 Details : If setting attributes on a file created under NFS that had
1836 never been opened it would be possible to oops the client
1837 if the file had no objects.
1840 Frequency : always for liblustre
1842 Description: liblustre client does MDS+OSTs setattr RPC for each write
1843 Details : When doing a write from a liblustre client, the client
1844 incorrectly issued an RPC to the MDS and each OST the file was
1845 striped over in order to update the timestamps. When writing
1846 with small chunks and many clients this could overwhelm the MDS
1847 with RPCs. In all cases it would slow down the write because
1848 these RPCs are unnecessary.
1850 Severity : enhancement
1852 Description: allow number of MDS service threads to be changed at module load
1853 Details : It is now possible to change the number of MDS service threads
1854 running. Adding "options mds mds_num_threads={N}" to the MDS's
1855 /etc/modprobe.conf will set the number of threads for the next
1856 time Lustre is restarted (assuming the "mds" module is also
1857 reloaded at that time). The default number of threads will
1858 stay the same, 32 for most systems.
1863 Description: OST crash if filesystem is unformatted or corrupt
1864 Details : If an OST is started on a device that has never been formatted
1865 or if the filesystem is corrupt and cannot even mount then the
1866 error handling cleanup routines would dereference a NULL pointer.
1871 Description: NULL pointer deref in llap_from_page.
1872 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1873 incorrect) mapping if the page was truncated/reclaimed while it was
1874 searched for. Check for this condition and skip such pages when
1875 doing readahead. Introduce extra check to llap_from_page() to
1876 verify page->mapping->host is non-NULL (so page is not anonymous).
1879 Frequency : Sometimes when using sys_sendfile
1881 Description: "page not covered by a lock" warnings from ll_readpage
1882 Details : sendfile called ll_readpage without right page locks present.
1883 Now we introduced ll_file_sendfile that does necessary locking
1884 around call to generic_file_sendfile() much like we do in
1888 Frequency : with certain MDS communication failures at client mount time
1890 Description: NULL pointer deref after failed client mount
1891 Details : a client connection request may delayed by the network layer
1892 and not be sent until after the PTLRPC layer has timed out the
1893 request. If the client fails the mount immediately it will try
1894 to clean up before the network times out the request. Add a
1895 reference from the request import to the obd device and delay
1896 the cleanup until the network drops the request.
1899 Frequency : occasionally during client (re)connect
1901 Description: assertion failure during client (re)connect
1902 Details : processing a client connection request may be delayed by the
1903 client or server longer than the client connect timeout. This
1904 causes the client to resend the connection request. If the
1905 original connection request is replied in this interval, the
1906 client may trip an assertion failure in ptlrpc_connect_interpret()
1907 which thought it would be the only running connect process.
1910 Frequency : only with obd_echo servers and clients that are rebooted
1912 Description: kernel BUG accessing uninitialized data structure
1913 Details : When running an obd_echo server it did not start the ping_evictor
1914 thread, and when a client was evicted an uninitialized data
1915 structure was accessed. Start the ping_evictor in the RPC
1916 service startup instead of the OBD startup.
1918 Severity : enhancement
1919 Bugzilla : 10193 (patchless)
1920 Description: Remove dependency on various unexported kernel interfaces.
1921 Details : No longer need reparent_to_init, exit_mm, exit_files,
1922 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1925 Frequency : rare (only users of deprecated and unsupported LDAP config)
1927 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1928 Details : LDAP apparently contains 'lustreName' attributes instead of
1929 'name'. A simple remapping of the name is sufficient.
1932 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1934 Description: waiting_locks_callback trips kernel BUG if client is evicted
1935 Details : Running with the dump_on_timeout debug flag turned on makes
1936 it possible that the waiting_locks_callback() can try to dump
1937 the Lustre kernel debug logs from an interrupt handler. Defer
1938 this log dumping to the expired_lock_main() thread.
1940 Severity : enhancement
1942 Description: Support NFS exporting on 2.6 kernels.
1943 Details : Implement non-rawops metadata methods for NFS server to use without
1944 changing NFS server code.
1947 Frequency : very rare (synthetic metadata workload only)
1949 Description: two racing renames might cause an MDS thread to deadlock
1950 Details : Running the "racer" program may cause one MDS thread to rename
1951 a file from being the source of a rename to being the target of
1952 a rename at exactly the same time that another thread is doing
1953 so, and the second thread has already enqueued these locks after
1954 doing a lookup of the target and is trying to relock them in
1955 order. Ensure that we don't try to re-lock the same resource.
1958 Frequency : only very large systems with liblustre clients
1960 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1961 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1962 all OSTs in parallel. This allows the eviction of stale liblustre
1963 clients to proceed much faster than if they were done in series,
1964 and also offers similar improvements for other set_info RPCs.
1969 Description: excessive CPU usage during initial read phase on client
1970 Details : During the initial read phase on a client, it would agressively
1971 retry readahead on the file, consuming too much CPU and impacting
1972 performance (since 1.4.5.8). Improve the readahead algorithm
1973 to avoid this, and also improve some other common cases (read
1974 of small files in particular, where "small" is files smaller than
1975 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1980 Description: MDS crash when receiving packet with unknown intent.
1981 Details : Do not LBUG in unknown intent case, just return -EFAULT
1983 Severity : enhancement
1984 Bugzilla : 9293, 9385
1985 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1986 Details : Do not serialize getattr (non-intent version) and statfs.
1989 Frequency : occasional, when OST network is overloaded/intermittent
1991 Description: client evicted by OST after bulk IO timeout
1992 Details : If a client sends a bulk IO request (read or write) the OST
1993 may evict the client if it is unresposive to its data GET/PUT
1994 request. This is incorrect if the network is overloaded (takes
1995 too long to transfer the RPC data) or dropped the OST GET/PUT
1996 request. There is no need to evict the client at all, since
1997 the pinger and/or lock callbacks will handle this, and the
1998 client can restart the bulk request.
2001 Frequency : Always when mmapping file with no objects
2003 Description: client crashes when mmapping file with no objects
2004 Details : Check that we actually have objects in a file before doing any
2005 operations on objects in ll_vm_open, ll_vm_close and
2011 Description: Request leak when working with deleted CWD
2012 Details : Introduce advanced request refcount tracking for requests
2013 referenced from lustre intent.
2015 Severity : Enhancement
2017 Description: Cache open file handles on client.
2018 Details : MDS now will return special lock along with openhandle, if
2019 requested and client is allowed to hold openhandle, even if unused,
2020 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2021 closing files for every read/write openration.
2023 Severity : Enhancement
2025 Description: Cache open negative dentries on client when possible.
2026 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2027 negative dentries on lock revocation.
2032 Description: Remounting a client read-only wasn't possible with a zconf mount
2033 Details : It wasn't possible to remount a client read-only with llmount.
2035 Severity : enhancement
2036 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2037 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2038 MPICH is included to add Lustre-specific ADIO interfaces.
2039 This is based closely on the UFS ADIO layer and only differs
2040 in file creation, in order to allow the OST striping to be set.
2041 This is user-contributed code and not supported by CFS.
2046 Description: extended inode attributes (immutable, append-only) work improperly
2047 when 2.4 and 2.6 kernels are used on client/server or vice versa
2048 Details : Introduce kernel-independent values for these flags.
2050 Severity : enhancement
2053 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2054 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2055 granularity. Now, Fractional MB granularity is be supported,
2056 this is very useful for low memory system.
2058 Severity : enhancement
2060 Description: Getattr by fid
2061 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2062 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2063 have locks and avoids getattr rpc requests for it on every stat.
2066 Frequency : Always, for filesystems larger than 2TB
2068 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2069 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2070 filesystems larger than 2TB to be mounted with mballoc enabled.
2073 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2075 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2076 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2077 filesystems larger than 2TB is mounted then the kernel will
2078 silently corrupt the start of the filesystem. CONFIG_LBD is
2079 enabled for all CFS-supported kernels, but the possibility of
2080 this happening with a modified kernel config exists.
2082 Severity : enhancement
2084 Description: add client O_DIRECT support for 2.6 kernels
2085 Details : It is now possible to do O_DIRECT reads and writes to files
2086 in the Lustre client mountpoint on 2.6 kernel clients.
2088 Severity : enhancement
2090 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2091 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2092 parallel, not waiting for response from every OST before sending
2093 a rpc to the next OST.
2098 Description: setattr vs write race when updating file timestamps
2099 Details : Client processes that update a file timestamp into the past
2100 right after writing to the file (e.g. tar) it is possible that
2101 the updated file modification time can be reset to the current
2102 time due to a race between processing the setattr and write RPC.
2104 Severity : enhancement
2106 Description: Bring 'lfs find' closer in line with regular Linux find.
2107 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2108 -print0 options and obtains all the needed info through the lustre
2111 Severity : enhancement
2113 Description: support up to 1024 configured devices on one node
2114 Details : change obd_dev array from statically allocated to dynamically
2115 allocated structs as they are first used to reduce memory usage
2120 Description: Flush dirty partially truncated pages during truncate
2121 Details : Immediatelly flush partially truncated pages in filter_setattr,
2122 this way we completely avoid having any pages in page cache on OST
2123 and can retire ugly workarounds during writes to flush such pages.
2128 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2129 Details : protect inode from truncation within vfs_unlink() context
2130 just take a reference before calling vfs_unlink() and release it
2131 when parent's i_sem is free.
2136 Description: last_id value checked outside lock on OST caused LASSERT failure
2137 Details : If there were multiple MDS->OST object precreate requests in
2138 flight, it was possible that the OST's last object id was checked
2139 outside a lock and incorrectly tripped an assertion. Move checks
2140 inside locks, and discard old precreate requests.
2143 Frequency : always, if extents are used on OSTs
2145 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2146 Details : OSTs using the extents format would not zero the high 16 bits of
2147 the index physical block number. This is not a problem for any
2148 OST filesystems smaller than 16TB, and no kernels support ext3
2149 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2150 new/modified files) and can be fixed for existing filesystems
2151 with e2fsprogs-1.39-cfs1.
2156 Description: import connection selection may be incorrect if timer wraps
2157 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2158 import connections to be ignored if the 32-bit jiffies counter
2159 wraps. Use a 64-bit jiffies counter.
2162 Frequency : very large clusters immediately after boot
2164 Description: LNET request buffers exhausted under heavy short-term load
2165 Details : If a large number of client requests are generated on a service
2166 that has previously never seen so many requests it is possible
2167 that the request buffer growth cannot keep up with the spike in
2168 demand. Instead of dropping incoming requests, they are held in
2169 the LND until the RPC service can accept more requests.
2172 Frequency : Sometimes during replay
2174 Description: Assertion failure in ll_local_open after replay.
2175 Details : If replay happened on an open request reply before we were able
2176 to set replay handler, reply will become not swabbed tripping the
2177 assertion in ll_local_open. Now we set the handler right after
2178 recognising of open request
2181 Frequency : very rare
2183 Description: kernel reports "badness in vsnprintf"
2184 Details : Reading from the "recovery_status" /proc file in small chunks
2185 may cause a negative length in lprocfs_obd_rd_recovery_status()
2186 call to vsnprintf() (which is otherwise harmless). Exit early
2187 if there is no more space in the output buffer.
2189 Severity : enhancement
2191 Description: clear OBD RPC statistics by writing to them
2192 Details : It is now possible to clear the OBD RPC statistics by writing
2193 to the "stats" file.
2198 Description: Client mtime is not the same on different clients after utimes
2199 Details : In some cases, the client was using the utimes() syscall on
2200 a file cached on another node. The clients now validate the
2201 ctime from the MDS + OSTs to determine which one is right.
2206 Description: Inability to activate failout mode
2207 Details : lconf script incorrectly assumed that in pythong string's numeric
2208 value is used in comparisons.
2211 Frequency : always with multiple stripes per file
2213 Description: Inefficient object allocation for mutli-stripe files
2214 Details : When selecting which OSTs to stripe files over, for files with
2215 a stripe count that divides evenly into the number of OSTs,
2216 the MDS is always picking the same starting OST for each file.
2217 Return the OST selection heuristic to the original design.
2222 Description: mount failures may take full timeout to return an error
2223 Details : Under some heavy load conditions it is possible that a
2224 failed mount can wait for the full obd_timeout interval,
2225 possibly several minutes, before reporting an error.
2226 Instead return an error as soon as the status is known.
2228 Frequency : quota enabled and large files being deleted
2230 Description: releasing more than 4GB of quota at once hangs OST
2231 Details : If a user deletes more than 4GB of files on a single OST it
2232 will cause the OST to spin in an infinite loop. Release
2233 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2238 Description: statfs data retrieved from /proc may be stale or zero
2239 Details : When reading per-device statfs data from /proc, in the
2240 {kbytes,files}_{total,free,avail} files, it may appear
2241 as zero or be out of date.
2244 Frequency : systems with MD RAID1 external journal devices
2246 Description: lconf's call to blkid is confused by RAID1 journal devices
2247 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2248 of returning all block devices that match the journal UUID.
2251 Frequency : always, for aggregate stripe size over 4GB
2253 Description: assertion fails when trying to use 4GB stripe size
2254 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2255 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2258 Frequency : always on ppc64
2260 Description: the first write on an ext3 filesystem with mballoc got stuck
2261 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2262 perform endianness conversion.
2264 ------------------------------------------------------------------------------
2266 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2268 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2269 INTEROPERATE with older versions automatically. Please read the
2270 user documentation before upgrading any part of a live system.
2271 * WARNING: Lustre networking configuration changes are required with
2272 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2275 * Support for kernels:
2276 2.6.9-22.0.2.EL (RHEL 4)
2277 2.6.5-7.244 (SLES 9)
2278 2.6.12.6 vanilla (kernel.org)
2281 Severity : enhancement
2282 Bugzilla : 7981/8208
2283 Description: Introduced Lustre Networking (LNET)
2284 Details : LNET is new networking infrastructure for Lustre, it includes
2285 a reorganized network configuration mode (see the user
2286 documentation for full details) as well as support for routing
2287 between different network fabrics. Lustre Networking Devices
2288 (LNDs) for the supported network fabrics have also been
2289 created for this new infrastructure.
2291 Severity : enhancement
2292 Description: Introduced Access control lists
2293 Details : clients can set ACLs on files and directories in order to have
2294 more fine-grained permissions than the standard Unix UGO+RWX.
2295 The MDS must be started with the "-o acl" mount option.
2297 Severity : enhancement
2298 Description: Introduced filesystem quotas
2299 Details : Administrators may now establish per-user quotas on the
2302 Severity : enhancement
2304 Description: Configuration change for the XT3
2305 The PTLLND is now used to run Lustre over Portals on the XT3
2306 The configure option(s) --with-cray-portals are no longer used.
2307 Rather --with-portals=<path-to-portals-includes> is used to
2308 enable building on the XT3. In addition to enable XT3 specific
2309 features the option --enable-cray-xt3 must be used.
2314 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2315 Details : A race between allocating a new llcd and re-getting the llcd_lock
2316 allowed another thread to grab newly-allocated llcd.
2318 Severity : enhancement
2320 Description: 2.6 OST async journal commit and locking fix to improve performance
2321 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2322 2.6 kernels are now async as they already were in 2.4 kernels so
2323 that they can commit concurrently with the network bulk transfer.
2324 For block-allocated files the filter allocation semaphore is held
2325 to avoid filesystem fragmentation during allocation. BKL lock
2326 removed for 2.6 xattr operations where it is no longer needed.
2331 Description: lconf incorrectly determined whether two IP networks could talk
2332 Details : In some more complicated routing and multiple-network
2333 configurations, lconf will avoid trying to make a network
2334 connection to a disjoint part of the IP space. It was doing the
2335 math incorrectly for one set of cases.
2340 Description: Fix for potential infinite loop processing records in an llog.
2341 Details : If an llog record is corrupted/zeroed, it is possible to loop
2342 forever in llog_process(). Validate the llog record length
2343 and skip the remainder of the block on error.
2346 Frequency : occasional (liblustre only)
2348 Description: liblustre could not open files whose last component is a symlink
2349 Details : sysio_path_walk() would incorrectly pass the open intent to
2350 intermediate path components.
2353 Frequency : rare (liblustre only with non-standard tuning)
2354 Bugzilla : 7201 (7350)
2355 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2356 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2357 after it has been released, only do this for referenced locks
2359 Severity : enhancement
2361 Description: specifying an (invalid) directory default stripe_size of -1
2362 would reset the directory default striping
2363 Details : stripe_size -1 was used internally to signal directory stripe
2364 removal, now use "all default" to signal dir stripe removal
2365 as a directory striping of "all default" is not useful
2368 Frequency : common for large clusters running liblustre clients
2370 Description: doing an ls when liblustre clients are running is slow
2371 Details : sending a glimpse AST to a liblustre client waits for every AST
2372 to time out, as liblustre clients will not respond. Since they
2373 cannot cache data we refresh the OST lock LVB from disk instead.
2375 Severity : enhancement
2377 Description: doing an ls at the same time as file IO can be slow
2378 Details : enqueue and other "small" requests can be blocked behind many
2379 large IO requests. Create a new OST IO portal for non-IO
2380 requests so they can be processed faster.
2383 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2385 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2386 created file with mode 000
2387 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2388 allow an ftruncate() as a non-root user to a file with mode 000.
2389 Lustre now respects this flag to disable mode checks when
2390 truncating a file owned by the user
2393 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2396 Description: Revoking locks from clients that went dead or catatonic might take
2398 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2399 cancellation of such locks instant on servers without waiting for
2400 any reply from clients. Clients drops these locks when cancel
2401 notification from server is received without replying.
2404 Frequency : liblustre-only, when liblustre client dies or becomes busy
2406 Description: Doing ls on Linux clients can take a long time with active
2408 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2409 granting such locks to it in the first place if possible. Locks
2410 are taken by proxy on the OST during the read or write and
2411 dropped immediately afterward. Add connect flags handling, do
2412 not grant locks to liblustre clients for glimpse ASTs.
2414 Severity : enhancement
2416 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2417 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2418 sequential reads and behaves badly with applications doing random
2419 reads. Improve it by reading ahead at least read region, and
2420 avoiding excessive large RPC for small reads.
2422 Severity : enhancement
2424 Description: Creating more than 1000 files for a single job may cause a load
2425 imbalance on the OSTs if there are also a large number of OSTs.
2426 Details : qos_prep_create() uses an OST index reseed value that is an
2427 even multiple of the number of available OSTs so that if the
2428 reseed happens in the middle of the object allocation it will
2429 still utilize the OSTs as uniformly as possible.
2434 Description: OST or MDS may oops in ping_evictor_main()
2435 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2436 but doesn't restart at beginning of obd_exports_timed list
2439 Severity : enhancement
2441 Description: improve by-nid export eviction on the MDS and OST
2442 Details : allow multiple exports with the same NID to be evicted at one
2443 time without re-searching the exports list.
2446 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2448 Description: MDS may oops in groups_free()
2449 Details : in rare race conditions a newly allocated group_info struct is
2450 freed again, and this can be NULL. The 2.4 compatibility code
2451 for groups_free() checked for a NULL pointer, but 2.6 did not.
2454 Frequency : common for liblustre clients doing little filesystem IO
2455 Bugzilla : 9352, 7313
2456 Description: server may evict liblustre clients accessing contended locks
2457 Details : if a client is granted a lock or receives a completion AST
2458 with a blocking AST already set it would not reply to the AST
2459 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2462 Frequency : lfs setstripe, only systems with more than 160 OSTs
2464 Description: unable to set striping with a starting offset beyond OST 160
2465 Details : llapi_create_file() incorrectly limited the starting stripe
2466 index to the maximum single-file stripe count.
2469 Frequency : LDAP users only
2471 Description: lconf did not handle in-kernel recovery with LDAP properly
2472 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2474 Severity : enhancement
2476 Description: bind OST threads to NUMA nodes to improve performance
2477 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2478 node and do their allocations there to localize memory access
2480 Severity : enhancement
2482 Description: llmount can determine client NID directly from Myrinet (GM)
2483 Details : the client NID code from gmnalnid was moved directly into
2484 llmount, removing the need to use this or specifying the
2485 client NID explicitly when mounting GM clients with zeroconf
2488 Frequency : if client is started with down MDS
2490 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2491 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2492 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2493 ptlrpc_import_delayed_req() because we didn't check if the
2494 request was interrupted, and we also didn't break out of the
2495 event loop if there was no timeout
2500 Description: data loss during non-page-aligned writes to a single file from
2501 both multiple nodes and multiple threads on one node at same time
2502 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2503 inconsistency led to false short-reads, that were cached and later
2504 used by ->prepare_write() to fill in partially written page,
2505 leading to data loss.
2508 Frequency : always, if lconf --abort_recovery used
2510 Description: lconf --abort_recovery fails with 'Operation not supported'
2511 Details : lconf was attempting to abort recovery on the MDT device and not
2514 Severity : enhancement
2516 Description: remove cleanup logs
2517 Details : replace lconf-generated cleanup logs with lustre internal
2518 cleanup routines. Eliminates the need for client-cleanup and
2521 Severity : enhancement
2523 Description: add support for EAs (user and system) on lustre filesystems
2524 Details : it is now possible to store extended attributes in the Lustre
2525 client filesystem, and with the user_xattr mount option it
2526 is possible to allow users to store EAs on their files also
2528 Severity : enhancement
2530 Description: Add possibility (config option) to show minimal available OST free
2532 Details : When compiled with --enable-mindf configure option, statfs(2)
2533 (and so, df) will return least minimal free space available from
2534 all OSTs as amount of free space on FS, instead of summary of
2535 free spaces of all OSTs.
2537 Severity : enhancement
2539 Description: do not expand extent locks acquired on OST-side
2540 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2541 by server: they are not cached anyway.
2544 Frequency : when mmap is used/binaries executed from Lustre
2546 Description: Unmmap pages before throwing them away from read cache.
2547 Details : llap_shrink cache now attempts to unmap pages before discarding
2548 them (if unmapping failed - do not discard). SLES9 kernel has
2549 extra checks that trigger if this unmapping is not done first.
2554 Description: lconf didn't resolve symlinks before checking to see whether a
2555 given mountpoint was already in use
2558 Frequency : when migrating failover services
2559 Bugzilla : 6395, 9514
2560 Description: When migrating a subset of services from a node (e.g. failback
2561 from a failover service node) the remaining services would
2562 time out and evict clients.
2563 Details : lconf --force (implied by --failover) sets the global obd_timeout
2564 to 5 seconds in order to quickly disconnect, but this caused
2565 other RPCs to time out too quickly. Do not change the global
2566 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2568 Severity : enhancement
2569 Frequency : if MDS is started with down OST
2570 Bugzilla : 9439,5706
2571 Description: Allow startup/shutdown of an MDS without depending on the
2572 availability of the OSTs.
2573 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2574 Add appropriate locking and lov-osc refcounts for safe
2575 cleaning. Add osc abort_inflight calls in case the
2576 synchronize never started.
2579 Frequency : occasional (Cray XT3 only)
2581 Description: root not authorized to access files in CRAY_PORTALS environment
2582 Details : The client process capabilities were not honoured on the MDS in
2583 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2584 been accessed by an authorized user then root was able to access
2585 the file on the local client also. The root user capabilities
2586 are now allowed on the MDS, as this environment has secure UID.
2589 Frequency : occasional
2591 Description: ldiskfs "too long searching" message happens too often
2592 Details : A debugging message (otherwise harmless) prints too often on
2593 the OST console. This has been reduced to only happen when
2594 there are fragmentation problems on the filesystem.
2599 Description: Division by zero in statfs when all OSCs are inactive
2600 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2601 lov_statfs divides by value returned by lov_get_stripecnt().
2605 Bugzilla : 9489, 3273
2606 Description: First write from each client to each OST was only 4kB in size,
2607 to initialize client writeback cache, which caused sub-optimal
2608 RPCs and poor layout on disk for the first writen file.
2609 Details : Clients now request an initial cache grant at (re)connect time
2610 and so that they can start streaming writes to the cache right
2611 away and always do full-sized RPCs if there is enough data.
2612 If the OST is rebooted the client also re-establishes its grant
2613 so that client cached writes will be honoured under the grant.
2618 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2619 Details : Now I/O RPCs go to different portal number and (presumably) fast
2620 lock requests (and glimses) and other RPCs get their own service
2621 threads pool that should be able to service those RPCs
2624 Severity : enhancement
2626 Description: Ability to exchange lustre version between client and servers and
2627 issue warnings at client side if client is too old. Also for
2628 liblustre clients there is ability to refuse connection of too old
2630 Details : New 'version' field is added to connect data structure that is
2631 filled with version info. That info is later checked by server and
2635 Frequency : rare, liblustre only.
2636 Bugzilla : 9296, 9581
2637 Description: Two simultaneous writes from liblustre at offset within same page
2638 might proceed at the same time overwriting eachother with stale
2640 Details : I/O lock withing llu_file_prwv was released too early, before data
2641 actually was hitting the wire. Extended lock-holding time until
2642 server acknowledges receiving data.
2645 Frequency : extremely rare. Never observed in practice.
2647 Description: avoid generating lustre_handle cookie of 0.
2648 Details : class_handle_hash() generates handle cookies by incrementing
2649 global counter, and can hit 0 occasionaly (this is unlikely, but
2650 not impossible, because initial value of cookie counter is
2651 selected randonly). Value of 0 is used as a sentinel meaning
2652 "unassigned handle" --- avoid it. Also coalesce two critical
2653 sections in this function into one.
2655 Severity : enhancement
2657 Description: allow liblustre clients to delegate truncate locking to OST
2658 Details : To avoid overhead of locking, liblustre client instructs OST to
2659 take extent lock in ost_punch() on client's behalf. New connection
2660 flag is added to handle backward compatibility.
2662 Severity : enhancement
2663 Bugzilla : 4928, 7341, 9758
2664 Description: allow number of OST service threads to be specified
2665 Details : a module parameter allows the number of OST service threads
2666 to be specified via "options ost ost_num_threads={N}" in the
2667 OSS's /etc/modules.conf or /etc/modprobe.conf.
2671 Bugzilla : 6146, 9635, 9895
2672 Description: servers crash with bad pointer in target_handle_connect()
2673 Details : In rare cases when a client is reconnecting it was possible that
2674 the connection request was the last reference for that export.
2675 We would temporarily drop the export reference and get a new
2676 one, but this may have been the last reference and the export
2677 was just destroyed. Get new reference before dropping old one.
2679 Severity : enhancement
2680 Frequency : if client is started with failover MDS
2682 Description: Allow multiple MDS hostnames in the mount command
2683 Details : Try to read the configuration from all specified MDS
2684 hostnames during a client mount in case the "primary"
2687 Severity : enhancement
2689 Description: Stop sending data to evicted clients as soon as possible.
2690 Details : Check if the client we are about to send or are sending data to
2691 was evicted already. (Check is done every second of waiting,
2692 for which l_wait_event interface was extended to allow checking
2693 of exit condition at specified intervals).
2696 Frequency : rare, normally only when NFS exporting is done from client
2698 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2700 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2703 Frequency : always, if srand() is called before liblustre initialization
2705 Description: Liblustre uses system PRNG disturbing its usage by user application
2706 Details : Introduce internal to lustre fast and high-quality PRNG for
2707 lustre usage and make liblustre and some other places in generic
2708 lustre code to use it.
2710 Severity : enhancement
2711 Bugzilla : 9477, 9557, 9870
2712 Description: Verify that the MDS configuration logs are updated when xml is
2713 Details : Check if the .xml configuration logs are newer than the config
2714 logs stored on the MDS and report an error if this is the case.
2715 Request --write-conf, or allow starting with --old_conf.
2717 Severity : enhancement
2719 Description: Handle symlinks in the path when checking if Lustre is mounted.
2720 Details : Resolve intermediate symlinks when checking if a client has
2721 mounted a filesystem to avoid duplicate client mounts.
2726 Description: lconf can hit an error exception but still return success.
2727 Details : The lconf command catches the Command error exception at the top
2728 level script context and will exit with the associated exit
2729 status, but doesn't ensure that this exit status is non-zero.
2734 Description: failure of ptlrpc thread startup can cause oops
2735 Details : Starting a ptlrpc service thread can fail if there are a large
2736 number of threads or the server memory is very fragmented.
2737 Handle this without oopsing.
2740 Frequency : always, only if liblustre and non-default acceptor port was used
2742 Description: liblustre cannot connect to servers with non-default acceptor port
2743 Details : tcpnal_set_default_params() was not called and was therefore
2744 ignoring the environment varaible TCPNAL_PORT, as well as other
2745 TCPNAL_ environment variables
2750 Description: two objects could be created on the same OST for a single file
2751 Details : If an OST is down, in some cases it was possible to create two
2752 objects on a single OST for a single file. No problems other
2753 than potential performance impact and spurious error messages.
2757 Bugzilla : 5681, 9562
2758 Description: Client may oops in ll_unhash_aliases
2759 Details : Client dcache may become inconsistent in race condition.
2760 In some cases "getcwd" can fail if the current directory is
2766 Description: Inode refcounting problems in NFS export code
2767 Details : link_raw functions used to call d_instantiate without obtaining
2768 extra inode reference first.
2772 Bugzilla : 9942, 9903
2773 Description: Referencing freed requests leading to crash, memleaks with NFS.
2774 Details : We used to require that call to ll_revalidate_it was always
2775 followed by ll_lookup_it. Also with revalidate_special() it is
2776 possible to call ll_revalidate_it() twice for the same dentry
2777 even if first occurence returned success. This fix changes semantic
2778 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2779 reference on a request referred from the intent.
2780 ll_intent_release() then releases such a request.
2783 Frequency : rare, normally benchmark loads only
2785 Description: unlinked inodes were kept in memory on the client
2786 Details : If a client is repeatedly creating and unlinking files it
2787 can accumulate a lot of stale inodes in the inode slab cache.
2788 If there is no other client load running this can cause the
2789 client node to run out of memory. Instead flush old inodes
2790 from client cache that have the same inode number as a new inode.
2793 Frequency : SLES9 2.6.5 kernel and long filenames only
2794 Bugzilla : 9969, 10379
2795 Description: utime reports stale NFS file handle
2796 Details : SLES9 uses out-of-dentry names in some cases, which confused
2797 the lustre dentry revalidation. Change it to always use the
2801 Frequency : rare, unless heavy write-truncate concurrency is continuous
2802 Bugzilla : 4180, 6984, 7171, 9963, 9331
2803 Description: OST becomes very slow and/or deadlocked during object unlink
2804 Details : filter_destroy() was holding onto the parent directory lock
2805 while truncating+unlinking objects. For very large objects this
2806 may block other threads for a long time and slow overall OST
2807 responsiveness. It may also be possible to get a lock ordering
2808 deadlock in this case, or run out of journal credits because of
2809 the combined truncate+unlink. Solution is to do object truncate
2810 first in one transaction without parent lock, and then do the
2811 final unlink in a new transaction with the parent lock. This
2812 reduces the lock hold time dramatically.
2815 Frequency : rare, 2.4 kernels only
2817 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2818 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2819 before unmounting their filesystem in order to avoid deadlock.
2820 The kernel_locked() function in 2.4 kernels only checks whether
2821 the kernel lock is held, not whether it is this process that is
2822 holding it as 2.6 kernels do.
2827 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2828 Details : The client ptlrpc code may be trying to reconnect to a down
2829 server before a previous connection attempt has timed out.
2830 Increase the reconnect interval to be longer than the connection
2831 timeout interval to avoid sending duplicate connections to
2835 Frequency : echo_client brw_test command
2837 Description: fix echo_client to work with OST preallocated code
2838 Details : OST preallocation code (5137) didn't take echo_client IO path
2839 into account: echo_client calls filter methods outside of any
2840 OST thread and, hence, there is no per-thread preallocated
2841 pages and buffers to use. Solution: hijack pga pages for IO. As
2842 a byproduct, this avoids unnecessary data copying.
2846 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2847 Description: Client can oops in mdc_commit_close() after open replay
2848 Details : It was possible for the MDS to return an open request with no
2849 transaction number in mds_finish_transno() if the client was
2850 evicted, but without actually returning an error. Clients
2851 would later try to replay that open and may trip an assertion
2852 Simplify the client close codepath, and always return an error
2853 from the MDS in case the open is not successful.
2856 Frequency : rare, 2.6 OSTs only
2858 Description: OST may deadlock under high load on fragmented files
2859 Details : If there was a heavy load and highly-fragmented OST filesystems
2860 it was possible to have all the OST threads deadlock waiting on
2861 allocation of biovecs, because the biovecs were not released
2862 until the entire RPC IO was completed. Instead, release biovecs
2863 as soon as they are complete to ensure forward IO progress.
2865 Severity : enhancement
2867 Description: Support for specifying external journal device at mount
2868 Details : If an OST or MDS device is formatted with an external journal
2869 device, this device major/minor is stored in the ext3 superblock
2870 and may not be valid for failover. Allow detecting and
2871 specifying the external journal at mount time.
2876 Description: Mounting an MDS with pending unlinked files may cause oops
2877 Details : target_finish_recovery() calls mds_postrecov() which returned
2878 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2879 considers this an error and immediately begins cleaning up the
2880 lov, just after starting the mds_lov process
2882 Severity : enhancement
2884 Description: Implement 'lfs df' to report actual free space on per-OST basis
2885 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2886 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2887 usage of objects; '-h' to report in human readable format.
2889 ------------------------------------------------------------------------------
2891 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2898 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2899 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2900 at startup during orphan unlinking. Instead, skip buddy update
2901 before setup, as it will be regenerated after recovery is complete.
2906 Description: If an OST is inactive, its locks might reference stale inodes.
2907 Details : lov_change_cbdata() must iterate over all namespaces, even if
2908 they are inactive to clear inode references from the lock.
2910 Severity : enhancement
2911 Frequency : occasional, if non-standard max_dirty_mb used
2913 Description: Client will block write RPCs if not enough grant
2914 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2915 then the client will block writes while waiting for another RPC
2916 to complete instead of consuming its dirty limit. With change
2917 we get improved performance when max_dirty_mb is small.
2919 Severity : enhancement
2920 Bugzilla : 3389, 6253
2921 Description: Add support for supplementary groups on the MDS.
2922 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2923 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2924 lookups for user supplementary groups into a cache.
2928 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2929 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2930 for liblustre O_EXCL is enforced by MDS.
2934 Description: readdir never returns NULL in liblustre.
2935 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2936 dentry in struct dirent.
2940 Frequency : liblustre only, depends on application IO pattern
2941 Description: liblustre clients evicted if not contacting servers
2942 Details : Don't put liblustre clients into the ping_evictor list, so
2943 they will not be evicted by the pinger ever.
2945 Severity : enhancement
2947 Description: Add ability to evict clients by NID from MDS.
2948 Details : By echoing "nid:$NID" string into
2949 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2950 $NID would be instantly evicted from this MDS and from all active
2951 OSTs connected to it.
2955 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2956 Details : lookup_it_finish() used to query file size from OSTs that was not
2961 Description: service threads change working directory to that of init
2962 Details : Starting lustre service threads may pin the working directory
2963 of the parent thread, making that filesystem busy. Threads
2964 now change to the working directory of init to avoid this.
2968 Frequency : during shutdown only
2969 Description: shutdown with a failed MDS or OST can cause unmount to hang
2970 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2975 Frequency : 2.6 only
2976 Description: chmod/chown may include an extra supplementary group
2977 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2978 supplementary group and if none is specified this is used.
2981 Bugzilla : 5479 (6816)
2983 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2984 Details : If lookup is in progress on a file that is unlinked we might try
2985 to revalidate the inode and fail in revalidate after lookup is
2986 complete and ll_file_open() enqueues the open again but
2987 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2990 Frequency : always, if lconf --abort_recovery used
2992 Description: lconf --abort_recovery fails with 'Operation not supported'
2993 Details : lconf was attempting to abort recovery on the MDT device and not
2996 ------------------------------------------------------------------------------
2998 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3003 Frequency : rare (only unsupported configurations with a node running as an
3005 Bugzilla : 6514, 5137
3006 Description: Mounting a Lustre file system on a node running as an OST could
3008 Details : OSTs now preallocates memory needed to write out data at
3009 startup, instead of when needed, to avoid having to
3010 allocate memory in possibly low memory situations.
3011 Specifically, if the file system is mounted on on OST,
3012 memory pressure could force it to try to write out data,
3013 which it needed to allocate memory to do. Due to the low
3014 memory, it would be unable to do so and the node would
3015 become unresponsive.
3017 Severity : enhancement
3019 Description: Addition of lconf --service command line option
3020 Details : lconf now accepts a '--service <arg>' option, which is
3021 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3023 Severity : enhancement
3025 Description: Failover mode is now the default for OSTs.
3026 Details : By default, OSTs will now run in failover mode. To return to
3027 the old behaviour, add '--failout' to the lmc line for OSTs.
3029 Severity : enhancement
3031 Description: Health checks are now provided for MDS and OSTs
3032 Details : Additional detailed health check information on MSD and OSTs
3033 is now provided through the procfs health_check value.
3036 Frequency : occasional, depends on IO load
3038 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3039 after numerous create/delete cycles
3040 Details : The ext3 inode allocation policy would not allocate new inodes
3041 very well on the OSTs because there are no new directories
3042 being created. Instead we look for groups with free space if
3043 the parent directories are nearly full.
3048 Description: Network or server problems during mount may cause partially
3049 mounted clients instead of returning an error.
3050 Details : The config llog parsing code may overwrite the error return
3051 code during mount error handling, returning success instead
3057 Description: MDS can fail to allocate large reply buffers
3058 Details : After long uptimes the MDS can fail to allocate large reply
3059 buffers (e.g. zconf client mount config records) due to memory
3060 fragmentation or consumption by the buffer cache. Preallocate
3061 some large reply buffers so that these replies can be sent even
3062 under memory pressure.
3066 Frequency : rare (liblustre)
3067 Description: fsx running with liblustre complained that using truncate() to
3068 extend the file doesn't work. This patch corrects that issue.
3069 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3070 ATTR_SIZE and lsm use in llu_setattr_raw.
3074 Frequency : rare, only 2.6 kernels
3075 Description: Unusual file access patterns on the MDS may result in inode
3076 data being lost in very rare circumstances.
3077 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3078 code in the 2.6 kernel caused the inode and/or EA data not to
3079 be read from disk, causing single-file corruption.
3083 Frequency : rare, only 2.6 filesystems using extents
3084 Description: Heavy concurrent write and delete load may cause data corruption.
3085 Details : It was possible under high-load situations to have an extent
3086 metadata block in the block device cache from a just-unlinked
3087 file overwrite a newly-allocated data block. We now unmap any
3088 metadata buffers that alias just-allocated data blocks.
3092 Frequency : filesystems with default stripe_count larger than 77
3093 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3094 Details : lconf specifies an inode size of 4096 bytes when the default
3095 stripe_count is larger than 77. This conflicts with the default
3096 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3097 this case to avoid pinning too much memory for large EAs.
3099 ------------------------------------------------------------------------------
3101 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3106 Frequency : rare (extremely heavy IO load with hundreds of clients)
3108 Description: Client is evicted, gets IO error writing to file
3109 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3110 caused clients to be evicted for AST timeouts. The fixes in
3111 bug 5192 mean we no longer need to have such short AST timeouts
3112 so ldlm_timeout has been increased.
3115 Frequency : occasional during --force or --failover shutdown under load
3116 Bugzilla : 5949, 4834
3117 Description: Server oops/LBUG if stopped with --force or --failover under load
3118 Details : a collection of import/export refcount and cleanup ordering
3119 issues fixed for safer force cleanup
3122 Frequency : only filesystems larger than 120 OSTs
3123 Bugzilla : 5990, 6223
3124 Description: lfs getstripe would oops on a very large filesystem
3125 Details : lov_getconfig used kfree on vmalloc'd memory
3128 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3129 Bugzilla : 6242, 6243
3130 Description: reading from files that had been truncated to a non-zero size
3131 but never opened returned no data
3132 Details : ll_file_read() reads zeros from no-object files to EOF
3137 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3138 Details : There's little harm in aborting MDS/OSS recovery and letting it
3139 try again, so I removed the LASSERT and return an error instead.
3141 Severity : enhancement
3143 Description: New debugging infrastructure for tracking down data corruption
3144 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3145 (b) cover more of the client-side code path, and (c) try to narrow
3146 down where problems occurred
3150 Bugzilla : 3819, 4364, 4397, 6313
3151 Description: Racing close and eviction MDS could cause assertion in mds_close
3152 Details : It was possible to get multiple mfd references during close and
3153 client eviction, leading to one thread referencing a freed mfd.
3155 Severity: : enhancement
3156 Bugzilla : 3262, 6359
3157 Description: Attempts to reconnect to servers are now more aggressive.
3158 Details : This builds on the enhanced upcall-less recovery that was added
3159 in 1.4.2. When trying to reconnect to servers, clients will
3160 now try each server in the failover group every 10 seconds. By
3161 default, clients would previously try one server every 25 seconds.
3166 Description: After recovery, certain operations trigger a failed
3167 assertion on a client.
3168 Details : Failing over an mds, using lconf -d --failover, while a
3169 client was doing a readdir() call would cause the client to
3170 LBUG after recovery completed and the readdir() was resent.
3172 Severity : enhancement
3174 Description: Default groups are now added by lconf
3175 Details : You can now run lconf --group <servicename> without having to
3176 manually add groups with lmc.
3179 Frequency : occasional
3181 Description: Nodes with an elan id of 0 trigger a failed assertion
3184 Frequency : always when accessing e.g. tty/console device nodes
3186 Description: tty and some other devices nodes cannot be used on lustre
3187 Details : file's private_data field is used by device data and lustre
3188 values in there got lost. New field was added to struct file to
3189 store fs-specific private data.
3192 Frequency : when exporting Lustre via NFS
3194 Description: NFSD failed occasionally when looking up a path component
3195 Details : NFSD is looking up ".." which was broken in ext3 directories
3196 that had grown large enough to become hashed.
3199 Frequency : Clusters with multiple interfaces not on the same subnet
3201 Description: Nodes will repeatedly try to reconnect to an interface which it
3202 cannot reach and report an error to the log.
3203 Details : Extra peer list entries will be created by lconf with some peers
3204 unreachable. lconf now validates the peer before adding it.
3207 Frequency : Only if a default stripe is set on the filesystem root.
3209 Description: Setting a default stripe on the filesystem root prevented the
3210 filesystem from being remounted.
3211 Details : The client was sending extra request flags in the root getattr
3212 request and did not allocate a reply buffer for the dir EA.
3215 Frequency : occasional, higher if lots of files are accessed by one client
3216 Bugzilla : 6159, 6097
3217 Description: Client trips assertion regarding lsm mismatch/magic
3218 Details : While revalidating inodes the VFS looks up inodes with ifind()
3219 and in rare cases can find an inode that is being freed.
3220 The ll_test_inode() code will free the lsm during ifind()
3221 when it finds an existing inode and then the VFS later attaches
3222 this free lsm to a new inode.
3226 Bugzilla : 6422, 7030
3227 Description: MDS deadlock between mkdir and client eviction
3228 Details : Creating a new file via mkdir or mknod (starting a transaction
3229 and getting the ns lock) can deadlock with client eviction
3230 (gets ns lock and trying to finish a synchronous transaction).
3233 Frequency : occasional
3234 Description: While starting a server, the fsfilt_ext3 module could not be
3236 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3237 kernels. Previously, lconf would still use the ext3 name
3238 when trying to load modules. Now, it will correctly use
3239 ext3 on 2.4 and ldiskfs on 2.6.
3241 Severity : enhancement
3242 Description: The default stripe count has been changed to 1
3243 Details : The interpretation of the default stripe count (0, to lfs
3244 or lmc) has been changed to mean striping across a single
3245 OST, rather than all available. For general usage we have
3246 found a stripe count of 1 or 2 works best.
3248 Severity : enhancement
3249 Description: Add support for compiling against Cray portals.
3250 Details : Conditional compiling for some areas that are different
3254 Frequency : occasional
3255 Bugzilla : 6409, 6834
3256 Description: Creating files with an explicit stripe count may lead to
3257 a failed assertion on the MDS
3258 Details : If some OSTs are full or unavailable, creating files may
3259 trigger a failed assertion on the MDS. Now, Lustre will
3260 try to use other servers or return an error to the
3264 Frequency : occasional
3266 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3267 caused lock timeout during readahead (since 1.4.2).
3268 Details : Processes doing readahead might match a lock that hasn't been
3269 granted yet if there are overlapping and conflicting lock
3270 requests. The readahead process waits on ungranted lock
3271 (original lock is CBPENDING), while OST waits for that process
3272 to cancel CBPENDING read lock and eventually evicts client.
3274 Severity : enhancement
3276 Description: Initial enabling of flock support for clients
3277 Details : Implements fcntl advisory locking and file status functions.
3278 This feature is provided as an optional mount flag (default
3279 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3280 locking are implemented yet, and those that are are not guaranteed
3281 to be completely correct in production environments.
3282 mount -t lustre -o [flock|noflock] ...
3285 Frequency : occasional
3287 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3288 assertion in the ext3 JBD (journaling) layer.
3289 Details : The b_committed_data struct is protected by the big kernel lock
3290 in 2.4 kernels, serializing journal_commit_transaction() and
3291 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3292 access to this struct. In 2.6 kernels there is finer grained
3293 locking to improve SMP performance of the JBD layer.
3297 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3302 Description: merge and cleanup kernel patches.
3305 -----------------------------------------------------------------------------
3307 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3309 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3310 versions of Lustre. Please update all servers and clients to
3311 version 1.4.2 or later at the same time. You must also run
3312 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3313 to update the configuration logs.
3315 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3316 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3317 - swab llog records of type '0' so we get proper header size/idx (5861)
3318 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3319 - fix rename of one directory over another leaking an inode (5953)
3320 - avoid SetPageDirty on 2.6 (5981)
3321 - don't re-add just-being-destroyed locks to the waiting list (5653)
3322 - when creating new directories, inherit the parent's custom
3323 striping settings if present parent (3048)
3324 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3325 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3326 - don't hold client locks on temporary worklist from l_lru (5666)
3327 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3328 - automatically evict dead clients (5921)
3329 - Update file size properly in create+truncate+fstat case (6196)
3330 - Do not unhash mountpoint dentries, do not allow removal of
3332 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3333 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3334 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3335 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3336 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3337 - Correctly handle reads of files with no objects (6243)
3338 - lctl recover will also mark a device active if deactivate used (5933)
3340 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3341 - allow --write-conf on an MDS with different nettype than client (5619)
3342 - don't write config llogs to MDS for mounts not from that MDS (5617)
3343 - lconf should create multiple TCP connections from a client (5201)
3344 - init scripts are now turned off by default; run chkconfig --on
3345 lustre and chkconfig --on lustrefs to use them
3346 - upcalls are no longer needed for clients to recover to failover
3348 - add --abort-recovery option to lconf to abort recovery on device
3350 - add support for an arbitrary number of OSTs (3026)
3351 - Quota support protocol changes.
3352 - forward compatibility changes to wire structs (6007)
3353 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3354 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3355 - improved "lustre status" script
3356 - initialize blocksize for non-regular files (6062)
3357 - added --disable-server and --disable-client configure options (5782)
3358 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3359 - Vanilla 2.4.29 support
3360 - increase maximum number of obd devices to 520 (6242)
3361 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3362 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3363 SLES 9 2.6.5 kernels
3364 - limit stripes per file to 160 (the maximum EA size) (6093)
3366 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3369 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3370 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3371 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3372 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3373 - small ext3 extents cleanups and fixes (5733)
3374 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3375 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3376 - avoid needless client->OST connect, fix handle mismatch (5317)
3377 - fix DLM error path that led to out-of-sync client, long delays (5779)
3378 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3379 - fix several locking issues related to i_size (5492,5624,5654,5672)
3380 - don't move pending lock onto export if it is already evicted (5683)
3381 - fix kernel oops when creating .foo in unlinked directory (5548)
3382 - fix deadlock in obdfilter statistics vs. object create (5811)
3383 - use time_{before,after} to avoid timer jiffies wrap (5882)
3384 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3385 - Do not leak request if server was not able to process it (5154)
3386 - If mds_open unable to find parent dir, make that negative lookup(5154)
3387 - don't create new directories with extent-mapping (5909, 5936)
3389 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3390 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3391 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3392 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3393 - support for RHEL4 kernel on i686 (5773)
3394 - provide error messages when incompatible logs are encountered (5898)
3396 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3397 * version 1.4.0.10 (1.4.1 release candidate 1)
3399 - don't keep a lock reference when lock is not granted (4238)
3400 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3401 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3402 - drop import inflight refcount on signal_completed_replay error (5255)
3403 - unlock page after async write error during send (3677)
3404 - handle missing objects in filter_preprw_read properly (5265)
3405 - no transno return for symlink open, don't save no-trasno open (3440)
3406 - don't try to complete elan receive that already failed (4012)
3407 - free RPC server reply state on error (5406)
3408 - clean up thread from ptlrpc_start_thread() on error (5160)
3409 - readahead could read extra page into cache that wasn't ejected (5388)
3410 - prevent races in class_attach/setup/cleanup/detach (5260)
3411 - don't dereference de->d_inode after l_dput of de (5458)
3412 - use "int" for stripe value returned from lock_to_stripe (5544)
3413 - mballoc allocation and error-checking fixes in 2.6 (5504)
3414 - block device patches to fix I/O request sizes in 2.6 (5482)
3415 - look up hostnames for IB nals (5602)
3416 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3417 - don't start multiple acceptors for the same port (5277)
3418 - fix incorrect LASSERT in mds_getattr_name (5635)
3419 - export a proc file for general "ping" checking (5628)
3420 - fix "lfs check" to not block when the MDS is down (5628)
3422 - service request history (4965)
3423 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3424 - create an "evict_client" /proc entry on OSTs, like the MDS has
3425 - fix mount usage message, return errors per mount(8) (5168)
3426 - change grep [] to grep "[]" in tests so they work in more UMLs
3427 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3428 - remove ancient LOV_MAGIC_V0 EA support (5047)
3429 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3430 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3431 - allow readahead allocations to fail when low on memory (5383)
3432 - mmap locking landed again, after considerable improvement (2828)
3433 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3435 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3438 - send OST transaction number in read/write reply to free req (4966)
3439 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3440 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3441 - the watchdog thread now runs as interruptible (5246)
3442 - flock/lockf fixes (but it's still disabled, pending 5135)
3443 - don't use EXT3 constants in llite code (5094)
3444 - memory shortage at startup could cause assertion (5176)
3446 - reorganization of lov code
3447 - single portals codebase
3449 - add extents/mballoc support (5025)
3450 - direct I/O reads in the obdfilter (4048)
3451 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3453 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3456 - send OST transaction number in read/write reply to free req (4966)
3457 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3458 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3459 - the watchdog thread now runs as interruptible (5246)
3460 - handle missing objects in filter_preprw_read properly (5265)
3461 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3462 - drop import inflight refcount on signal_completed_replay error (5255)
3463 - unlock page after async write error during send (3677)
3464 - return original error code on reconstructed replies (3761)
3465 - no transno return for symlink open, don't save no-trasno open (3440)
3467 - add pid to ldlm debugging output (4922)
3468 - bump the watchdog timeouts -- we can't handle 30sec yet
3469 - extra debugging for orphan dentry/inode bug (5259)
3471 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3474 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3475 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3476 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3477 - fix NULL dereference in /proc/sys/portals/routes (4827)
3478 - allow failed mdc_close() operations to be interrupted (4561)
3479 - stop precreate on OST before MDS would time out on it (4778)
3480 - don't send partial-page writes before EOF from client (4410)
3481 - discard client grant for sub-page writes on large-page clients (4520)
3482 - don't free dentries not owned by NFS code, check generation (4806)
3483 - fix lsm leak if mds_create_objects() fails (4801)
3484 - limit debug_daemon file size, always print CERROR messages (4789)
3485 - use transno after validating reply (3892)
3486 - process timed out requests if import state changes (3754)
3487 - update mtime on OST during writes, return in glimpse (4829)
3488 - add mkfsoptions to LDAP (4679)
3489 - use ->max_readahead method instead of zapping global ra (5039)
3490 - don't interrupt __l_wait_event() during strace
3492 - add software watchdogs to catch hung threads quickly (4941)
3493 - make lustrefs init script start after nfs is mounted
3494 - fix CWARN/ERROR duplication (4930)
3495 - return async write errors to application if possible (2248)
3496 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3497 - print NAL number in %x format (4645)
3498 - update barely-supported suse-2.4.21-171 series (4842)
3499 - support for sles 9 %post scripts
3500 - support for building 2.6 kernel-source packages
3501 - support for sles km_* packages
3503 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3506 - ignore -ENOENT errors in osc_destroy (3639)
3507 - notify osc create thread that OSC is being cleaned up (4600)
3508 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3509 - reconstruct ost_handle() like mds_handle() (4657)
3510 - create a new thread to do import eviction to avoid deadlock (3969)
3511 - let lconf resolve symlinked-to devices (4629)
3512 - don't unlink "objects" from directory with default EA (4554)
3513 - hold socknal file ref over connect in case target is down (4394)
3514 - allow more than 32000 subdirectories in a single directory (3244)
3515 - fix blocks count for O_DIRECT writes (3751)
3516 - OST returns ENOSPC from object create when no space left (4539)
3517 - don't send truncate RPC if file size isn't changing (4410)
3518 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3519 - bind to privileged port in socknal and tcpnal (3689)
3521 - rate limit CERROR/CWARN console message to avoid overload (4519)
3522 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3523 - basic mmap support (3918)
3524 - kernel patch series update from b1_4 (4711)
3526 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3529 - avoid crash during MDS cleanup with OST shut down (2775)
3530 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3531 - don't use bad inodes on the MDS (3744)
3532 - dynamic object preallocation to improve recovery speed (4236)
3533 - don't hold spinlock over lock dumping or change debug flags (4401)
3534 - don't zero obd_dev when it is force cleaned (3651)
3535 - print grants to console if they go negative (4431)
3536 - "lctl deactivate" will stop automatic recovery attempts (3406)
3537 - look for existing locks in ldlm_handle_enqueue() (3764)
3538 - don't resolve lock handle twice in recovery avoiding race (4401)
3539 - revalidate should check working dir is a directory (4134)
3541 - don't always mark "slow" obdfilter messages as errors (4418)
3543 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3546 - don't close LustreDB during write_conf until it is done (3860)
3547 - fix typo in lconf for_each_profile (3821)
3548 - allow dumping logs from multiple threads at one time (3820)
3549 - don't allow multiple threads in OSC recovery (3812)
3550 - fix debug_size parameters (3864)
3551 - fix mds_postrecov to initialize import for llog ctxt (3121)
3552 - replace config semaphore with spinlock (3306)
3553 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3554 - don't allow enqueue to complete on a destroyed export (3822)
3555 - down write_lock before checking llog header bitmap (3825)
3556 - recover from lock replay timeout (3764)
3557 - up llog sem before sending rpc (3652)
3558 - reduce ns lock hold times when setting kms (3267)
3559 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3560 - fix NULL deref and obd_dev leak on setup error (3312)
3561 - replace some LBUG about llog ops with error handling (3841)
3562 - don't match INVALID dentries from d_lookup and spin (3784)
3563 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3564 - fix invalid assertion in ptlrpc_set_wait (3880)
3566 - add libwrap support for the TCP acceptor (3996)
3567 - add /proc/sys/portals/routes for non-root route listing (3994)
3568 - allow setting MDS UUID in .xml (2580)
3569 - print the stack of a process that LBUGs (4228)
3571 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3574 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3575 - make sure to unset replay flag from failed open requests (3440)
3576 - if default stripe count is 0, use OST count for inode size (3636)
3577 - update parent mtime/ctime on client for create/unlink (2611)
3578 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3579 - free recovery state on server during a forced cleanup (3571)
3580 - unregister_reply for resent reqs (3063)
3581 - loop back devices mounting and status check on 2.6 (3563)
3582 - fix resource-creation race that can provoke i_size == 0 (3513)
3583 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3584 - more debugging for page-accounting assertion (3746)
3585 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3586 - avoid deadlock after precreation failure (3758)
3587 - fix race and lock order deadlock in orphan handling (3450, 3750)
3588 - add validity checks when grabbing inodes from l_ast_data (3599)
3590 - add /proc/.../recovery_status to obdfilter (3428)
3591 - lightweight CDEBUG infrastructure, debug daemon (3668)
3592 - change default OSC RPC parameters to be better on small clusters
3593 - turn off OST read cache for files smaller than 32MB
3594 - install man pages and include them in rpms (3100)
3595 - add new init script for (un)mounting lustre filesystems (2593)
3596 - run chkconfig in %post for init scripts (3701)
3597 - drop scimac NAL (unmaintained)
3599 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3602 - clean kiobufs before and after use (3485)
3603 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3604 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3605 - change init script to fail more clearly if not run as root (1528)
3606 - allow clients to reconnect during replay (1742)
3607 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3608 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3609 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3610 - fix cond_resched (3554)
3611 - only evict unfinished clients after recovery (3515)
3612 - allow bulk resend, prevent data loss (3570)
3613 - dynamic ptlrpc request buffer allocation (2102)
3614 - don't allow unlinking open directory if it isn't empty (2904)
3615 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3616 - remove extraneous obd dereference causing LASSERT failure (3334)
3617 - don't use get_cycles() when creating temp. files on the mds (3156)
3618 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3619 - handle EEXIST for set-stripe, set proper directory name (3336)
3621 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3622 - fix ksocknal_fmb_callback() error messages (2918)
3624 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3627 - don't copy lvb into (possibly NULL) reply on error (2983)
3628 - don't deref dentry after dput, don't free lvb on error (2922)
3629 - use the kms to determine writeback rpc length (2947)
3630 - increment oti_logcookies when osc is inactive (2948)
3631 - update client's i_blocks count via lvb messages (2543)
3632 - handle intent open/close of special files properly (1557)
3633 - mount MDS with errors=remount-ro, like obdfilter (2009)
3634 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3635 - don't use cancelling-locks' kms values (2947)
3636 - use highest lock extent for kms, not last one (2925)
3637 - don't dereference ERR_PTR() dentry in error handling path (3107)
3638 - fix thread race in portals_debug_dumplog() (3122)
3639 - create lprocfs device entries at setup instead of at attach (1519)
3640 - common AST error handler, don't evict client on completion race (3145)
3641 - zero nameidata in detach_mnt in 2.6 (3118)
3642 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3643 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3644 - initialize RPC timeout timer earlier for 2.6 (3219)
3645 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3646 - print nal/nid for unknown nid (3258)
3647 - additional checks for oscc recovery before doing precreate (3284)
3648 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3649 - don't crash in mdc_close for bad permissions on open (3285)
3650 - zero i_rdev for non-device files (3147)
3651 - clear page->private before handing to FS, better assertion (3119)
3652 - tune the read pipeline (3236)
3653 - fix incorrect decref of invalidated dentry (2350)
3654 - provide read-ahead stats and refine rpc in flight stats (3328)
3655 - don't hold journal transaction open across create RPC (3313)
3656 - update atime on MDS at close time (3265)
3657 - close LDAP connection when recovering to avoid server load (3315)
3658 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3659 - don't leak open file on MDS after open resend (3325)
3660 - serialize filter_precreate and filter_destroy_precreated (3329)
3661 - loop device shouldn't call sync_dev() for nul device (3092)
3662 - clear page cache after eviction (2766)
3663 - resynchronize MDS->OST in background (2824)
3664 - refuse to mount the same filesystem twice on same mountpoint (3394)
3665 - allow llmount to create routes for mounting behind routers (3320)
3666 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3667 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3668 - fix rare problem with rename on htree directories (3417)
3670 - allow default OST striping configuration per directory (1414)
3671 - fix compilation for qswnal for 2.6 kernels (3125)
3672 - increase maximum number of MDS request buffers for large systems
3673 - change liblustreapi to be useful for external progs like lfsck (3098)
3674 - increase local configuration timeout for slow disks (3353)
3675 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3677 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3680 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3681 - don't overwrite extent policy data in reply if lock was blocked (2901)
3682 - drop filter export grants atomically with removal from device (2663)
3683 - del obd_self_export from work_list in class_disconnect_exports (2908)
3684 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3685 - swab reply message in mdc_close, other PPC fixes (2464)
3686 - fix destroying of named logs (2325)
3687 - overwrite old logs when running lconf --write_conf (2264)
3688 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3689 - fix race in target_handle_connect (2898)
3690 - mds_reint_create() should take same inode create lock (2926)
3691 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3692 - don't close files for self_export to avoid uninitialized obd (2936)
3693 - allow MDS with the same name as client node (2939)
3694 - hold dentry reference for closed log files for unlink (2325)
3695 - reserve space for all logs during transactions (2059)
3696 - don't evict page beyond end of stripe extent (2925)
3697 - don't oops on a deleted current working directory (2399)
3698 - handle hard links to targets without a parent properly (2517)
3699 - don't dereference NULL lock when racing during eviction (2867)
3700 - don't grow lock extents when lots of conflicting locks (2919)
3702 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3705 - account for cache space usage on clients to avoid data loss (974)
3706 - lfsck support in lustre kernel code (2349)
3707 - reduce journal credits needed for BRW writes (2370)
3708 - orphan handling to avoid losing space on client/server crashes
3709 - ptlrpcd can be blocked, stopping ALL progress (2477)
3710 - use lock value blocks to assist in proper KMS, faster stat (1021)
3711 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3712 - recovery for initial connections (2355)
3713 - fixes for mds_cleanup_orphans (1934)
3714 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3715 - block all file creations until orphan recovery completes (1901)
3716 - client remove rq_connection from request struct (2423)
3717 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3718 - recovery timer race (2670)
3719 - mdc_close recovey bug (2532)
3720 - ptlrpc cleanup bug (2710)
3721 - mds timeout on local locks (2588)
3722 - namespace lock held during RPCs (2431)
3723 - handle interrupted sync write properly (2503)
3724 - don't try to handle a message that hasn't been replied to (2699)
3725 - client assert failure during cleanup after abort recovery (2701)
3726 - leak mdc device after failed mount (2712)
3727 - ptlrpc_check_set allows timedout requests to complete (2714)
3728 - wait for inflight reqs when ptlrpcd finishes (2710)
3729 - make sure unregistered services are removed from the srv_list
3730 - reset bulk XID's when resending them (caught by 1138 test)
3731 - unregister_bulk after timeout
3732 - fix lconf error (2694)
3733 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3734 - readahead locks pages, leaves pending causing memory pressure (2673)
3735 - increase OST request buffers to 4096 on large machines (2729)
3736 - fix up permission of existing directories in simple_mkdir (2661)
3737 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3738 - don't assign transno to errored transactions (2742)
3739 - don't delete objects on OST if given a bogus objid from MDS (2751)
3740 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3741 - if rq_no_resend, then timeout request after recovery (2432)
3742 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3743 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3744 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3745 - MDS recovery shouldn't delete objects that it has given out (2730)
3746 - if enqueue arrives after completion, don't clobber LVB (2819)
3747 - don't unlock pages twice when trigger_group_io returns error (2814)
3748 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3749 - don't write pages to disk if there was an error (1450)
3750 - don't ping imports that have recovery disabled (2676)
3751 - take buffered bytes into account when balancing socknal conn (2817)
3752 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3753 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3754 - remove little-endian swabbing of llog records (1987)
3755 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3756 - retry reposting request buffers if they fail (1191)
3757 - grow extent at grant time to avoid granting a revoked lock (2809)
3758 - lock revoke doesn't evict page if covered by a second lock (2765)
3759 - disable VM readahead to avoid reading outside lock extents (2805)
3761 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3762 - updated kernel patches for hp-2.4.20 kernel (2681)
3764 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3767 - fix truncated write corruption (2366)
3768 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3770 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3771 - ensure len(uuid) < 37 in lmc (1171)
3772 - fix ia64 OOPS in llog_test (2255)
3773 - zero end of page at obdfilter for partial page writes (2648)
3774 - don't leave stale dentries around after renames (bug 2428)
3775 - fix timeouts when evicting a client with a single lock held (2642)
3776 - set deadline for the initial HELLO message to drain (2634)
3777 - print out dotted-quad IP addresses in the socknal (2302)
3779 - additional debugging for MDS client eviction problem (2443)
3780 - fix mkfsoptions support for osts (2603, 2604)
3782 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3785 - add series for the vanilla 2.6.0 kernel
3786 - add series for the vanilla 2.4.24 kernel
3787 - add series for a cray x86/64 UL kernel drop
3788 - fix xattr patches for the vanilla 2.4.19 series
3790 - generate true UUIDs in lmc (1171)
3791 - have portals stack dumping break in UML (2466)
3792 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3793 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3794 - wake callback waiting threads on client eviction (2460)
3795 - Add --ptldebug and --subsystem to lmc (1719)
3796 - update assertion to allow safe interrupt allocation
3797 - set rq_no_resend for cancel requests (2432)
3798 - recalculate ptlrpcd timeout after resend (2494)
3799 - call vfs_rmdir when removing pending directories (2368)
3800 - fix renaming a file to itself (2429)
3801 - lmc creates a default one-stripe lov (2454)
3802 - expand procfs space to handle large clusters (2326)
3803 - increase UML stack to avoid overflow
3804 - update lconf's list of debug and subsystem masks
3805 - fix lfs find --obd (2510)
3806 - /proc tunable for disabling filter read caching (2591)
3807 - stop rpm packages from altering slapd.conf (2301)
3808 - disable nagle in the socknal under 0conf (2578)
3809 - choose mds inode size based on stripe count (2572)
3810 - fix kernel-source rpm problems (2516)
3812 - add --disable-doc to avoid pdf generation (2421)
3813 - update documentation, tests, type-os, comments
3814 - avoid format warnings on ia64
3815 - remove the TOE NAL
3816 - tiny code cleanups by removing unused fields
3818 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3821 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3822 - remove the most blatant lies from BUILDING file (2371)
3823 - change default debug level to reasonable production setting
3824 - reduce client side cache size to reduce cache flush time
3825 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3826 - make TCP zerocopy and pinger support enabled by default (2476)
3827 - sync writes completed after process exits caused crashes (2319)
3828 - maintain correct mount count on the MDS (2356)
3829 - backout 1557, because 2316 wasn't really fixed
3830 - better file I/O statistics gathering in /proc
3831 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3832 - another kernel patch to fix zero-copy TCP function export
3833 - don't take duplicate lock when processing re-sent getattr (2420)
3834 - lctl uses obd_self_export instead of creating new conn (2353)
3835 - MDS/OST recovery case which requires object creation asserted (2425)
3836 - move lfs from /usr/sbin to /usr/bin in packages
3837 - fix race between mds_client_add and mds_client_free (2417)
3838 - use kmalloc instead of slabs in portals (2430)
3839 - don't create duplicate records when a failover MDS is present (2442)
3840 - remove unnecessary mount age check (2332)
3841 - don't remove directory inodes from locks prematurely (2451)
3842 - don't break if MDS service name is the same as hostname (2103)
3843 - fix races in client write RPC generation when cache full (2482)
3845 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3848 - remove now-unused request->rq_obd (278)
3849 - if an allocation fails, print out how much memory we've used (1933)
3850 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3851 - add the "configurable stack size" patch to most series files (1256)
3852 - ability to write large log records, for 100+ OST configs (2306)
3853 - fix NULL deref when filter_prep fails (2314)
3854 - fix operator precedence error in filter_sync
3855 - dynamic allocation of socknal TX descriptors (2315)
3856 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3857 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3858 - ensure that CWARN messages in Portals always get to the syslog
3859 - __init/__exit are not for prototype decls (ldlm_init/exit)
3860 - x86-64 compile warning fixes
3861 - fix gateway LMC keyword conflict (2318)
3862 - fix MDS lock inversions in getattr/reint paths (1844)
3863 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3864 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3865 - DLM race condition prevented some lock evictions (2328)
3866 - ENOMEM detection and retry on socknal sends (2230)
3867 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3868 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3869 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3870 - better error messages when a client is rejected during recovery (1505)
3871 - avoid cancelling locks which were never granted, after failure (2330)
3872 - fix i_sem/journal inversion in mds_client_add (2333)
3873 - fix truncate/getattr lock cycle deadlock (2334)
3874 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3875 - fix two rare exit paths which could leak an l_lock() ref (2321)
3876 - fencepost error in MDS/OST orphan recovery (2226)
3877 - make log record alignment 8 bytes (1988)
3878 - lstripe now fails when requested offset > ost_count (2237)
3879 - ensure that all kernel series have a complete list.h (1607)
3880 - fix crashes in special-file operations (2316)
3881 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3883 - allow configurable automake binary, for testing new versions
3884 - small update to the lfs documentation
3886 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3888 * fix negative export reference count in fsfilt_sync (2312)
3890 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3891 * release candidate 0.9.1
3893 - orphans are moved into the PENDING directory for possible recovery
3894 - replayed opens now open by fid for orphan/rename safety (1042)
3895 - last close of an orphan inode generates a transno (683)
3896 - chdir() and mount() now pin the directory entry (1020)
3897 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3898 - discard very old requests without processing them (1502)
3899 - remove some common, well-understood CERRORs (1505)
3900 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3901 - clear "grant" flags in OST replies until OST grant code lands (1644)
3902 - fix read performance by not clobbering i_blksize on client (1598)
3903 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3904 - make LRU size a /proc tunable, clears locks when reduced (707)
3905 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3906 - server locks take references on exports now (1558)
3907 - build fixes for 2.4.20-rh trees (1663)
3908 - return an error from lov_create if all OSCs are inactive (1751)
3909 - fix import levels when a reconnect happens without a timeout (1597)
3910 - exit early from mds_open if we get a lookup error (1749)
3911 - partial page read at EOF wouldn't wait for disk before sending (1642)
3912 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3913 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3914 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3915 - write/truncate lock inversion (1639)
3916 - Don't auto-load obdclass, portals modules during cleanup (1495)
3917 - fix timestamps from jumping to "now" (1763)
3918 - extra journal assertions (1648)
3919 - add an extra multiunlink test (1771)
3920 - fix read_record/write_record API (1776)
3921 - fix leak of offset_extent, possible incorrect i_size later (1772)
3922 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3923 - Debugging for the kqswnal_get_idle_tx problems (1820)
3924 - Allow recovery to be attempted multiple times (1536)
3925 - Write out MDS last_rcvd file after it is first created (1600)
3926 - Fix tx_descriptor leak in failed transmit situations (1827)
3927 - ext3 journaling fixes for assertion failure after IO error (1871)
3928 - class_export_put() on freed export after completion AST error (1896)
3929 - Fix revalidate looping in VFS (1322)
3930 - Don't access a freed export during MDS_REINT timeout (1521)
3931 - Add open-unlink recovery support on the MDS (1673,1764)
3932 - Return an error if no MDS data was read from last_rcvd (1946)
3933 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3934 - Don't setup a disk device that doesn't match exported UUID (317)
3935 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3936 - avoid committing NULL handle in force close
3937 - local.sh is now a one-stripe LOV configuration
3938 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3939 - fix ext3 htree duplicate directory entry corruption (1516)
3940 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3941 - update to new LOV EA format (2097)
3942 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3944 2003-06-15 Phil Schwan <phil@clusterfs.com>
3947 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3948 - per-import/export recovery handling (958, 931, 959)
3949 - multiple last-rcvd slots, for serving multiple FSes (949)
3950 - connections are again shared between multiple imp/exports (963, 964)
3951 - "umount -f" would hang if any requests needed to be sent (393, 978)
3952 - avoid pinning large req buffer by copying for queued messages (989)
3953 - add "uuid" to "lctl device" command to help upcalls (991)
3954 - "open" RPCs with transnos would confuse recovery counters (1037)
3955 - do proper endian conversion of all wire messages (288, 340, 891)
3956 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3957 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3958 - fix ptlrpc_connection leak in target_handle_connect (1174)
3959 - fix import refcounting bug in OST and MDS cleanup (1134)
3960 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3961 - fix very unlikely obd_types race condition (501)
3962 - remove osc_open hack for echo_client (1187)
3963 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3964 - a failure in read_inode2 leads to deadlock (1139)
3965 - cancel ack-locks as soon as transaction is committed (1072)
3966 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3967 - make sure to commitrw after any preprw to avoid deadlock (1162)
3968 - failing to execute a file in a lustre FS would lock inode (1203)
3969 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3970 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3971 - obd_uuid2tgt would walk off the end of the list (1255)
3972 - on IA64 the osc would give portals incorrect bulk size (1258)
3973 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3974 - fix lock inversion caused by new llite matching code (1282)
3975 - limit the number of dirty pages on a client to 10MB (1286)
3976 - timed out locks were not being corrected cancelled (1289)
3977 - fix O_DIRECT above 4GB on IA-32 (1292)
3978 * major user-visible changes
3979 - fail out/fail over policy now controlled by the upcall (993)
3981 - add OBD_PING to check server availability and failure (954)
3982 - lustre messages are now sent in sending host order (288, 340, 891)
3983 - add eadatalen to MDS getattr reply (340)
3984 - OST read replies may contain second buffer, with per-page status (593)
3986 2003-03-11 Phil Schwan <phil@clusterfs.com>
3989 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3990 - failed open()s could cause deadlock; fixed (867, 869)
3991 - stop cancelling OST locks when files are closed (481)
3992 - overlapping XID spaces caused network corruption (851, 853)
3993 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3994 - setattr_raw added, to do single-RPC, server-side setattrs
3995 - lmc/lconf syntax change for OST UUIDs
3996 - fix crashy race condition between ptlrpc_free_req and osc_close
3997 - don't use request in mdc_enqueue if we hit a timeout (889)
3998 - don't set the inode i_size for regular files from the MDS (896)
3999 - handle out of order completion AST (842)
4000 - don't LBUG if a lock request times out after receiving AST (913)
4001 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4002 - if a bad lock AST arrives, send an error instead of dropping entirely
4003 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4004 - fix leak in bulk IO when only partially completed (899, 900, 926)
4005 - fix O_DIRECT for ia64 (55)
4006 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4007 - C-z after timeout could hang a process forever; fixed (977)
4009 - client-side I/O cache (678, 924, 929, 941, 970)
4011 - READPAGE and SETATTRs which don't take server-side locks get
4014 2003-02-11 Phil Schwan <phil@clusterfs.com>
4017 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4018 - Fix fsfilt_extN_readpage() to read a full page of directory
4019 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4020 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4021 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4022 - Most intent operations were removed, in favour of a new RPC mode
4023 that does a single RPC to the server and bypasses most of the VFS
4024 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4025 - Aggressively cancel local locks on DLM servers
4026 - mds_reint_unlink sends EA to the client if it's the last nlink.
4027 client uses that EA to unlink OST objects.
4028 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4029 - recursive symlinks were fixed (439)
4030 - fixed NULL deref in DEBUG_REQ
4031 - filter_update_lastobjid no longer calls sync, which annoyed extN
4032 - fixed multi-client small-writes to a single file problem (445)
4033 - fixed mtime updates during file writes (607)
4034 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4035 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4036 - fixed Linux OST size reporting problem (444, 656)
4037 - OST now updates object mtime with writes or setattr (607, 619)
4038 - client verifies file size before zeroing page past EOF (445)
4039 - OST now writes last allocated objid to disk with allocation (108)
4040 - LOV on echo now works (409)
4042 - mds_reint_unlink sends a new buffer, with the EA included. this
4043 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4044 set if a regular file was being unlinked, and it was the last link
4045 - use PtlGet from the target for bulk writes (315)
4046 - OST now updates object mtime with writes or setattr (607, 619)
4047 - LDLM now has a grant-time callback to revalidate locked items, if
4049 - Many MDS operations were reorganized to combat race conditions
4051 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4052 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4054 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4057 - Fully reactivate OST imports after reconnection (512, others)
4058 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4059 - More graceful error handling for truncating on dead OST (515)
4060 - Don't error out unless we're actually accessing dead stripes (474)
4061 - Fix garbage sizes when stripes are missing (410)
4062 - LRU counters were broken, causing constant lock purge (433, 432)
4063 - garbage on read from stripes with failed OSTs (441)
4064 - mark OSCs as active before reconnecting during recovery (438)
4065 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4066 - lfind did not preserve OST order in output (443)
4067 - symlinks cause hung clients, incorrect data (439)
4068 - stop dereferencing request after dropping refcount (457)
4069 - don't LASSERT(spin_is_locked) on non-SMP (455)
4070 - fixes for many rename() bugs
4071 - fstat didn't correctly synchronize attributes (399)
4072 - server must handle lock cancellation during blocking AST prep (487)
4073 - bulk descriptors were free()d too soon (511)
4074 - fix paths in lconf, which would load incorrect modules (451, 507)
4075 - fix confusing lconf 'host not found' error message (386)
4076 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4077 - fix race condition in mdc_blocking_ast() for inode access (526)
4078 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4079 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4080 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4081 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4082 - fix lstripe in several ways (kernel side) (527)
4083 - fix request leak in ldlm_cli_enqueue (262)
4084 - incorrect OSC was marked inactive after OST failure
4085 - call mds_fs_cleanup before unmounting filesystem (524)
4086 - fix races between taking ns_lock and ldlm_lock_change_resource
4087 - fix races updating LOV export open file list
4088 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4089 - fix recovery NULL deref in ldlm_cli_cancel_unused
4090 - fix some DLM races by using new hash table for lock handles (419)
4091 - permit the client to specify desired inodes, at replay
4092 - duplicate requests when we queue them for replay reintegration
4093 - fix last_rcvd offset calculation
4094 - sync after each recovered transaction, so we always make progress
4095 - never, not always, ERESTART requests without transnos
4096 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4098 - skip replay if the MDS says that the client is already connected
4099 - don't check for a recovery-enabled export to match lctl's UUID
4100 - don't INC_USE_COUNT for phantom exports
4101 - don't crash when cleaning up phantom exports (567)
4102 - don't double-finish or set replay data for errored mdc_open requests
4103 - abort requests when they time out, so we don't get old replies
4104 - send/receive replies for AST messages again
4105 - if the client says that it doesn't have the lock, cancel it on the
4107 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4108 mark it as destroyed, it will all work out when decref is called
4109 - fix module use counts (22, 581)
4111 - ASTs now expect a reply (server cancels lock on error reply)
4113 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4116 - fix many simultaneous client startup (392)
4117 - fix dentry->d_it clobbering
4118 - credentials weren't being shipped for readdir/getattr operations
4119 - remove invalid assertions triggered during some concurrent MD
4121 - proper Lustre versions added (336, 389)
4122 - fix memory leak for create error case (398)
4123 - fix LOV locking bug that would get cli/srv out of sync
4124 - fix echo client over LOV (409)
4125 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4126 - fix double-O_EXCL intent crash (424)
4127 - avoid sending multiple lock CANCELs (352)
4129 - MDS can do multi-client recovery (modulo bugs in new code)
4131 - many updates, edits, cleanups
4133 2002-11-18 Phil Schwan <phil@clusterfs.com>
4136 - fix null d_it dereference (346)
4137 - fix full OST/dbench hang (333)
4138 - fix permission problem with file removal (286)
4139 - fix removal of OSCs from LOV when they fail
4140 - fix NULL deref during bulk timeout (214)
4141 - fix problems related to multiple filesystems on one MDS (241)
4142 - fixed serious subtle metadata locking bugs
4143 - free locks on clients when inodes are removed due to memory
4145 - fix inode pointer in lock data (285)
4146 - partial support for multiple MDS on a single host (241)
4147 - data locks weren't cancelled at clear_inode time (290, 311)
4148 - intent locks could lead to unbounded lock growth (205)
4149 - added a maximum lock count, an LRU list, and a flusher
4150 - fix multiple rename (365)
4151 - properly abstracted the echo client
4152 - OSC locked 1 byte too many; fixed
4153 - rewrote brw callback code:
4154 - fixed recovery bugs related to LOVs (306)
4155 - fixed too-many-pages-in-one-write crash (191)
4156 - fixed (again) crash in sync_io_timeout (214)
4157 - probably fixed callback-related race (385)
4159 - Add capability to MDS protocol
4160 - LDLM cancellations and callbacks on different portals
4162 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4165 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4166 - defer open object destruction to close time (601981, 138)
4167 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4168 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4169 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4170 - don't call obd_setattr in ll_file_release for destroyed objects
4171 * protocol change to lustre_msg: move |version| and add |flags|
4172 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4173 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4174 * added replay of create, unlink, link and rename operations during
4175 MDS failover; recovery should be much more robust now
4176 * remove failed OSCs from LOVs (only lov_create uses this so far)
4177 * the lustre-HOWTO was brought (more) up to date (582544)
4179 2002-10-23 Phil Schwan <phil@clusterfs.com>
4182 - in-use dentries weren't being reused properly (617851)
4183 - prevent multiple LDLM setup (599178)
4184 - fix LOV size calculations for truncate (617853)
4185 - fix client handling of MDS intent errors (POSIX)
4186 - fix permission bug in lovstripe.c test (624321)
4187 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4188 - truncate past end of file could corrupt data
4189 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4190 - a race in recovery could return ETIMEDOUT to apps (623947)
4191 - building outside the source directory was fixed
4192 * the lustre-HOWTO was brought (more) up to date (582544)
4193 * major progress was made on recovery functionality
4195 2002-10-10 Phil Schwan <phil@clusterfs.com>
4198 - recovery deadlock fix
4199 - rm -rf causes LBUG fix (617817)
4200 - file open by multiple tasks fix (618962)
4201 - directory permissions bugs (602707 and 620007)
4202 - journal_stop fixed with locking (611313)
4203 - O_APPEND failures resolved (618273, perhaps 614459)
4204 - lconf PATH fix (619770)
4205 - IA64 build fix (621450)
4206 - RPC buffer sizes scale with amount of memory
4208 2002-10-01 Phil Schwan <phil@clusterfs.com>
4211 - locks would be cancelled without throwing away data pages,
4212 resulting in inconsistent data (605627)
4213 - inode attributes were not always being refreshed (605627, 612449)
4214 - lconf now continues to cleanup after lctl reports an error
4215 - MDS now enforces user permissions (602707)
4216 - lprocfs cleanup fixed, but not yet enabled (614157)
4217 - fixed infinite server hang, should a client not respond to an AST
4218 - avoid going into recovery if user calls readlink() with a buffer
4219 that's too small (613941)
4220 - AST RPCs no longer require replies (614867) -- this may be changed
4221 - don't crash server if client sends an IOV that's too big (611336)
4222 - fixed lock conversion deadlock (611892)
4223 - fixed the following of symlinks (614622)
4224 * recovery: the server can remove locks from a client that dies, other
4225 clients can make progress
4226 * more extN patch fixes
4227 * compile-time configurable ptlrpc buffer allocations
4229 - collaborative read cache document
4230 - Lustre Lite Performance CDR document-in-progress
4232 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4235 - fix typo in patch-2.4.18
4237 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4240 - clear ptlrpc request each time in handle_incoming_request()
4241 - unlink of files now destroys the object on the OST
4243 2002-09-19 Peter Braam <braam@clusterfs.com>
4245 * add hard link support
4246 * change obdfile creation method
4247 * kernel patch changed
4249 2002-09-19 Peter Braam <braam@clusterfs.com>
4252 - stack overflow bug in extN fixed
4254 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4256 * documentation updates
4257 - add man pages for config tools
4258 - update tests/README to describe testing with new config tools
4259 - finish metadata API descriptions
4260 * bug fixes and cleanups
4261 - statfs workaround for 16TB limit
4262 - LOV stripe allocation improved, can stripe on subset of OSTs
4263 - LOV file size/IO offset was wrong for files > 4GB in size
4264 - object EA data was being dropped, caused files to be unreadable
4265 - memory overflow with non-LOV OST caused memory corruption
4266 - fixed regression tests to work with new config tools, obdfilter
4267 - fixed bug when directory size became larger than 1 block
4268 - fixed bug (for single client case) when PWD was deleted
4269 - invalidate local directory pages when doing intent-based ops
4270 - avoid LDLM oops when lock callback contained bad data
4272 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4274 * documentation updates
4275 * bug fixes and cleanups
4276 - configuration tools
4279 - 64-bit compile warnings
4280 - 64-bit internal statfs data
4282 * test_brw on persistent OST devices
4284 * lprocfs (disabled)
4286 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4288 * documentation updates
4289 * bug fixes and cleanups
4290 * configuration tools
4292 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4295 * many small fixes to 0_5_4
4296 * io/network handling
4297 * thinkos in MDS operations
4299 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4302 * crucial basic fixes to 0.5.3
4303 * IOR, Iozone work over Elan
4306 2002-08-07 Phil Schwan <phil@clusterfs.com>
4307 * version 0_5_3, our first alpha
4308 * we use the new Portals iovs
4309 * documentation updates
4310 * bug fixes and cleanups
4311 * small changes in the DLM wire protocol
4313 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4314 * version 0_5_1 with some initial stability,
4315 * locking on MD and file I/O.
4316 * documentation updates
4317 * several bug fixes since 0.5.0
4318 * small changes in wire protocol
4320 2002-07-18 Phil Schwan <phil@clusterfs.com>
4322 * delivered as Lustre Light Alpha
4323 * fixed a crash after handling invalid MDS requests
4324 * fixed directory pages for architectures with non-4k pages sizes
4326 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4327 * release version v0_4_4
4328 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4330 * Fixes a number of interruption problems with OST operations.
4331 * Update documentation for portals header changes
4332 * Move all wire protocol structs/defines to lustre_idl.h
4333 * Fixes symlink length bug.
4334 * Add tcpdump to repository.
4336 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4337 * release version v0_4_3
4338 * Fixes statfs for inodes on extN.
4339 * Fixes bug in runtests which would delete /etc/hosts.
4340 * Use 64-bit object IDs wherever possible (not into VFS though)
4341 Remove ost_get_info, which is unused by lustre, and out of date.
4343 2002-07-03 Peter Braam <braam@clusterfs.com>
4344 * release version v0_4_2 Fixes a lookup error (type not passed)
4345 * move forward to head of Portals
4346 * move forward to latest Lustre kernel
4348 2002-06-25 Peter Braam <braam@clusterfs.com>
4349 * release version v0_4_1. Hopefully stable on single node use.