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: 1.40.2-cfs5
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: Kernel BUG at ... ldiskfs/ldiskfs/mballoc.c:3853
586 Details : Removing the underlying device (e.g. via a SCSI rescan) can cause a
587 kernel BUG in ldiskfs_mb_discard_inode_preallocations().
589 --------------------------------------------------------------------------------
591 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
593 * Support for kernels:
594 2.6.5-7.283 (SLES 9),
595 2.6.9-55.EL (RHEL 4),
596 2.6.16.46-0.14 (SLES 10),
597 2.6.18.8 vanilla (kernel.org)
598 * Client support for unpatched kernels:
599 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
600 2.6.16 - 2.6.22 vanilla (kernel.org)
601 * Due to recently discovered recovery problems, we do not recommend
602 using patchless RHEL 4 clients with this or any earlier release.
603 * Recommended e2fsprogs version: 1.39.cfs8
604 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
605 * Starting with this release, the ldiskfs backing filesystem required
606 by Lustre is now in its own package, lustre-ldiskfs. This package
607 should be installed. It is versioned separately from Lustre and
608 may be released separately in future.
612 Description: block reactivating mgc import until all deactivates complete
613 Details : Fix race when failing back MDT/MGS to itself (testing)
616 Frequency : at statup only
618 Description: mds_lov_synchronize race leads to various problems
619 Details : simultaneous MDT->OST connections at startup can cause the
620 sync to abort, leaving the OSC in a bad state.
622 Severity : enhancement
624 Description: add optional extra BUILD_VERSION info
625 Details : add a new environment variable (namely LUSTRE_VERS) which allows
626 to override the lustre version.
629 Frequency : 2.6.18 servers only
631 Description: ll_kern_mount() doesn't release the module reference
632 Details : The ldiskfs module reference count never drops down to 0
633 because ll_kern_mount() doesn't release the module reference.
638 Description: server LBUG when using old ost_num_threads parameter
639 Details : Accept the old ost_num_threads parameter but warn that it
640 is deprecated, and fix an off-by-one error that caused an LBUG.
645 Description: Transient SCSI error results in persistent IO issue
646 Details : iobuf->dr_error is not reinitialized to 0 between two
650 Frequency : sometimes when underlying device returns I/O errors
652 Description: OSTs not going read-only during write failures
653 Details : OSTs are not remounted read-only when the journal commit threads
654 get I/O errors because fsfilt_ext3 calls journal_start/stop()
655 instead of the ext3 wrappers.
659 Description: poor connect scaling with increasing client count
660 Details : Don't run filter_grant_sanity_check for more than 100 exports
661 to improve scaling for large numbers of clients.
664 Frequency : SLES10 only
666 Description: sanity-quota.sh quotacheck failed: rc = -22
667 Details : Quotas cannot be enabled on SLES10.
670 Frequency : liblustre clients only
672 Description: getdirentries does not give error when run on compute nodes
673 Details : getdirentries does not fail when the size specified as an argument
674 is too small to contain at least one entry
676 Severity : enhancement
678 Description: Add LNET router traceability for debug purposes
679 Details : If a checksum failure occurs with a router as part of the
680 IO path, the NID of the last router that forwarded the bulk data
681 is printed so it can be identified.
686 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
687 Details : Due to a race condition, liblustre clients were occasionally
690 Severity : enhancement
692 Description: lfs setstripe use optional parameters instead of postional
695 Severity : enhancement
697 Description: Nanosecond timestamp support for ldiskfs
698 Details : The on-disk ldiskfs filesystem has added support for nanosecond
699 resolution timestamps. There is not yet support for this at
700 the Lustre filesystem level.
703 Frequency : during server recovery
705 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
706 Details : request with rq_no_resend flag not awake l_wait_event if they get a
710 Frequency : nfs export on patchless client
712 Description: connectathon hang when test nfs export over patchless client
713 Details : Disconnected dentry cannot be found with lookup, so we do not need
714 to unhash it or make it invalid
717 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
718 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
719 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
724 Description: open req refcounting wrong on reconnect
725 Details : If reconnect happened between getting open reply from server and
726 call to mdc_set_replay_data in ll_file_open, we will schedule
727 replay for unreferenced request that we are about to free.
728 Subsequent close will crash in variety of ways.
729 Check that request is still eligible for replay in
730 mdc_set_replay_data().
735 Description: disable writes to filesystem when reading health_check file
736 Details : the default for reading the health_check proc file has changed
737 to NOT do a journal transaction and write to disk, because this
738 can cause reads of the /proc file to hang and block HA state
739 checking on a healthy but otherwise heavily loaded system. It
740 is possible to return to the previous behaviour during configure
741 with --enable-health-write.
743 Severity : enhancement
745 Description: 64-bit inode version
746 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
747 to the inode. This will be required for version-based recovery.
752 Description: MDS fails to start if a duplicate client export is detected
753 Details : in some rare cases it was possible for a client to connect to
754 an MDS multiple times. Upon recovery the MDS would detect this
755 and fail during startup. Handle this more gracefully.
757 Severity : enhancement
759 Description: Add -o localflock option to simulate old noflock
761 Details : This will achieve local-only flock/fcntl locks
767 Description: log_commit_thread vs filter_destroy race leads to crash
768 Details : Take import reference before releasing llog record semaphore
773 Description: Wrong request locking in request set processing
774 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
775 imp_delayed_list, in this place should be used imp_lock.
778 Frequency : when reconnection
780 Description: Grant Leak when osc reconnect to OST
781 Details : When osc reconnect ost, OST(filter) should check whether it
782 should grant more space to client by comparing fed_grant and
783 cl_avail_grant, and return the granted space to client instead
784 of "new granted" space, because client will call osc_init_grant
785 to update the client grant space info.
788 Frequency : when client reconnect to OST
790 Description: Grant Leak when osc do resend and replay bulk write
791 Details : When osc reconnect to OST, OST(filter)should clear grant info of
792 bulk write request, because the grant info while be sync between
793 OSC and OST when reconnect, and we should ignore the grant info
794 these of resend/replay write req.
799 Description: Grant space more than avaiable left space sometimes.
800 Details : When then OST is about to be full, if two bulk writing from
801 different clients came to OST. Accord the avaliable space of the
802 OST, the first req should be permitted, and the second one
803 should be denied by ENOSPC. But if the seconde arrived before
804 the first one is commited. The OST might wrongly permit second
805 writing, which will cause grant space > avaiable space.
808 Frequency : when client is evicted
810 Description: Grant might be wrongly erased when osc is evicted by OST
811 Details : when the import is evicted by server, it will fork another
812 thread ptlrpc_invalidate_import_thread to invalidate the
813 import, where the grant will be set to 0. While the original
814 thread will update the grant it got when connecting. So if
815 the former happened latter, the grant will be wrongly errased
816 because of this race.
821 Description: Checking Stale with correct fid
822 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
823 to check whether it is stale, because sometimes, we want the
824 enqueue happened anyway, and op_data.fid2 will not be initialized.
826 Severity : enhancement
828 Description: update patchless client
829 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
832 Frequency : only with 2.4 kernel
834 Description: random memory corruption
835 Details : size of struct ll_inode_info is to big for union inode.u and this
836 can be cause of random memory corruption.
841 Description: Memory leak in recovery
842 Details : Lov_mds_md was not free in an error handler in mds_create_object.
843 It should also check obd_fail before fsfilt_start, otherwise if
844 fsfilt_start return -EROFS,(failover mds during mds recovery).
845 then the req will return with repmsg->transno = 0 and rc = EROFS.
846 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
847 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
848 be freed no matter whether fsfilt_commit success or not.
851 Frequency : only with huge count clients
853 Description: Prevents from taking the superblock lock in llap_from_page for
855 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
856 for llap_from_page call in ll_removepage prevents from taking the
857 superblock lock for a soon died page.
862 Description: Not check open intent error before release open handle
863 Details : in some rare cases, the open intent error is not checked before
864 release open handle, which may cause
865 ASSERTION(open_req->rq_transno != 0), because it tries to release
866 the failed open handle.
871 Description: Set cat log bitmap only after create log success.
872 Details : in some rare cases, the cat log bitmap is set too early. and it
873 should be set only after create log success.
877 Description: Accessing a block bevice can re-enable I/O when Lustre is
878 tearing down a device.
879 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
884 Description: service threads may hog cpus when there are a lot of requests
885 Details : Insert cond_resched to give other threads a chance to use some CPU
890 Description: the cat log was not initialized in recovery
891 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
892 unlink log on mds will not be initialized until mds post
893 recovery. And also in mds post recovery, the unlink log will
894 initialization will be done asynchronausly, so there will be race
895 between add unlink log and unlink log initialization.
899 Description: brw_stats were being printed incorrectly
900 Details : brw_stats were being printed as log2 but all of them were not
901 recorded as log2. Also remove some code duplication arising from
902 filter_tally_{read,write}.
906 Frequency : rare, only in recovery.
907 Description: ASSERTION(req->rq_type != LI_POISON) failed
908 Details : imp_lock should be held while iterating over imp_sending_list for
909 prevent destroy request after get timeout in ptlrpc_queue_wait.
913 Description: replay-single.sh test 52 fails
914 Details : A lock's skiplist need to be cleanup when it being unlinked
915 from its resource list.
919 Description: Short directio read returns full requested size rather than
921 Details : Direct I/O operations should return actual amount of bytes
922 transferred rather than requested size.
924 Severity : enhancement
926 Description: metadata RPC reduction (e.g. for rm performance)
927 Details : decrease the amount of synchronous RPC between clients and servers
928 by canceling conflicing lock before the operation on the client
929 and packing thier handles into the main operation RPC to server.
931 Severity : enhancement
933 Description: Async OSC create to avoid the blocking unnecessarily.
934 Details : If a OST has no remain object, system will block on the creating
935 when need to create a new object on this OST. Now, ways use
936 pre-created objects when available, instead of blocking on an
937 empty osc while others are not empty. If we must block, we block
938 for the shortest possible period of time.
940 Severity : enhancement
942 Description: refine locking for avoid write wrong info into lov_objid file
943 Details : fix possible races with add new target and write/update data in
946 --------------------------------------------------------------------------------
948 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
953 Frequency : on some architectures
955 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
956 Details : Uninitialized flags sometimes cause configuration commands to
960 Frequency : patchless clients only
962 Description: missing __iget() symbol export
963 Details : The __iget() symbol export is missing. To avoid the need for
964 this on patchless clients the deathrow inode reaper is turned
965 off, and we depend on the VM to clean up old inodes. This
966 dependency was during via the fix for bug 12181.
968 --------------------------------------------------------------------------------
970 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
972 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
973 INTEROPERATE with older versions automatically. In many cases a
974 special upgrade step is needed. Please read the
975 user documentation before upgrading any part of a live system.
976 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
977 version will not interoperate with 1.6 betas before beta5 (1.5.95).
978 * WARNING: Lustre configuration and startup changes are required with
979 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
985 Severity : enhancement
987 Description: MountConf
988 Details : Lustre configuration is now managed via mkfs and mount
989 commands instead of lmc and lconf. New obd types (MGS, MGC)
990 are added for dynamic configuration management. See
991 https://mail.clusterfs.com/wikis/lustre/MountConf for
994 Severity : enhancement
996 Description: dynamic OST addition
997 Details : OSTs can now be added to a live filesystem
999 Severity : enhancement
1001 Description: startup order invariance
1002 Details : MDTs and OSTs can be started in any order. Clients only
1003 require the MDT to complete startup.
1005 Severity : enhancement
1007 Description: parallel, asynchronous orphan cleanup
1008 Details : orphan cleanup is now performed in separate threads for each
1009 OST, allowing parallel non-blocking operation.
1011 Severity : enhancement
1013 Description: optimized stripe assignment
1014 Details : stripe assignments are now made based on ost space available,
1015 ost previous usage, and OSS previous usage, in order to try
1016 to optimize storage space and networking resources.
1018 Severity : enhancement
1020 Description: Permanently set tunables
1021 Details : All writable /proc/fs/lustre tunables can now be permanently
1022 set on a per-server basis, at mkfs time or on a live system.
1024 Severity : enhancement
1026 Description: Lustre message v2
1027 Details : Add lustre message format v2.
1029 Severity : enhancement
1031 Description: client OST exclusion list
1032 Details : Clients can be started with a list of OSTs that should be
1033 declared "inactive" for known non-responsive OSTs.
1037 Description: SPEC SFS validation failure on NFS v2 over lustre.
1038 Details : Changes the blocksize for regular files to be 2x RPC size,
1039 and not depend on stripe size.
1041 Severity : enhancement
1043 Description: Multiple MD RPCs in flight.
1044 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1045 To avoid overly-overloading MDS, introduce a limit on number of
1046 MDS RPCs in flight for a single client and add /proc controls
1047 to adjust this limit.
1049 Severity : enhancement
1051 Description: client read/write statistics
1052 Details : Add client read/write call usage stats for performance
1053 analysis of user processes.
1054 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1055 file access. extents_stats shows chunk size distribution.
1056 extents_stats_per_process show chunk size distribution per
1059 Severity : enhancement
1061 Description: mds statistics
1062 Details : Add detailed mds operations statistics in
1063 /proc/fs/lustre/mds/*/stats.
1067 Description: Failure of copying files with lustre special EAs.
1068 Details : Client side always return success for setxattr call for lustre
1069 special xattr (currently only "trusted.lov").
1074 Description: Refcount LNET uuids
1075 Details : The global LNET uuid list grew linearly with every startup;
1076 refcount repeated list entries instead of always adding to
1079 Severity : enhancement
1081 Description: Dynamic service threads
1082 Details : Within a small range, start extra service threads
1083 automatically when the request queue builds up.
1086 Frequency : mixed-endian client/server environments
1088 Description: mixed-endian crashes
1089 Details : The new msg_v2 system had some failures in mixed-endian
1092 Severity : enhancement
1094 Description: Easy OST removal
1095 Details : OSTs can be permanently deactivated with e.g. 'lctl
1096 conf_param lustre-OST0001.osc.active=0'
1098 Severity : enhancement
1100 Description: MGS proc entries
1101 Details : Added basic proc entries for the MGS showing what filesystems
1104 Severity : enhancement
1106 Description: provide MGS failover
1107 Details : Added config lock reacquisition after MGS server failover.
1109 Severity : enhancement
1111 Description: add Linux 2.4 support
1112 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1116 Description: a large application tries to do I/O to the same resource and dies
1117 in the middle of it.
1118 Details : Check the req->rq_arrival time after the call to
1119 ost_brw_lock_get(), but before we do anything about
1120 processing it & sending the BULK transfer request. This
1121 should help move old stale pending locks off the queue as
1122 quickly as obd_timeout.
1125 Frequency : when an incorrect nid is specified during startup
1127 Description: ptlrpc connect to non-existant node causes kernel crash
1128 Details : LNET can't be re-entered from an event callback, which
1129 happened when we expire a message after the export has been
1130 cleaned up. Instead, hand the zombie cleanup off to another
1133 Severity : enhancement
1135 Description: plain/inodebits lock performance improvement
1136 Details : Grouping plain/inodebits in granted list by their request modes
1137 and bits policy, thus improving the performance of search through
1141 Frequency : only if OST filesystem is corrupted
1143 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1144 Details : for a short time RPCs with bulk IO are in the replay list,
1145 but replay of bulk IOs is unimplemented. If the OST filesystem
1146 is corrupted due to disk cache incoherency and then replay is
1147 started it is possible to trip an assertion. Avoid putting
1148 committed RPCs into the replay list at all to avoid this issue.
1151 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1153 Description: System hang on startup
1154 Details : This bug allowed the liblustre (e.g. catamount) client to
1155 return to the app before handling all startup RPCs. This
1156 could leave the node unresponsive to lustre network traffic
1157 and manifested as a server ptllnd timeout.
1159 Severity : enhancement
1161 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1162 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1163 boolean to control whether to print peer debug info when a
1164 client's RPC times out.
1167 Frequency : only for kernels with patches from Lustre below 1.4.3
1169 Description: Remove old rdonly API
1170 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1173 Frequency : only for devices with external journals
1175 Description: Set external device read-only also
1176 Details : During a commanded failover stop, we set the disk device
1177 read-only while the server shuts down. We now also set any
1178 external journal device read-only at the same time.
1181 Frequency : when upgrading from 1.4 while trying to change parameters
1183 Description: The wrong (new) MDC name was used when setting parameters for
1184 upgraded MDT's. Also allows changing of OSC (and MDC)
1185 parameters if --writeconf is specified at tunefs upgrade time.
1188 Frequency : when setting specific ost indicies
1190 Description: QOS code breaks on skipped indicies
1191 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1192 created with --index need not be sequential.
1194 Severity : enhancement
1196 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1197 Details : The uninit_groups feature works in conjunction with the kernel
1198 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1199 up the pass1 processing of e2fsck. This is a read-only feature
1200 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1201 mount filesystems that have had this feature enabled.
1203 Severity : enhancement
1205 Description: Improve multi-block allocation algorithm to avoid fragmentation
1206 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1207 allocation locality and avoid filesystem fragmentation.
1209 ------------------------------------------------------------------------------
1211 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1213 * Support for kernels:
1214 2.4.21-47.0.1.EL (RHEL 3)
1215 2.6.5-7.283 (SLES 9)
1216 2.6.9-42.0.10.EL (RHEL 4)
1217 2.6.12.6 vanilla (kernel.org)
1218 2.6.16.27-0.9 (SLES 10)
1219 * Recommended e2fsprogs version: 1.39.cfs5
1221 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1225 Frequency : occasional, depends on client load and configuration
1226 Bugzilla : 12181, 12203
1227 Description: data loss for recently-modified files
1229 Details : In some cases it is possible that recently written or created
1230 files may not be written to disk in a timely manner (this should
1231 normally be within 30s unless client IO load is very high).
1232 The problem appears as zero-length files or files that are a
1233 multiple of 1MB in size after a client crash or client eviction
1234 that are missing data at the end of the file.
1236 This problem is more likely to be hit on clients where files are
1237 repeatedly created and unlinked in the same directory, clients
1238 have a large amount of RAM, have many CPUs, the filesystem has
1239 many OSTs, the clients are rebooted frequently, and/or the files
1240 are not accessed by other nodes after being written.
1242 The presence of the problem can be detected by looking at
1243 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1244 smaller than the second (nr_unused) then dirty files will not
1245 be flushed automatically to disk. "sync; sleep 10" should be
1246 run several times on the node before unmounting it to update
1247 Lustre (this is also safe to run on nodes without this problem).
1249 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1250 that can cause this same problem, so customers using that kernel
1251 also need to update the kernel in addition to Lustre. In order
1252 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1254 It is normal that files written just before a client crash (less
1255 than 30s) may not yet have been flushed to disk, even for local
1259 Frequency : frequent on thin XT3 nodes
1261 Description: UUID collision on thin XT3 Linux nodes
1262 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1263 randomly, since we relied on an insufficiently-seeded PRNG.
1268 Description: OSS hangs after "All ost request buffers busy"
1269 Details : A deadlock between quota and journal operations caused OSS
1270 hangs after printing "All ost request buffers busy."
1273 Frequency : always on liblustre builds
1275 Description: Cleanup compiler warnings on liblustre
1278 Frequency : always on liblustre builds on XT3
1280 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1284 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1285 > 32000 subdirectories
1286 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1287 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1288 subdirectory count crosses 32000. This will aid e2fsck to
1289 correctly handle more than 32000 subdirectories.
1292 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1294 Description: System hang on startup
1295 Details : This bug allowed the liblustre (e.g. catamount) client to
1296 return to the app before handling all startup RPCs. This
1297 could leave the node unresponsive to lustre network traffic
1298 and manifested as a server ptllnd timeout.
1300 Severity : enhancement
1302 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1303 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1304 boolean to control whether to print peer debug info when a
1305 client's RPC times out.
1310 Description: make O_SYNC working on 2.6 kernels
1311 Details : 2.6 kernels use different method for mark pages for write,
1312 so need add a code to lustre for O_SYNC work.
1317 Description: Failure to close file and release space on NFS
1318 Details : Put inode details into lock acquired in ll_intent_file_open.
1319 Use mdc_intent_lock in ll_intent_open to properly
1320 detect all kind of errors unhandled by mdc_enqueue.
1325 Description: proc file read during shutdown sometimes raced obd removal,
1327 Details : Add lock to prevent obd access after proc file removal.
1330 Frequency : Only for files larger than 4GB on 32-bit clients.
1332 Description: improperly doing page alignment of locks
1333 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1334 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1339 Description: RPCs being resent when they shouldn't be
1340 Details : Some RPCs that should not be resent are being resent. This
1341 can cause inconsistencies in the RPC state machine. Do not
1342 resend such requests.
1345 Frequency : rare, only with NFS export
1347 Description: Crash on NFS re-export node
1348 Details : under very unusual load conditions an assertion is hit in
1349 ll_intent_file_open()
1352 Frequency : only if OST filesystem is corrupted
1354 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1355 Details : for a short time RPCs with bulk IO are in the replay list,
1356 but replay of bulk IOs is unimplemented. If the OST filesystem
1357 is corrupted due to disk cache incoherency and then replay is
1358 started it is possible to trip an assertion. Avoid putting
1359 committed RPCs into the replay list at all to avoid this issue.
1364 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1365 Details : Large single O_DIRECT read and write calls can fail to allocate
1366 a sufficiently large buffer to process the request. In case of
1367 allocation failure the allocation is retried with a smaller
1368 buffer and broken into smaller requests.
1370 Severity : enhancement
1372 Description: Add -o localflock option to simulate old noflock behaviour.
1373 Details : This will achieve local-only flock/fcntl locks coherentness.
1378 Description: versioning check is incomplete
1379 Details : Checking the version difference of client vs. server, report
1380 error if the gap is too big.
1385 Description: add support PG_writeback bit
1386 Details : add support for PG_writeback bit for Lustre, for more carefull
1387 work with page cache in 2.6 kernel. This also fix some deadlocks
1388 and remove hack for work O_SYNC with 2.6 kernel.
1390 Severity : enhancement
1392 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1393 Details : The uninit_groups feature works in conjunction with the kernel
1394 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1395 up the pass1 processing of e2fsck. This is a read-only feature
1396 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1397 mount filesystems that have had this feature enabled.
1399 Severity : enhancement
1401 Description: Improve multi-block allocation algorithm to avoid fragmentation
1402 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1403 allocation locality and avoid filesystem fragmentation.
1405 ------------------------------------------------------------------------------
1407 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1409 * Support for kernels:
1410 2.6.9-42.0.3.EL (RHEL 4)
1411 2.6.5-7.276 (SLES 9)
1412 2.4.21-47.0.1.EL (RHEL 3)
1413 2.6.12.6 vanilla (kernel.org)
1414 2.6.16.21-0.8 (SLES10)
1415 * Recommended e2fsprogs version: 1.39.cfs2-0
1417 * The backwards-compatible /proc/sys/portals symlink has been removed
1418 in this release. Before upgrading, please ensure that you change
1419 any configuration scripts or /etc/sysctl.conf files that access
1420 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1421 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1422 in advance of the upgrade on any system running Lustre 1.4.6 or
1423 newer, since /proc/sys/lnet was added in that version.
1424 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1428 Frequency : only when quota is used
1430 Description: avoid scanning export list for quota master
1431 Details : Change the algorithms to avoid scanning export list in order
1432 to improve the efficiency.
1435 Frequency : MDS failover only, very rarely
1437 Description: "went back in time" messages on mds failover
1438 Details : The greatest transno may be lost when the current operation
1439 finishes with an error (transno==0) and the client's last_rcvd
1440 record is over-written. Save the greatest transno in the
1441 mds_last_transno for this case.
1444 Frequency : always for specific kernels and striping counts
1446 Description: client may get "Matching packet too big" without ACL support
1447 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1448 when trying to access files in certain configurations. The
1449 clients should in fact be denied when mounting because they do
1450 not understand ACLs.
1453 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1455 Description: many clients connecting with IO in progress causes connect timeouts
1456 Details : Avoid synchronous journal commits to avoid delays caused by many
1457 clients connecting/disconnecting when bulk IO is in progress.
1458 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1459 OST_IO_PORTAL to avoid delays behind potentially many pending
1463 Frequency : occasionally with multiple writers to a single file
1465 Description: shared writes to file may result in wrong size reported by stat()
1466 Details : Allow growing of kms when extent lock is cancelled
1469 Frequency : always with random mmap IO to multi-striped file
1471 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1472 Details : Only if the hole is at the end of OST object so that kms is too
1473 small. Fix is to increase kms accordingly in ll_nopage.
1476 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1478 Description: writes to a missing object would leak memory on the OST
1479 Details : If there is an inconsistency between the MDS and OST filesystems,
1480 such that the MDS references an object that doesn't exist, writes
1481 to that object will leak memory due to incorrect cleanup in the
1482 error handling path, eventually running out of memory on the OST.
1487 Description: Creating too long symlink causes lustre errors
1488 Details : Check symlink and name lengths before sending requests to MDS.
1491 Frequency : only if flock is enabled (not on by default)
1493 Description: posix locks not released on fd closure on 2.6.9+
1494 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1495 kernels, this caused such locks not to be released on fd close and
1496 then assertions on fs unmount about still used locks.
1499 Frequency : MDS failover only, very rarely
1501 Description: clients may get ASSERTION(granted_lock != NULL)
1502 Details : When request was taking a long time, and a client was resending
1503 a getattr by name lock request. The were multiple lock requests
1504 with the same client lock handle and
1505 mds_getattr_name->fixup_handle_for_resent_request found one of the
1506 lock handles but later failed with ASSERTION(granted_lock != NULL).
1511 Description: handle->h_buffer_credits > 0, assertion failure
1512 Details : h_buffer_credits is zero after truncate, causing assertion
1513 failure. This patch extends the transaction or creates a new
1517 Frequency : NFS re-export or patchless client
1518 Bugzilla : 11179, 10796
1519 Description: Crash on NFS re-export node (__d_move)
1520 Details : We do not want to hash the dentry if we don't have a lock.
1521 But if this dentry is later used in d_move, we'd hit uninitialised
1522 list head d_hash, so we just do this to init d_hash field but
1523 leave dentry unhashed.
1526 Frequency : NFS re-export or patchless client
1528 Description: NFS exports has problem with symbolic link
1529 Details : lustre client didn't properly install dentry when re-exported
1530 to NFS or running patchless client.
1533 Frequency : NFS re-export or patchless client
1535 Description: Various nfs/patchless fixes.
1536 Details : fixes reuse disconected alias for lookup process - this fixes
1537 warning "find_exported_dentry: npd != pd",
1538 fix permission error with open files at nfs.
1539 fix apply umask when do revalidate.
1542 Frequency : occasional
1544 Description: Crash on NFS re-export node
1545 Details : calling clear_page() on the wrong pointer triggered oops in
1546 generic_mapping_read().
1549 Frequency : rarely, using O_DIRECT IO
1551 Description: unaligned directio crashes client with LASSERT
1552 Details : check for unaligned buffers before trying any requests.
1555 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1557 Description: stale data returned from RAID cache
1558 Details : If only a small amount of IO is done to the RAID device before
1559 reading it again it is possible to get stale data from the RAID
1560 cache instead of reading it from disk.
1563 Frequency : always for sles10 kernel
1565 Description: sles10 support
1566 Details : ll_follow_link: compile fixes and using of nd_set_link
1567 under newer kernels.
1570 Frequency : depends on arch, kernel and compiler version, always on sles10
1573 Description: recursive or deep enough symlinks cause stack overflow
1574 Details : getting rid of large stack-allocated variable in
1578 Frequency : depends on hardware
1580 Description: lustre write performance loss in the SLES10 kernel
1581 Details : the performance loss is caused by using of write barriers in the
1582 ext3 code. The SLES10 kernel turns barrier support on by
1583 default. The fix is to undo that change for ldiskfs.
1585 ------------------------------------------------------------------------------
1587 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1589 * Support for kernels:
1590 2.6.9-42.0.3EL (RHEL 4)
1591 2.6.5-7.276 (SLES 9)
1592 2.4.21-47.0.1.EL (RHEL 3)
1593 2.6.12.6 vanilla (kernel.org)
1597 Frequency : quota enabled and large files being deleted
1599 Description: releasing more than 4GB of quota at once hangs OST
1600 Details : If a user deletes more than 4GB of files on a single OST it
1601 will cause the OST to spin in an infinite loop. Release
1602 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1607 Description: statfs data retrieved from /proc may be stale or zero
1608 Details : When reading per-device statfs data from /proc, in the
1609 {kbytes,files}_{total,free,avail} files, it may appear
1610 as zero or be out of date.
1613 Frequency : systems with MD RAID1 external journal devices
1615 Description: lconf's call to blkid is confused by RAID1 journal devices
1616 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1617 of returning all block devices that match the journal UUID.
1620 Frequency : always, for aggregate stripe size over 4GB
1622 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1623 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1624 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1627 Frequency : always if "lfs find" used on a local file/directory
1629 Description: "lfs find" segfaults if used on a local file/directory
1630 Details : The case where a directory component was not specified wasn't
1631 handled correctly. Handle this properly.
1634 Frequency : always on ppc64
1636 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1637 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1638 perform endianness conversion.
1641 Frequency : rarely (truncate to non-zero file size after write under load)
1642 Bugzilla : 10730, 10687
1643 Description: Files padded with zeros to next 4K multiple
1644 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1645 it is possible that files that are truncated to a non-zero size
1646 immediately after being written are filled with zero bytes beyond
1647 the truncated size. No file data is lost.
1649 Severity : enhancement
1651 Description: Allow recovery/failover for liblustre clients.
1652 Details : liblustre clients were unaware of failover configurations until
1655 Severity : enhancement
1657 Description: user file locks should fail when not mounting with flock option
1658 Details : Set up an error-returning stub in ll_file_operations.lock field
1659 to prevent incorrect behaviour when client is mounted without
1660 flock option. Also, set up properly f_op->flock field for
1664 Frequency : always on ia64
1666 Description: "lfs df" loops on printing out MDS statfs information
1667 Details : The obd_ioctl_data was not initialized and in some systems
1668 this caused a failure during the ioctl that did not return
1669 an error. Initialize the struct and return an error on failure.
1672 Frequency : SLES 9 only
1674 Description: Error of copying files with lustre special EAs as root
1675 Details : Client side always return success for setxattr call for lustre
1676 special xattr (currently only "trusted.lov").
1679 Frequency : rarely on clusters with both ia64+i386 clients
1681 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1682 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1683 clients are doing concurrent writes to the same file it is
1684 possible that the ia64 clients may LASSERT because the OST
1685 extent locks are not PAGE_SIZE aligned. Ensure that grown
1686 locks are always aligned on the request boundary.
1689 Frequency : specific use, occasional
1691 Description: Overwriting in use executable truncates on-disk binary image
1692 Details : If one node attempts to overwrite an executable in use by
1693 another node, we now correctly return ETXTBSY instead of
1694 truncating the file.
1696 Severity : enhancement
1698 Description: Async OSC create to avoid the blocking unnecessarily.
1699 Details : If a OST has no remain object, system will block on the creating
1700 when need to create a new object on this OST. Now, ways use
1701 pre-created objects when available, instead of blocking on an
1702 empty osc while others are not empty. If we must block, we block
1703 for the shortest possible period of time.
1708 Description: chmod on Lustre root is propagated to other clients
1709 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1710 invalid by the follow_mount time.
1713 Frequency : liblustre clients only
1715 Description: Race in 'instant cancel' lock handling could lead to such locks
1716 never to be granted in case of SMP MDS
1717 Details : Do not destroy not yet granted but cbpending locks in
1721 Frequency : replay/resend of open
1723 Description: non null lock assetion failure in mds_intent_policy
1724 Details : Trying to replay/resend lockless open requests resulted in
1725 mds_open() returning 0 with no lock. Now it sets a flag if
1726 a lock is going to be returned.
1728 Severity : enhancement
1730 Description: Checksum enhancements
1731 Details : New checksum enhancements allow for resending RPCs that failed
1734 Severity : enhancement
1736 Description: Tunables on number of dirty pages in cacche
1737 Details : Allow to set limit on number of dirty pages cached.
1742 Description: client crash on unmount - lock still has references
1743 Details : In some error handling cases it was possible to leak a lock
1744 reference on a client while accessing a file. This was not
1745 harmful to the client during operation, but would cause the
1746 client to crash when the filesystem is unmounted.
1749 Frequency : specific case, rare
1751 Description: ETXTBSY on mds though file not in use by client
1752 Details : ETXTBSY is no longer incorrectly returned when attempting to
1753 chmod or chown a directory that the user previously tried to
1754 execute or a currently-executing binary.
1757 Frequency : extremely rare except on liblustre-based clients
1759 Description: Lustre space not freed when files are deleted
1760 Details : Clean up open-unlinked files after client eviction. Previously
1761 the unlink was skipped and the files remained as orphans.
1766 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1767 Details : In some cases it was possible to send a blocking callback to
1768 a client doing a glimpse, even though that client didn't get
1769 a lock granted. When the glimpse lock is cancelled on the OST
1770 the freed lock is left on the waiting list and corrupted the list.
1773 Frequency : all core dumps
1775 Description: Broke core dumps to lustre
1776 Details : Negative dentry may be unhashed if parent does not have UPDATE
1777 lock, but some callers, e.g. do_coredump, expect dentry to be
1778 hashed after successful create, hash it in ll_create_it.
1780 ------------------------------------------------------------------------------
1782 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1784 * Support for kernels:
1785 2.6.9-42.0.2.EL (RHEL 4)
1786 2.6.5-7.276 (SLES 9)
1787 2.4.21-40.EL (RHEL 3)
1788 2.6.12.6 vanilla (kernel.org)
1792 Frequency : always on RHEL 3
1794 Description: Number of open files grows over time
1795 Details : The number of open files grows over time, whether or not
1796 Lustre is started. This was due to a filp leak introduced
1797 by one of our kernel patches.
1799 ------------------------------------------------------------------------------
1801 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1803 * Support for kernels:
1804 2.6.9-42.EL (RHEL 4)
1805 2.6.5-7.276 (SLES 9)
1806 2.4.21-40.EL (RHEL 3)
1807 2.6.12.6 vanilla (kernel.org)
1812 Bugzilla : 5719, 9635, 9792, 9684
1813 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1814 Details : If a server is under heavy load and cannot reply to new
1815 connection requests before the client resends the (re)connect,
1816 the connection handling code can behave badly if two service
1817 threads are concurrently handing separate (re)connections from
1818 the same client. Add better locking to the connection handling
1819 code, and ensure that only a single connection will be processed
1820 for a given client UUID, even if the lock is dropped.
1822 Severity : enhancement
1824 Description: add TCP zero-copy support to kernel
1825 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1826 sends if the hardware supports scatter-gather and checksumming.
1827 This allows socklnd to do client-write and server-read more
1828 efficiently and reduce CPU utilization from skbuf copying.
1831 Frequency : only if NFS exporting from client
1833 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1834 Details : If setting attributes on a file created under NFS that had
1835 never been opened it would be possible to oops the client
1836 if the file had no objects.
1839 Frequency : always for liblustre
1841 Description: liblustre client does MDS+OSTs setattr RPC for each write
1842 Details : When doing a write from a liblustre client, the client
1843 incorrectly issued an RPC to the MDS and each OST the file was
1844 striped over in order to update the timestamps. When writing
1845 with small chunks and many clients this could overwhelm the MDS
1846 with RPCs. In all cases it would slow down the write because
1847 these RPCs are unnecessary.
1849 Severity : enhancement
1851 Description: allow number of MDS service threads to be changed at module load
1852 Details : It is now possible to change the number of MDS service threads
1853 running. Adding "options mds mds_num_threads={N}" to the MDS's
1854 /etc/modprobe.conf will set the number of threads for the next
1855 time Lustre is restarted (assuming the "mds" module is also
1856 reloaded at that time). The default number of threads will
1857 stay the same, 32 for most systems.
1862 Description: OST crash if filesystem is unformatted or corrupt
1863 Details : If an OST is started on a device that has never been formatted
1864 or if the filesystem is corrupt and cannot even mount then the
1865 error handling cleanup routines would dereference a NULL pointer.
1870 Description: NULL pointer deref in llap_from_page.
1871 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1872 incorrect) mapping if the page was truncated/reclaimed while it was
1873 searched for. Check for this condition and skip such pages when
1874 doing readahead. Introduce extra check to llap_from_page() to
1875 verify page->mapping->host is non-NULL (so page is not anonymous).
1878 Frequency : Sometimes when using sys_sendfile
1880 Description: "page not covered by a lock" warnings from ll_readpage
1881 Details : sendfile called ll_readpage without right page locks present.
1882 Now we introduced ll_file_sendfile that does necessary locking
1883 around call to generic_file_sendfile() much like we do in
1887 Frequency : with certain MDS communication failures at client mount time
1889 Description: NULL pointer deref after failed client mount
1890 Details : a client connection request may delayed by the network layer
1891 and not be sent until after the PTLRPC layer has timed out the
1892 request. If the client fails the mount immediately it will try
1893 to clean up before the network times out the request. Add a
1894 reference from the request import to the obd device and delay
1895 the cleanup until the network drops the request.
1898 Frequency : occasionally during client (re)connect
1900 Description: assertion failure during client (re)connect
1901 Details : processing a client connection request may be delayed by the
1902 client or server longer than the client connect timeout. This
1903 causes the client to resend the connection request. If the
1904 original connection request is replied in this interval, the
1905 client may trip an assertion failure in ptlrpc_connect_interpret()
1906 which thought it would be the only running connect process.
1909 Frequency : only with obd_echo servers and clients that are rebooted
1911 Description: kernel BUG accessing uninitialized data structure
1912 Details : When running an obd_echo server it did not start the ping_evictor
1913 thread, and when a client was evicted an uninitialized data
1914 structure was accessed. Start the ping_evictor in the RPC
1915 service startup instead of the OBD startup.
1917 Severity : enhancement
1918 Bugzilla : 10193 (patchless)
1919 Description: Remove dependency on various unexported kernel interfaces.
1920 Details : No longer need reparent_to_init, exit_mm, exit_files,
1921 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1924 Frequency : rare (only users of deprecated and unsupported LDAP config)
1926 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1927 Details : LDAP apparently contains 'lustreName' attributes instead of
1928 'name'. A simple remapping of the name is sufficient.
1931 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1933 Description: waiting_locks_callback trips kernel BUG if client is evicted
1934 Details : Running with the dump_on_timeout debug flag turned on makes
1935 it possible that the waiting_locks_callback() can try to dump
1936 the Lustre kernel debug logs from an interrupt handler. Defer
1937 this log dumping to the expired_lock_main() thread.
1939 Severity : enhancement
1941 Description: Support NFS exporting on 2.6 kernels.
1942 Details : Implement non-rawops metadata methods for NFS server to use without
1943 changing NFS server code.
1946 Frequency : very rare (synthetic metadata workload only)
1948 Description: two racing renames might cause an MDS thread to deadlock
1949 Details : Running the "racer" program may cause one MDS thread to rename
1950 a file from being the source of a rename to being the target of
1951 a rename at exactly the same time that another thread is doing
1952 so, and the second thread has already enqueued these locks after
1953 doing a lookup of the target and is trying to relock them in
1954 order. Ensure that we don't try to re-lock the same resource.
1957 Frequency : only very large systems with liblustre clients
1959 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1960 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1961 all OSTs in parallel. This allows the eviction of stale liblustre
1962 clients to proceed much faster than if they were done in series,
1963 and also offers similar improvements for other set_info RPCs.
1968 Description: excessive CPU usage during initial read phase on client
1969 Details : During the initial read phase on a client, it would agressively
1970 retry readahead on the file, consuming too much CPU and impacting
1971 performance (since 1.4.5.8). Improve the readahead algorithm
1972 to avoid this, and also improve some other common cases (read
1973 of small files in particular, where "small" is files smaller than
1974 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1979 Description: MDS crash when receiving packet with unknown intent.
1980 Details : Do not LBUG in unknown intent case, just return -EFAULT
1982 Severity : enhancement
1983 Bugzilla : 9293, 9385
1984 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1985 Details : Do not serialize getattr (non-intent version) and statfs.
1988 Frequency : occasional, when OST network is overloaded/intermittent
1990 Description: client evicted by OST after bulk IO timeout
1991 Details : If a client sends a bulk IO request (read or write) the OST
1992 may evict the client if it is unresposive to its data GET/PUT
1993 request. This is incorrect if the network is overloaded (takes
1994 too long to transfer the RPC data) or dropped the OST GET/PUT
1995 request. There is no need to evict the client at all, since
1996 the pinger and/or lock callbacks will handle this, and the
1997 client can restart the bulk request.
2000 Frequency : Always when mmapping file with no objects
2002 Description: client crashes when mmapping file with no objects
2003 Details : Check that we actually have objects in a file before doing any
2004 operations on objects in ll_vm_open, ll_vm_close and
2010 Description: Request leak when working with deleted CWD
2011 Details : Introduce advanced request refcount tracking for requests
2012 referenced from lustre intent.
2014 Severity : Enhancement
2016 Description: Cache open file handles on client.
2017 Details : MDS now will return special lock along with openhandle, if
2018 requested and client is allowed to hold openhandle, even if unused,
2019 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2020 closing files for every read/write openration.
2022 Severity : Enhancement
2024 Description: Cache open negative dentries on client when possible.
2025 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2026 negative dentries on lock revocation.
2031 Description: Remounting a client read-only wasn't possible with a zconf mount
2032 Details : It wasn't possible to remount a client read-only with llmount.
2034 Severity : enhancement
2035 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2036 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2037 MPICH is included to add Lustre-specific ADIO interfaces.
2038 This is based closely on the UFS ADIO layer and only differs
2039 in file creation, in order to allow the OST striping to be set.
2040 This is user-contributed code and not supported by CFS.
2045 Description: extended inode attributes (immutable, append-only) work improperly
2046 when 2.4 and 2.6 kernels are used on client/server or vice versa
2047 Details : Introduce kernel-independent values for these flags.
2049 Severity : enhancement
2052 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2053 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2054 granularity. Now, Fractional MB granularity is be supported,
2055 this is very useful for low memory system.
2057 Severity : enhancement
2059 Description: Getattr by fid
2060 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2061 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2062 have locks and avoids getattr rpc requests for it on every stat.
2065 Frequency : Always, for filesystems larger than 2TB
2067 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2068 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2069 filesystems larger than 2TB to be mounted with mballoc enabled.
2072 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2074 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2075 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2076 filesystems larger than 2TB is mounted then the kernel will
2077 silently corrupt the start of the filesystem. CONFIG_LBD is
2078 enabled for all CFS-supported kernels, but the possibility of
2079 this happening with a modified kernel config exists.
2081 Severity : enhancement
2083 Description: add client O_DIRECT support for 2.6 kernels
2084 Details : It is now possible to do O_DIRECT reads and writes to files
2085 in the Lustre client mountpoint on 2.6 kernel clients.
2087 Severity : enhancement
2089 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2090 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2091 parallel, not waiting for response from every OST before sending
2092 a rpc to the next OST.
2097 Description: setattr vs write race when updating file timestamps
2098 Details : Client processes that update a file timestamp into the past
2099 right after writing to the file (e.g. tar) it is possible that
2100 the updated file modification time can be reset to the current
2101 time due to a race between processing the setattr and write RPC.
2103 Severity : enhancement
2105 Description: Bring 'lfs find' closer in line with regular Linux find.
2106 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2107 -print0 options and obtains all the needed info through the lustre
2110 Severity : enhancement
2112 Description: support up to 1024 configured devices on one node
2113 Details : change obd_dev array from statically allocated to dynamically
2114 allocated structs as they are first used to reduce memory usage
2119 Description: Flush dirty partially truncated pages during truncate
2120 Details : Immediatelly flush partially truncated pages in filter_setattr,
2121 this way we completely avoid having any pages in page cache on OST
2122 and can retire ugly workarounds during writes to flush such pages.
2127 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2128 Details : protect inode from truncation within vfs_unlink() context
2129 just take a reference before calling vfs_unlink() and release it
2130 when parent's i_sem is free.
2135 Description: last_id value checked outside lock on OST caused LASSERT failure
2136 Details : If there were multiple MDS->OST object precreate requests in
2137 flight, it was possible that the OST's last object id was checked
2138 outside a lock and incorrectly tripped an assertion. Move checks
2139 inside locks, and discard old precreate requests.
2142 Frequency : always, if extents are used on OSTs
2144 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2145 Details : OSTs using the extents format would not zero the high 16 bits of
2146 the index physical block number. This is not a problem for any
2147 OST filesystems smaller than 16TB, and no kernels support ext3
2148 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2149 new/modified files) and can be fixed for existing filesystems
2150 with e2fsprogs-1.39-cfs1.
2155 Description: import connection selection may be incorrect if timer wraps
2156 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2157 import connections to be ignored if the 32-bit jiffies counter
2158 wraps. Use a 64-bit jiffies counter.
2161 Frequency : very large clusters immediately after boot
2163 Description: LNET request buffers exhausted under heavy short-term load
2164 Details : If a large number of client requests are generated on a service
2165 that has previously never seen so many requests it is possible
2166 that the request buffer growth cannot keep up with the spike in
2167 demand. Instead of dropping incoming requests, they are held in
2168 the LND until the RPC service can accept more requests.
2171 Frequency : Sometimes during replay
2173 Description: Assertion failure in ll_local_open after replay.
2174 Details : If replay happened on an open request reply before we were able
2175 to set replay handler, reply will become not swabbed tripping the
2176 assertion in ll_local_open. Now we set the handler right after
2177 recognising of open request
2180 Frequency : very rare
2182 Description: kernel reports "badness in vsnprintf"
2183 Details : Reading from the "recovery_status" /proc file in small chunks
2184 may cause a negative length in lprocfs_obd_rd_recovery_status()
2185 call to vsnprintf() (which is otherwise harmless). Exit early
2186 if there is no more space in the output buffer.
2188 Severity : enhancement
2190 Description: clear OBD RPC statistics by writing to them
2191 Details : It is now possible to clear the OBD RPC statistics by writing
2192 to the "stats" file.
2197 Description: Client mtime is not the same on different clients after utimes
2198 Details : In some cases, the client was using the utimes() syscall on
2199 a file cached on another node. The clients now validate the
2200 ctime from the MDS + OSTs to determine which one is right.
2205 Description: Inability to activate failout mode
2206 Details : lconf script incorrectly assumed that in pythong string's numeric
2207 value is used in comparisons.
2210 Frequency : always with multiple stripes per file
2212 Description: Inefficient object allocation for mutli-stripe files
2213 Details : When selecting which OSTs to stripe files over, for files with
2214 a stripe count that divides evenly into the number of OSTs,
2215 the MDS is always picking the same starting OST for each file.
2216 Return the OST selection heuristic to the original design.
2221 Description: mount failures may take full timeout to return an error
2222 Details : Under some heavy load conditions it is possible that a
2223 failed mount can wait for the full obd_timeout interval,
2224 possibly several minutes, before reporting an error.
2225 Instead return an error as soon as the status is known.
2227 Frequency : quota enabled and large files being deleted
2229 Description: releasing more than 4GB of quota at once hangs OST
2230 Details : If a user deletes more than 4GB of files on a single OST it
2231 will cause the OST to spin in an infinite loop. Release
2232 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2237 Description: statfs data retrieved from /proc may be stale or zero
2238 Details : When reading per-device statfs data from /proc, in the
2239 {kbytes,files}_{total,free,avail} files, it may appear
2240 as zero or be out of date.
2243 Frequency : systems with MD RAID1 external journal devices
2245 Description: lconf's call to blkid is confused by RAID1 journal devices
2246 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2247 of returning all block devices that match the journal UUID.
2250 Frequency : always, for aggregate stripe size over 4GB
2252 Description: assertion fails when trying to use 4GB stripe size
2253 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2254 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2257 Frequency : always on ppc64
2259 Description: the first write on an ext3 filesystem with mballoc got stuck
2260 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2261 perform endianness conversion.
2263 ------------------------------------------------------------------------------
2265 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2267 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2268 INTEROPERATE with older versions automatically. Please read the
2269 user documentation before upgrading any part of a live system.
2270 * WARNING: Lustre networking configuration changes are required with
2271 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2274 * Support for kernels:
2275 2.6.9-22.0.2.EL (RHEL 4)
2276 2.6.5-7.244 (SLES 9)
2277 2.6.12.6 vanilla (kernel.org)
2280 Severity : enhancement
2281 Bugzilla : 7981/8208
2282 Description: Introduced Lustre Networking (LNET)
2283 Details : LNET is new networking infrastructure for Lustre, it includes
2284 a reorganized network configuration mode (see the user
2285 documentation for full details) as well as support for routing
2286 between different network fabrics. Lustre Networking Devices
2287 (LNDs) for the supported network fabrics have also been
2288 created for this new infrastructure.
2290 Severity : enhancement
2291 Description: Introduced Access control lists
2292 Details : clients can set ACLs on files and directories in order to have
2293 more fine-grained permissions than the standard Unix UGO+RWX.
2294 The MDS must be started with the "-o acl" mount option.
2296 Severity : enhancement
2297 Description: Introduced filesystem quotas
2298 Details : Administrators may now establish per-user quotas on the
2301 Severity : enhancement
2303 Description: Configuration change for the XT3
2304 The PTLLND is now used to run Lustre over Portals on the XT3
2305 The configure option(s) --with-cray-portals are no longer used.
2306 Rather --with-portals=<path-to-portals-includes> is used to
2307 enable building on the XT3. In addition to enable XT3 specific
2308 features the option --enable-cray-xt3 must be used.
2313 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2314 Details : A race between allocating a new llcd and re-getting the llcd_lock
2315 allowed another thread to grab newly-allocated llcd.
2317 Severity : enhancement
2319 Description: 2.6 OST async journal commit and locking fix to improve performance
2320 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2321 2.6 kernels are now async as they already were in 2.4 kernels so
2322 that they can commit concurrently with the network bulk transfer.
2323 For block-allocated files the filter allocation semaphore is held
2324 to avoid filesystem fragmentation during allocation. BKL lock
2325 removed for 2.6 xattr operations where it is no longer needed.
2330 Description: lconf incorrectly determined whether two IP networks could talk
2331 Details : In some more complicated routing and multiple-network
2332 configurations, lconf will avoid trying to make a network
2333 connection to a disjoint part of the IP space. It was doing the
2334 math incorrectly for one set of cases.
2339 Description: Fix for potential infinite loop processing records in an llog.
2340 Details : If an llog record is corrupted/zeroed, it is possible to loop
2341 forever in llog_process(). Validate the llog record length
2342 and skip the remainder of the block on error.
2345 Frequency : occasional (liblustre only)
2347 Description: liblustre could not open files whose last component is a symlink
2348 Details : sysio_path_walk() would incorrectly pass the open intent to
2349 intermediate path components.
2352 Frequency : rare (liblustre only with non-standard tuning)
2353 Bugzilla : 7201 (7350)
2354 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2355 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2356 after it has been released, only do this for referenced locks
2358 Severity : enhancement
2360 Description: specifying an (invalid) directory default stripe_size of -1
2361 would reset the directory default striping
2362 Details : stripe_size -1 was used internally to signal directory stripe
2363 removal, now use "all default" to signal dir stripe removal
2364 as a directory striping of "all default" is not useful
2367 Frequency : common for large clusters running liblustre clients
2369 Description: doing an ls when liblustre clients are running is slow
2370 Details : sending a glimpse AST to a liblustre client waits for every AST
2371 to time out, as liblustre clients will not respond. Since they
2372 cannot cache data we refresh the OST lock LVB from disk instead.
2374 Severity : enhancement
2376 Description: doing an ls at the same time as file IO can be slow
2377 Details : enqueue and other "small" requests can be blocked behind many
2378 large IO requests. Create a new OST IO portal for non-IO
2379 requests so they can be processed faster.
2382 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2384 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2385 created file with mode 000
2386 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2387 allow an ftruncate() as a non-root user to a file with mode 000.
2388 Lustre now respects this flag to disable mode checks when
2389 truncating a file owned by the user
2392 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2395 Description: Revoking locks from clients that went dead or catatonic might take
2397 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2398 cancellation of such locks instant on servers without waiting for
2399 any reply from clients. Clients drops these locks when cancel
2400 notification from server is received without replying.
2403 Frequency : liblustre-only, when liblustre client dies or becomes busy
2405 Description: Doing ls on Linux clients can take a long time with active
2407 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2408 granting such locks to it in the first place if possible. Locks
2409 are taken by proxy on the OST during the read or write and
2410 dropped immediately afterward. Add connect flags handling, do
2411 not grant locks to liblustre clients for glimpse ASTs.
2413 Severity : enhancement
2415 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2416 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2417 sequential reads and behaves badly with applications doing random
2418 reads. Improve it by reading ahead at least read region, and
2419 avoiding excessive large RPC for small reads.
2421 Severity : enhancement
2423 Description: Creating more than 1000 files for a single job may cause a load
2424 imbalance on the OSTs if there are also a large number of OSTs.
2425 Details : qos_prep_create() uses an OST index reseed value that is an
2426 even multiple of the number of available OSTs so that if the
2427 reseed happens in the middle of the object allocation it will
2428 still utilize the OSTs as uniformly as possible.
2433 Description: OST or MDS may oops in ping_evictor_main()
2434 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2435 but doesn't restart at beginning of obd_exports_timed list
2438 Severity : enhancement
2440 Description: improve by-nid export eviction on the MDS and OST
2441 Details : allow multiple exports with the same NID to be evicted at one
2442 time without re-searching the exports list.
2445 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2447 Description: MDS may oops in groups_free()
2448 Details : in rare race conditions a newly allocated group_info struct is
2449 freed again, and this can be NULL. The 2.4 compatibility code
2450 for groups_free() checked for a NULL pointer, but 2.6 did not.
2453 Frequency : common for liblustre clients doing little filesystem IO
2454 Bugzilla : 9352, 7313
2455 Description: server may evict liblustre clients accessing contended locks
2456 Details : if a client is granted a lock or receives a completion AST
2457 with a blocking AST already set it would not reply to the AST
2458 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2461 Frequency : lfs setstripe, only systems with more than 160 OSTs
2463 Description: unable to set striping with a starting offset beyond OST 160
2464 Details : llapi_create_file() incorrectly limited the starting stripe
2465 index to the maximum single-file stripe count.
2468 Frequency : LDAP users only
2470 Description: lconf did not handle in-kernel recovery with LDAP properly
2471 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2473 Severity : enhancement
2475 Description: bind OST threads to NUMA nodes to improve performance
2476 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2477 node and do their allocations there to localize memory access
2479 Severity : enhancement
2481 Description: llmount can determine client NID directly from Myrinet (GM)
2482 Details : the client NID code from gmnalnid was moved directly into
2483 llmount, removing the need to use this or specifying the
2484 client NID explicitly when mounting GM clients with zeroconf
2487 Frequency : if client is started with down MDS
2489 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2490 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2491 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2492 ptlrpc_import_delayed_req() because we didn't check if the
2493 request was interrupted, and we also didn't break out of the
2494 event loop if there was no timeout
2499 Description: data loss during non-page-aligned writes to a single file from
2500 both multiple nodes and multiple threads on one node at same time
2501 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2502 inconsistency led to false short-reads, that were cached and later
2503 used by ->prepare_write() to fill in partially written page,
2504 leading to data loss.
2507 Frequency : always, if lconf --abort_recovery used
2509 Description: lconf --abort_recovery fails with 'Operation not supported'
2510 Details : lconf was attempting to abort recovery on the MDT device and not
2513 Severity : enhancement
2515 Description: remove cleanup logs
2516 Details : replace lconf-generated cleanup logs with lustre internal
2517 cleanup routines. Eliminates the need for client-cleanup and
2520 Severity : enhancement
2522 Description: add support for EAs (user and system) on lustre filesystems
2523 Details : it is now possible to store extended attributes in the Lustre
2524 client filesystem, and with the user_xattr mount option it
2525 is possible to allow users to store EAs on their files also
2527 Severity : enhancement
2529 Description: Add possibility (config option) to show minimal available OST free
2531 Details : When compiled with --enable-mindf configure option, statfs(2)
2532 (and so, df) will return least minimal free space available from
2533 all OSTs as amount of free space on FS, instead of summary of
2534 free spaces of all OSTs.
2536 Severity : enhancement
2538 Description: do not expand extent locks acquired on OST-side
2539 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2540 by server: they are not cached anyway.
2543 Frequency : when mmap is used/binaries executed from Lustre
2545 Description: Unmmap pages before throwing them away from read cache.
2546 Details : llap_shrink cache now attempts to unmap pages before discarding
2547 them (if unmapping failed - do not discard). SLES9 kernel has
2548 extra checks that trigger if this unmapping is not done first.
2553 Description: lconf didn't resolve symlinks before checking to see whether a
2554 given mountpoint was already in use
2557 Frequency : when migrating failover services
2558 Bugzilla : 6395, 9514
2559 Description: When migrating a subset of services from a node (e.g. failback
2560 from a failover service node) the remaining services would
2561 time out and evict clients.
2562 Details : lconf --force (implied by --failover) sets the global obd_timeout
2563 to 5 seconds in order to quickly disconnect, but this caused
2564 other RPCs to time out too quickly. Do not change the global
2565 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2567 Severity : enhancement
2568 Frequency : if MDS is started with down OST
2569 Bugzilla : 9439,5706
2570 Description: Allow startup/shutdown of an MDS without depending on the
2571 availability of the OSTs.
2572 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2573 Add appropriate locking and lov-osc refcounts for safe
2574 cleaning. Add osc abort_inflight calls in case the
2575 synchronize never started.
2578 Frequency : occasional (Cray XT3 only)
2580 Description: root not authorized to access files in CRAY_PORTALS environment
2581 Details : The client process capabilities were not honoured on the MDS in
2582 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2583 been accessed by an authorized user then root was able to access
2584 the file on the local client also. The root user capabilities
2585 are now allowed on the MDS, as this environment has secure UID.
2588 Frequency : occasional
2590 Description: ldiskfs "too long searching" message happens too often
2591 Details : A debugging message (otherwise harmless) prints too often on
2592 the OST console. This has been reduced to only happen when
2593 there are fragmentation problems on the filesystem.
2598 Description: Division by zero in statfs when all OSCs are inactive
2599 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2600 lov_statfs divides by value returned by lov_get_stripecnt().
2604 Bugzilla : 9489, 3273
2605 Description: First write from each client to each OST was only 4kB in size,
2606 to initialize client writeback cache, which caused sub-optimal
2607 RPCs and poor layout on disk for the first writen file.
2608 Details : Clients now request an initial cache grant at (re)connect time
2609 and so that they can start streaming writes to the cache right
2610 away and always do full-sized RPCs if there is enough data.
2611 If the OST is rebooted the client also re-establishes its grant
2612 so that client cached writes will be honoured under the grant.
2617 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2618 Details : Now I/O RPCs go to different portal number and (presumably) fast
2619 lock requests (and glimses) and other RPCs get their own service
2620 threads pool that should be able to service those RPCs
2623 Severity : enhancement
2625 Description: Ability to exchange lustre version between client and servers and
2626 issue warnings at client side if client is too old. Also for
2627 liblustre clients there is ability to refuse connection of too old
2629 Details : New 'version' field is added to connect data structure that is
2630 filled with version info. That info is later checked by server and
2634 Frequency : rare, liblustre only.
2635 Bugzilla : 9296, 9581
2636 Description: Two simultaneous writes from liblustre at offset within same page
2637 might proceed at the same time overwriting eachother with stale
2639 Details : I/O lock withing llu_file_prwv was released too early, before data
2640 actually was hitting the wire. Extended lock-holding time until
2641 server acknowledges receiving data.
2644 Frequency : extremely rare. Never observed in practice.
2646 Description: avoid generating lustre_handle cookie of 0.
2647 Details : class_handle_hash() generates handle cookies by incrementing
2648 global counter, and can hit 0 occasionaly (this is unlikely, but
2649 not impossible, because initial value of cookie counter is
2650 selected randonly). Value of 0 is used as a sentinel meaning
2651 "unassigned handle" --- avoid it. Also coalesce two critical
2652 sections in this function into one.
2654 Severity : enhancement
2656 Description: allow liblustre clients to delegate truncate locking to OST
2657 Details : To avoid overhead of locking, liblustre client instructs OST to
2658 take extent lock in ost_punch() on client's behalf. New connection
2659 flag is added to handle backward compatibility.
2661 Severity : enhancement
2662 Bugzilla : 4928, 7341, 9758
2663 Description: allow number of OST service threads to be specified
2664 Details : a module parameter allows the number of OST service threads
2665 to be specified via "options ost ost_num_threads={N}" in the
2666 OSS's /etc/modules.conf or /etc/modprobe.conf.
2670 Bugzilla : 6146, 9635, 9895
2671 Description: servers crash with bad pointer in target_handle_connect()
2672 Details : In rare cases when a client is reconnecting it was possible that
2673 the connection request was the last reference for that export.
2674 We would temporarily drop the export reference and get a new
2675 one, but this may have been the last reference and the export
2676 was just destroyed. Get new reference before dropping old one.
2678 Severity : enhancement
2679 Frequency : if client is started with failover MDS
2681 Description: Allow multiple MDS hostnames in the mount command
2682 Details : Try to read the configuration from all specified MDS
2683 hostnames during a client mount in case the "primary"
2686 Severity : enhancement
2688 Description: Stop sending data to evicted clients as soon as possible.
2689 Details : Check if the client we are about to send or are sending data to
2690 was evicted already. (Check is done every second of waiting,
2691 for which l_wait_event interface was extended to allow checking
2692 of exit condition at specified intervals).
2695 Frequency : rare, normally only when NFS exporting is done from client
2697 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2699 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2702 Frequency : always, if srand() is called before liblustre initialization
2704 Description: Liblustre uses system PRNG disturbing its usage by user application
2705 Details : Introduce internal to lustre fast and high-quality PRNG for
2706 lustre usage and make liblustre and some other places in generic
2707 lustre code to use it.
2709 Severity : enhancement
2710 Bugzilla : 9477, 9557, 9870
2711 Description: Verify that the MDS configuration logs are updated when xml is
2712 Details : Check if the .xml configuration logs are newer than the config
2713 logs stored on the MDS and report an error if this is the case.
2714 Request --write-conf, or allow starting with --old_conf.
2716 Severity : enhancement
2718 Description: Handle symlinks in the path when checking if Lustre is mounted.
2719 Details : Resolve intermediate symlinks when checking if a client has
2720 mounted a filesystem to avoid duplicate client mounts.
2725 Description: lconf can hit an error exception but still return success.
2726 Details : The lconf command catches the Command error exception at the top
2727 level script context and will exit with the associated exit
2728 status, but doesn't ensure that this exit status is non-zero.
2733 Description: failure of ptlrpc thread startup can cause oops
2734 Details : Starting a ptlrpc service thread can fail if there are a large
2735 number of threads or the server memory is very fragmented.
2736 Handle this without oopsing.
2739 Frequency : always, only if liblustre and non-default acceptor port was used
2741 Description: liblustre cannot connect to servers with non-default acceptor port
2742 Details : tcpnal_set_default_params() was not called and was therefore
2743 ignoring the environment varaible TCPNAL_PORT, as well as other
2744 TCPNAL_ environment variables
2749 Description: two objects could be created on the same OST for a single file
2750 Details : If an OST is down, in some cases it was possible to create two
2751 objects on a single OST for a single file. No problems other
2752 than potential performance impact and spurious error messages.
2756 Bugzilla : 5681, 9562
2757 Description: Client may oops in ll_unhash_aliases
2758 Details : Client dcache may become inconsistent in race condition.
2759 In some cases "getcwd" can fail if the current directory is
2765 Description: Inode refcounting problems in NFS export code
2766 Details : link_raw functions used to call d_instantiate without obtaining
2767 extra inode reference first.
2771 Bugzilla : 9942, 9903
2772 Description: Referencing freed requests leading to crash, memleaks with NFS.
2773 Details : We used to require that call to ll_revalidate_it was always
2774 followed by ll_lookup_it. Also with revalidate_special() it is
2775 possible to call ll_revalidate_it() twice for the same dentry
2776 even if first occurence returned success. This fix changes semantic
2777 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2778 reference on a request referred from the intent.
2779 ll_intent_release() then releases such a request.
2782 Frequency : rare, normally benchmark loads only
2784 Description: unlinked inodes were kept in memory on the client
2785 Details : If a client is repeatedly creating and unlinking files it
2786 can accumulate a lot of stale inodes in the inode slab cache.
2787 If there is no other client load running this can cause the
2788 client node to run out of memory. Instead flush old inodes
2789 from client cache that have the same inode number as a new inode.
2792 Frequency : SLES9 2.6.5 kernel and long filenames only
2793 Bugzilla : 9969, 10379
2794 Description: utime reports stale NFS file handle
2795 Details : SLES9 uses out-of-dentry names in some cases, which confused
2796 the lustre dentry revalidation. Change it to always use the
2800 Frequency : rare, unless heavy write-truncate concurrency is continuous
2801 Bugzilla : 4180, 6984, 7171, 9963, 9331
2802 Description: OST becomes very slow and/or deadlocked during object unlink
2803 Details : filter_destroy() was holding onto the parent directory lock
2804 while truncating+unlinking objects. For very large objects this
2805 may block other threads for a long time and slow overall OST
2806 responsiveness. It may also be possible to get a lock ordering
2807 deadlock in this case, or run out of journal credits because of
2808 the combined truncate+unlink. Solution is to do object truncate
2809 first in one transaction without parent lock, and then do the
2810 final unlink in a new transaction with the parent lock. This
2811 reduces the lock hold time dramatically.
2814 Frequency : rare, 2.4 kernels only
2816 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2817 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2818 before unmounting their filesystem in order to avoid deadlock.
2819 The kernel_locked() function in 2.4 kernels only checks whether
2820 the kernel lock is held, not whether it is this process that is
2821 holding it as 2.6 kernels do.
2826 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2827 Details : The client ptlrpc code may be trying to reconnect to a down
2828 server before a previous connection attempt has timed out.
2829 Increase the reconnect interval to be longer than the connection
2830 timeout interval to avoid sending duplicate connections to
2834 Frequency : echo_client brw_test command
2836 Description: fix echo_client to work with OST preallocated code
2837 Details : OST preallocation code (5137) didn't take echo_client IO path
2838 into account: echo_client calls filter methods outside of any
2839 OST thread and, hence, there is no per-thread preallocated
2840 pages and buffers to use. Solution: hijack pga pages for IO. As
2841 a byproduct, this avoids unnecessary data copying.
2845 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2846 Description: Client can oops in mdc_commit_close() after open replay
2847 Details : It was possible for the MDS to return an open request with no
2848 transaction number in mds_finish_transno() if the client was
2849 evicted, but without actually returning an error. Clients
2850 would later try to replay that open and may trip an assertion
2851 Simplify the client close codepath, and always return an error
2852 from the MDS in case the open is not successful.
2855 Frequency : rare, 2.6 OSTs only
2857 Description: OST may deadlock under high load on fragmented files
2858 Details : If there was a heavy load and highly-fragmented OST filesystems
2859 it was possible to have all the OST threads deadlock waiting on
2860 allocation of biovecs, because the biovecs were not released
2861 until the entire RPC IO was completed. Instead, release biovecs
2862 as soon as they are complete to ensure forward IO progress.
2864 Severity : enhancement
2866 Description: Support for specifying external journal device at mount
2867 Details : If an OST or MDS device is formatted with an external journal
2868 device, this device major/minor is stored in the ext3 superblock
2869 and may not be valid for failover. Allow detecting and
2870 specifying the external journal at mount time.
2875 Description: Mounting an MDS with pending unlinked files may cause oops
2876 Details : target_finish_recovery() calls mds_postrecov() which returned
2877 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2878 considers this an error and immediately begins cleaning up the
2879 lov, just after starting the mds_lov process
2881 Severity : enhancement
2883 Description: Implement 'lfs df' to report actual free space on per-OST basis
2884 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2885 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2886 usage of objects; '-h' to report in human readable format.
2888 ------------------------------------------------------------------------------
2890 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2897 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2898 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2899 at startup during orphan unlinking. Instead, skip buddy update
2900 before setup, as it will be regenerated after recovery is complete.
2905 Description: If an OST is inactive, its locks might reference stale inodes.
2906 Details : lov_change_cbdata() must iterate over all namespaces, even if
2907 they are inactive to clear inode references from the lock.
2909 Severity : enhancement
2910 Frequency : occasional, if non-standard max_dirty_mb used
2912 Description: Client will block write RPCs if not enough grant
2913 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2914 then the client will block writes while waiting for another RPC
2915 to complete instead of consuming its dirty limit. With change
2916 we get improved performance when max_dirty_mb is small.
2918 Severity : enhancement
2919 Bugzilla : 3389, 6253
2920 Description: Add support for supplementary groups on the MDS.
2921 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2922 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2923 lookups for user supplementary groups into a cache.
2927 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2928 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2929 for liblustre O_EXCL is enforced by MDS.
2933 Description: readdir never returns NULL in liblustre.
2934 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2935 dentry in struct dirent.
2939 Frequency : liblustre only, depends on application IO pattern
2940 Description: liblustre clients evicted if not contacting servers
2941 Details : Don't put liblustre clients into the ping_evictor list, so
2942 they will not be evicted by the pinger ever.
2944 Severity : enhancement
2946 Description: Add ability to evict clients by NID from MDS.
2947 Details : By echoing "nid:$NID" string into
2948 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2949 $NID would be instantly evicted from this MDS and from all active
2950 OSTs connected to it.
2954 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2955 Details : lookup_it_finish() used to query file size from OSTs that was not
2960 Description: service threads change working directory to that of init
2961 Details : Starting lustre service threads may pin the working directory
2962 of the parent thread, making that filesystem busy. Threads
2963 now change to the working directory of init to avoid this.
2967 Frequency : during shutdown only
2968 Description: shutdown with a failed MDS or OST can cause unmount to hang
2969 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2974 Frequency : 2.6 only
2975 Description: chmod/chown may include an extra supplementary group
2976 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2977 supplementary group and if none is specified this is used.
2980 Bugzilla : 5479 (6816)
2982 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2983 Details : If lookup is in progress on a file that is unlinked we might try
2984 to revalidate the inode and fail in revalidate after lookup is
2985 complete and ll_file_open() enqueues the open again but
2986 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2989 Frequency : always, if lconf --abort_recovery used
2991 Description: lconf --abort_recovery fails with 'Operation not supported'
2992 Details : lconf was attempting to abort recovery on the MDT device and not
2995 ------------------------------------------------------------------------------
2997 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3002 Frequency : rare (only unsupported configurations with a node running as an
3004 Bugzilla : 6514, 5137
3005 Description: Mounting a Lustre file system on a node running as an OST could
3007 Details : OSTs now preallocates memory needed to write out data at
3008 startup, instead of when needed, to avoid having to
3009 allocate memory in possibly low memory situations.
3010 Specifically, if the file system is mounted on on OST,
3011 memory pressure could force it to try to write out data,
3012 which it needed to allocate memory to do. Due to the low
3013 memory, it would be unable to do so and the node would
3014 become unresponsive.
3016 Severity : enhancement
3018 Description: Addition of lconf --service command line option
3019 Details : lconf now accepts a '--service <arg>' option, which is
3020 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3022 Severity : enhancement
3024 Description: Failover mode is now the default for OSTs.
3025 Details : By default, OSTs will now run in failover mode. To return to
3026 the old behaviour, add '--failout' to the lmc line for OSTs.
3028 Severity : enhancement
3030 Description: Health checks are now provided for MDS and OSTs
3031 Details : Additional detailed health check information on MSD and OSTs
3032 is now provided through the procfs health_check value.
3035 Frequency : occasional, depends on IO load
3037 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3038 after numerous create/delete cycles
3039 Details : The ext3 inode allocation policy would not allocate new inodes
3040 very well on the OSTs because there are no new directories
3041 being created. Instead we look for groups with free space if
3042 the parent directories are nearly full.
3047 Description: Network or server problems during mount may cause partially
3048 mounted clients instead of returning an error.
3049 Details : The config llog parsing code may overwrite the error return
3050 code during mount error handling, returning success instead
3056 Description: MDS can fail to allocate large reply buffers
3057 Details : After long uptimes the MDS can fail to allocate large reply
3058 buffers (e.g. zconf client mount config records) due to memory
3059 fragmentation or consumption by the buffer cache. Preallocate
3060 some large reply buffers so that these replies can be sent even
3061 under memory pressure.
3065 Frequency : rare (liblustre)
3066 Description: fsx running with liblustre complained that using truncate() to
3067 extend the file doesn't work. This patch corrects that issue.
3068 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3069 ATTR_SIZE and lsm use in llu_setattr_raw.
3073 Frequency : rare, only 2.6 kernels
3074 Description: Unusual file access patterns on the MDS may result in inode
3075 data being lost in very rare circumstances.
3076 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3077 code in the 2.6 kernel caused the inode and/or EA data not to
3078 be read from disk, causing single-file corruption.
3082 Frequency : rare, only 2.6 filesystems using extents
3083 Description: Heavy concurrent write and delete load may cause data corruption.
3084 Details : It was possible under high-load situations to have an extent
3085 metadata block in the block device cache from a just-unlinked
3086 file overwrite a newly-allocated data block. We now unmap any
3087 metadata buffers that alias just-allocated data blocks.
3091 Frequency : filesystems with default stripe_count larger than 77
3092 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3093 Details : lconf specifies an inode size of 4096 bytes when the default
3094 stripe_count is larger than 77. This conflicts with the default
3095 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3096 this case to avoid pinning too much memory for large EAs.
3098 ------------------------------------------------------------------------------
3100 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3105 Frequency : rare (extremely heavy IO load with hundreds of clients)
3107 Description: Client is evicted, gets IO error writing to file
3108 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3109 caused clients to be evicted for AST timeouts. The fixes in
3110 bug 5192 mean we no longer need to have such short AST timeouts
3111 so ldlm_timeout has been increased.
3114 Frequency : occasional during --force or --failover shutdown under load
3115 Bugzilla : 5949, 4834
3116 Description: Server oops/LBUG if stopped with --force or --failover under load
3117 Details : a collection of import/export refcount and cleanup ordering
3118 issues fixed for safer force cleanup
3121 Frequency : only filesystems larger than 120 OSTs
3122 Bugzilla : 5990, 6223
3123 Description: lfs getstripe would oops on a very large filesystem
3124 Details : lov_getconfig used kfree on vmalloc'd memory
3127 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3128 Bugzilla : 6242, 6243
3129 Description: reading from files that had been truncated to a non-zero size
3130 but never opened returned no data
3131 Details : ll_file_read() reads zeros from no-object files to EOF
3136 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3137 Details : There's little harm in aborting MDS/OSS recovery and letting it
3138 try again, so I removed the LASSERT and return an error instead.
3140 Severity : enhancement
3142 Description: New debugging infrastructure for tracking down data corruption
3143 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3144 (b) cover more of the client-side code path, and (c) try to narrow
3145 down where problems occurred
3149 Bugzilla : 3819, 4364, 4397, 6313
3150 Description: Racing close and eviction MDS could cause assertion in mds_close
3151 Details : It was possible to get multiple mfd references during close and
3152 client eviction, leading to one thread referencing a freed mfd.
3154 Severity: : enhancement
3155 Bugzilla : 3262, 6359
3156 Description: Attempts to reconnect to servers are now more aggressive.
3157 Details : This builds on the enhanced upcall-less recovery that was added
3158 in 1.4.2. When trying to reconnect to servers, clients will
3159 now try each server in the failover group every 10 seconds. By
3160 default, clients would previously try one server every 25 seconds.
3165 Description: After recovery, certain operations trigger a failed
3166 assertion on a client.
3167 Details : Failing over an mds, using lconf -d --failover, while a
3168 client was doing a readdir() call would cause the client to
3169 LBUG after recovery completed and the readdir() was resent.
3171 Severity : enhancement
3173 Description: Default groups are now added by lconf
3174 Details : You can now run lconf --group <servicename> without having to
3175 manually add groups with lmc.
3178 Frequency : occasional
3180 Description: Nodes with an elan id of 0 trigger a failed assertion
3183 Frequency : always when accessing e.g. tty/console device nodes
3185 Description: tty and some other devices nodes cannot be used on lustre
3186 Details : file's private_data field is used by device data and lustre
3187 values in there got lost. New field was added to struct file to
3188 store fs-specific private data.
3191 Frequency : when exporting Lustre via NFS
3193 Description: NFSD failed occasionally when looking up a path component
3194 Details : NFSD is looking up ".." which was broken in ext3 directories
3195 that had grown large enough to become hashed.
3198 Frequency : Clusters with multiple interfaces not on the same subnet
3200 Description: Nodes will repeatedly try to reconnect to an interface which it
3201 cannot reach and report an error to the log.
3202 Details : Extra peer list entries will be created by lconf with some peers
3203 unreachable. lconf now validates the peer before adding it.
3206 Frequency : Only if a default stripe is set on the filesystem root.
3208 Description: Setting a default stripe on the filesystem root prevented the
3209 filesystem from being remounted.
3210 Details : The client was sending extra request flags in the root getattr
3211 request and did not allocate a reply buffer for the dir EA.
3214 Frequency : occasional, higher if lots of files are accessed by one client
3215 Bugzilla : 6159, 6097
3216 Description: Client trips assertion regarding lsm mismatch/magic
3217 Details : While revalidating inodes the VFS looks up inodes with ifind()
3218 and in rare cases can find an inode that is being freed.
3219 The ll_test_inode() code will free the lsm during ifind()
3220 when it finds an existing inode and then the VFS later attaches
3221 this free lsm to a new inode.
3225 Bugzilla : 6422, 7030
3226 Description: MDS deadlock between mkdir and client eviction
3227 Details : Creating a new file via mkdir or mknod (starting a transaction
3228 and getting the ns lock) can deadlock with client eviction
3229 (gets ns lock and trying to finish a synchronous transaction).
3232 Frequency : occasional
3233 Description: While starting a server, the fsfilt_ext3 module could not be
3235 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3236 kernels. Previously, lconf would still use the ext3 name
3237 when trying to load modules. Now, it will correctly use
3238 ext3 on 2.4 and ldiskfs on 2.6.
3240 Severity : enhancement
3241 Description: The default stripe count has been changed to 1
3242 Details : The interpretation of the default stripe count (0, to lfs
3243 or lmc) has been changed to mean striping across a single
3244 OST, rather than all available. For general usage we have
3245 found a stripe count of 1 or 2 works best.
3247 Severity : enhancement
3248 Description: Add support for compiling against Cray portals.
3249 Details : Conditional compiling for some areas that are different
3253 Frequency : occasional
3254 Bugzilla : 6409, 6834
3255 Description: Creating files with an explicit stripe count may lead to
3256 a failed assertion on the MDS
3257 Details : If some OSTs are full or unavailable, creating files may
3258 trigger a failed assertion on the MDS. Now, Lustre will
3259 try to use other servers or return an error to the
3263 Frequency : occasional
3265 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3266 caused lock timeout during readahead (since 1.4.2).
3267 Details : Processes doing readahead might match a lock that hasn't been
3268 granted yet if there are overlapping and conflicting lock
3269 requests. The readahead process waits on ungranted lock
3270 (original lock is CBPENDING), while OST waits for that process
3271 to cancel CBPENDING read lock and eventually evicts client.
3273 Severity : enhancement
3275 Description: Initial enabling of flock support for clients
3276 Details : Implements fcntl advisory locking and file status functions.
3277 This feature is provided as an optional mount flag (default
3278 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3279 locking are implemented yet, and those that are are not guaranteed
3280 to be completely correct in production environments.
3281 mount -t lustre -o [flock|noflock] ...
3284 Frequency : occasional
3286 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3287 assertion in the ext3 JBD (journaling) layer.
3288 Details : The b_committed_data struct is protected by the big kernel lock
3289 in 2.4 kernels, serializing journal_commit_transaction() and
3290 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3291 access to this struct. In 2.6 kernels there is finer grained
3292 locking to improve SMP performance of the JBD layer.
3296 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3301 Description: merge and cleanup kernel patches.
3304 -----------------------------------------------------------------------------
3306 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3308 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3309 versions of Lustre. Please update all servers and clients to
3310 version 1.4.2 or later at the same time. You must also run
3311 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3312 to update the configuration logs.
3314 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3315 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3316 - swab llog records of type '0' so we get proper header size/idx (5861)
3317 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3318 - fix rename of one directory over another leaking an inode (5953)
3319 - avoid SetPageDirty on 2.6 (5981)
3320 - don't re-add just-being-destroyed locks to the waiting list (5653)
3321 - when creating new directories, inherit the parent's custom
3322 striping settings if present parent (3048)
3323 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3324 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3325 - don't hold client locks on temporary worklist from l_lru (5666)
3326 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3327 - automatically evict dead clients (5921)
3328 - Update file size properly in create+truncate+fstat case (6196)
3329 - Do not unhash mountpoint dentries, do not allow removal of
3331 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3332 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3333 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3334 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3335 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3336 - Correctly handle reads of files with no objects (6243)
3337 - lctl recover will also mark a device active if deactivate used (5933)
3339 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3340 - allow --write-conf on an MDS with different nettype than client (5619)
3341 - don't write config llogs to MDS for mounts not from that MDS (5617)
3342 - lconf should create multiple TCP connections from a client (5201)
3343 - init scripts are now turned off by default; run chkconfig --on
3344 lustre and chkconfig --on lustrefs to use them
3345 - upcalls are no longer needed for clients to recover to failover
3347 - add --abort-recovery option to lconf to abort recovery on device
3349 - add support for an arbitrary number of OSTs (3026)
3350 - Quota support protocol changes.
3351 - forward compatibility changes to wire structs (6007)
3352 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3353 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3354 - improved "lustre status" script
3355 - initialize blocksize for non-regular files (6062)
3356 - added --disable-server and --disable-client configure options (5782)
3357 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3358 - Vanilla 2.4.29 support
3359 - increase maximum number of obd devices to 520 (6242)
3360 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3361 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3362 SLES 9 2.6.5 kernels
3363 - limit stripes per file to 160 (the maximum EA size) (6093)
3365 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3368 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3369 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3370 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3371 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3372 - small ext3 extents cleanups and fixes (5733)
3373 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3374 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3375 - avoid needless client->OST connect, fix handle mismatch (5317)
3376 - fix DLM error path that led to out-of-sync client, long delays (5779)
3377 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3378 - fix several locking issues related to i_size (5492,5624,5654,5672)
3379 - don't move pending lock onto export if it is already evicted (5683)
3380 - fix kernel oops when creating .foo in unlinked directory (5548)
3381 - fix deadlock in obdfilter statistics vs. object create (5811)
3382 - use time_{before,after} to avoid timer jiffies wrap (5882)
3383 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3384 - Do not leak request if server was not able to process it (5154)
3385 - If mds_open unable to find parent dir, make that negative lookup(5154)
3386 - don't create new directories with extent-mapping (5909, 5936)
3388 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3389 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3390 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3391 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3392 - support for RHEL4 kernel on i686 (5773)
3393 - provide error messages when incompatible logs are encountered (5898)
3395 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3396 * version 1.4.0.10 (1.4.1 release candidate 1)
3398 - don't keep a lock reference when lock is not granted (4238)
3399 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3400 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3401 - drop import inflight refcount on signal_completed_replay error (5255)
3402 - unlock page after async write error during send (3677)
3403 - handle missing objects in filter_preprw_read properly (5265)
3404 - no transno return for symlink open, don't save no-trasno open (3440)
3405 - don't try to complete elan receive that already failed (4012)
3406 - free RPC server reply state on error (5406)
3407 - clean up thread from ptlrpc_start_thread() on error (5160)
3408 - readahead could read extra page into cache that wasn't ejected (5388)
3409 - prevent races in class_attach/setup/cleanup/detach (5260)
3410 - don't dereference de->d_inode after l_dput of de (5458)
3411 - use "int" for stripe value returned from lock_to_stripe (5544)
3412 - mballoc allocation and error-checking fixes in 2.6 (5504)
3413 - block device patches to fix I/O request sizes in 2.6 (5482)
3414 - look up hostnames for IB nals (5602)
3415 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3416 - don't start multiple acceptors for the same port (5277)
3417 - fix incorrect LASSERT in mds_getattr_name (5635)
3418 - export a proc file for general "ping" checking (5628)
3419 - fix "lfs check" to not block when the MDS is down (5628)
3421 - service request history (4965)
3422 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3423 - create an "evict_client" /proc entry on OSTs, like the MDS has
3424 - fix mount usage message, return errors per mount(8) (5168)
3425 - change grep [] to grep "[]" in tests so they work in more UMLs
3426 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3427 - remove ancient LOV_MAGIC_V0 EA support (5047)
3428 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3429 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3430 - allow readahead allocations to fail when low on memory (5383)
3431 - mmap locking landed again, after considerable improvement (2828)
3432 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3434 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3437 - send OST transaction number in read/write reply to free req (4966)
3438 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3439 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3440 - the watchdog thread now runs as interruptible (5246)
3441 - flock/lockf fixes (but it's still disabled, pending 5135)
3442 - don't use EXT3 constants in llite code (5094)
3443 - memory shortage at startup could cause assertion (5176)
3445 - reorganization of lov code
3446 - single portals codebase
3448 - add extents/mballoc support (5025)
3449 - direct I/O reads in the obdfilter (4048)
3450 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3452 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3455 - send OST transaction number in read/write reply to free req (4966)
3456 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3457 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3458 - the watchdog thread now runs as interruptible (5246)
3459 - handle missing objects in filter_preprw_read properly (5265)
3460 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3461 - drop import inflight refcount on signal_completed_replay error (5255)
3462 - unlock page after async write error during send (3677)
3463 - return original error code on reconstructed replies (3761)
3464 - no transno return for symlink open, don't save no-trasno open (3440)
3466 - add pid to ldlm debugging output (4922)
3467 - bump the watchdog timeouts -- we can't handle 30sec yet
3468 - extra debugging for orphan dentry/inode bug (5259)
3470 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3473 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3474 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3475 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3476 - fix NULL dereference in /proc/sys/portals/routes (4827)
3477 - allow failed mdc_close() operations to be interrupted (4561)
3478 - stop precreate on OST before MDS would time out on it (4778)
3479 - don't send partial-page writes before EOF from client (4410)
3480 - discard client grant for sub-page writes on large-page clients (4520)
3481 - don't free dentries not owned by NFS code, check generation (4806)
3482 - fix lsm leak if mds_create_objects() fails (4801)
3483 - limit debug_daemon file size, always print CERROR messages (4789)
3484 - use transno after validating reply (3892)
3485 - process timed out requests if import state changes (3754)
3486 - update mtime on OST during writes, return in glimpse (4829)
3487 - add mkfsoptions to LDAP (4679)
3488 - use ->max_readahead method instead of zapping global ra (5039)
3489 - don't interrupt __l_wait_event() during strace
3491 - add software watchdogs to catch hung threads quickly (4941)
3492 - make lustrefs init script start after nfs is mounted
3493 - fix CWARN/ERROR duplication (4930)
3494 - return async write errors to application if possible (2248)
3495 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3496 - print NAL number in %x format (4645)
3497 - update barely-supported suse-2.4.21-171 series (4842)
3498 - support for sles 9 %post scripts
3499 - support for building 2.6 kernel-source packages
3500 - support for sles km_* packages
3502 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3505 - ignore -ENOENT errors in osc_destroy (3639)
3506 - notify osc create thread that OSC is being cleaned up (4600)
3507 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3508 - reconstruct ost_handle() like mds_handle() (4657)
3509 - create a new thread to do import eviction to avoid deadlock (3969)
3510 - let lconf resolve symlinked-to devices (4629)
3511 - don't unlink "objects" from directory with default EA (4554)
3512 - hold socknal file ref over connect in case target is down (4394)
3513 - allow more than 32000 subdirectories in a single directory (3244)
3514 - fix blocks count for O_DIRECT writes (3751)
3515 - OST returns ENOSPC from object create when no space left (4539)
3516 - don't send truncate RPC if file size isn't changing (4410)
3517 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3518 - bind to privileged port in socknal and tcpnal (3689)
3520 - rate limit CERROR/CWARN console message to avoid overload (4519)
3521 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3522 - basic mmap support (3918)
3523 - kernel patch series update from b1_4 (4711)
3525 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3528 - avoid crash during MDS cleanup with OST shut down (2775)
3529 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3530 - don't use bad inodes on the MDS (3744)
3531 - dynamic object preallocation to improve recovery speed (4236)
3532 - don't hold spinlock over lock dumping or change debug flags (4401)
3533 - don't zero obd_dev when it is force cleaned (3651)
3534 - print grants to console if they go negative (4431)
3535 - "lctl deactivate" will stop automatic recovery attempts (3406)
3536 - look for existing locks in ldlm_handle_enqueue() (3764)
3537 - don't resolve lock handle twice in recovery avoiding race (4401)
3538 - revalidate should check working dir is a directory (4134)
3540 - don't always mark "slow" obdfilter messages as errors (4418)
3542 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3545 - don't close LustreDB during write_conf until it is done (3860)
3546 - fix typo in lconf for_each_profile (3821)
3547 - allow dumping logs from multiple threads at one time (3820)
3548 - don't allow multiple threads in OSC recovery (3812)
3549 - fix debug_size parameters (3864)
3550 - fix mds_postrecov to initialize import for llog ctxt (3121)
3551 - replace config semaphore with spinlock (3306)
3552 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3553 - don't allow enqueue to complete on a destroyed export (3822)
3554 - down write_lock before checking llog header bitmap (3825)
3555 - recover from lock replay timeout (3764)
3556 - up llog sem before sending rpc (3652)
3557 - reduce ns lock hold times when setting kms (3267)
3558 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3559 - fix NULL deref and obd_dev leak on setup error (3312)
3560 - replace some LBUG about llog ops with error handling (3841)
3561 - don't match INVALID dentries from d_lookup and spin (3784)
3562 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3563 - fix invalid assertion in ptlrpc_set_wait (3880)
3565 - add libwrap support for the TCP acceptor (3996)
3566 - add /proc/sys/portals/routes for non-root route listing (3994)
3567 - allow setting MDS UUID in .xml (2580)
3568 - print the stack of a process that LBUGs (4228)
3570 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3573 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3574 - make sure to unset replay flag from failed open requests (3440)
3575 - if default stripe count is 0, use OST count for inode size (3636)
3576 - update parent mtime/ctime on client for create/unlink (2611)
3577 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3578 - free recovery state on server during a forced cleanup (3571)
3579 - unregister_reply for resent reqs (3063)
3580 - loop back devices mounting and status check on 2.6 (3563)
3581 - fix resource-creation race that can provoke i_size == 0 (3513)
3582 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3583 - more debugging for page-accounting assertion (3746)
3584 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3585 - avoid deadlock after precreation failure (3758)
3586 - fix race and lock order deadlock in orphan handling (3450, 3750)
3587 - add validity checks when grabbing inodes from l_ast_data (3599)
3589 - add /proc/.../recovery_status to obdfilter (3428)
3590 - lightweight CDEBUG infrastructure, debug daemon (3668)
3591 - change default OSC RPC parameters to be better on small clusters
3592 - turn off OST read cache for files smaller than 32MB
3593 - install man pages and include them in rpms (3100)
3594 - add new init script for (un)mounting lustre filesystems (2593)
3595 - run chkconfig in %post for init scripts (3701)
3596 - drop scimac NAL (unmaintained)
3598 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3601 - clean kiobufs before and after use (3485)
3602 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3603 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3604 - change init script to fail more clearly if not run as root (1528)
3605 - allow clients to reconnect during replay (1742)
3606 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3607 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3608 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3609 - fix cond_resched (3554)
3610 - only evict unfinished clients after recovery (3515)
3611 - allow bulk resend, prevent data loss (3570)
3612 - dynamic ptlrpc request buffer allocation (2102)
3613 - don't allow unlinking open directory if it isn't empty (2904)
3614 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3615 - remove extraneous obd dereference causing LASSERT failure (3334)
3616 - don't use get_cycles() when creating temp. files on the mds (3156)
3617 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3618 - handle EEXIST for set-stripe, set proper directory name (3336)
3620 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3621 - fix ksocknal_fmb_callback() error messages (2918)
3623 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3626 - don't copy lvb into (possibly NULL) reply on error (2983)
3627 - don't deref dentry after dput, don't free lvb on error (2922)
3628 - use the kms to determine writeback rpc length (2947)
3629 - increment oti_logcookies when osc is inactive (2948)
3630 - update client's i_blocks count via lvb messages (2543)
3631 - handle intent open/close of special files properly (1557)
3632 - mount MDS with errors=remount-ro, like obdfilter (2009)
3633 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3634 - don't use cancelling-locks' kms values (2947)
3635 - use highest lock extent for kms, not last one (2925)
3636 - don't dereference ERR_PTR() dentry in error handling path (3107)
3637 - fix thread race in portals_debug_dumplog() (3122)
3638 - create lprocfs device entries at setup instead of at attach (1519)
3639 - common AST error handler, don't evict client on completion race (3145)
3640 - zero nameidata in detach_mnt in 2.6 (3118)
3641 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3642 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3643 - initialize RPC timeout timer earlier for 2.6 (3219)
3644 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3645 - print nal/nid for unknown nid (3258)
3646 - additional checks for oscc recovery before doing precreate (3284)
3647 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3648 - don't crash in mdc_close for bad permissions on open (3285)
3649 - zero i_rdev for non-device files (3147)
3650 - clear page->private before handing to FS, better assertion (3119)
3651 - tune the read pipeline (3236)
3652 - fix incorrect decref of invalidated dentry (2350)
3653 - provide read-ahead stats and refine rpc in flight stats (3328)
3654 - don't hold journal transaction open across create RPC (3313)
3655 - update atime on MDS at close time (3265)
3656 - close LDAP connection when recovering to avoid server load (3315)
3657 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3658 - don't leak open file on MDS after open resend (3325)
3659 - serialize filter_precreate and filter_destroy_precreated (3329)
3660 - loop device shouldn't call sync_dev() for nul device (3092)
3661 - clear page cache after eviction (2766)
3662 - resynchronize MDS->OST in background (2824)
3663 - refuse to mount the same filesystem twice on same mountpoint (3394)
3664 - allow llmount to create routes for mounting behind routers (3320)
3665 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3666 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3667 - fix rare problem with rename on htree directories (3417)
3669 - allow default OST striping configuration per directory (1414)
3670 - fix compilation for qswnal for 2.6 kernels (3125)
3671 - increase maximum number of MDS request buffers for large systems
3672 - change liblustreapi to be useful for external progs like lfsck (3098)
3673 - increase local configuration timeout for slow disks (3353)
3674 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3676 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3679 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3680 - don't overwrite extent policy data in reply if lock was blocked (2901)
3681 - drop filter export grants atomically with removal from device (2663)
3682 - del obd_self_export from work_list in class_disconnect_exports (2908)
3683 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3684 - swab reply message in mdc_close, other PPC fixes (2464)
3685 - fix destroying of named logs (2325)
3686 - overwrite old logs when running lconf --write_conf (2264)
3687 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3688 - fix race in target_handle_connect (2898)
3689 - mds_reint_create() should take same inode create lock (2926)
3690 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3691 - don't close files for self_export to avoid uninitialized obd (2936)
3692 - allow MDS with the same name as client node (2939)
3693 - hold dentry reference for closed log files for unlink (2325)
3694 - reserve space for all logs during transactions (2059)
3695 - don't evict page beyond end of stripe extent (2925)
3696 - don't oops on a deleted current working directory (2399)
3697 - handle hard links to targets without a parent properly (2517)
3698 - don't dereference NULL lock when racing during eviction (2867)
3699 - don't grow lock extents when lots of conflicting locks (2919)
3701 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3704 - account for cache space usage on clients to avoid data loss (974)
3705 - lfsck support in lustre kernel code (2349)
3706 - reduce journal credits needed for BRW writes (2370)
3707 - orphan handling to avoid losing space on client/server crashes
3708 - ptlrpcd can be blocked, stopping ALL progress (2477)
3709 - use lock value blocks to assist in proper KMS, faster stat (1021)
3710 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3711 - recovery for initial connections (2355)
3712 - fixes for mds_cleanup_orphans (1934)
3713 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3714 - block all file creations until orphan recovery completes (1901)
3715 - client remove rq_connection from request struct (2423)
3716 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3717 - recovery timer race (2670)
3718 - mdc_close recovey bug (2532)
3719 - ptlrpc cleanup bug (2710)
3720 - mds timeout on local locks (2588)
3721 - namespace lock held during RPCs (2431)
3722 - handle interrupted sync write properly (2503)
3723 - don't try to handle a message that hasn't been replied to (2699)
3724 - client assert failure during cleanup after abort recovery (2701)
3725 - leak mdc device after failed mount (2712)
3726 - ptlrpc_check_set allows timedout requests to complete (2714)
3727 - wait for inflight reqs when ptlrpcd finishes (2710)
3728 - make sure unregistered services are removed from the srv_list
3729 - reset bulk XID's when resending them (caught by 1138 test)
3730 - unregister_bulk after timeout
3731 - fix lconf error (2694)
3732 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3733 - readahead locks pages, leaves pending causing memory pressure (2673)
3734 - increase OST request buffers to 4096 on large machines (2729)
3735 - fix up permission of existing directories in simple_mkdir (2661)
3736 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3737 - don't assign transno to errored transactions (2742)
3738 - don't delete objects on OST if given a bogus objid from MDS (2751)
3739 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3740 - if rq_no_resend, then timeout request after recovery (2432)
3741 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3742 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3743 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3744 - MDS recovery shouldn't delete objects that it has given out (2730)
3745 - if enqueue arrives after completion, don't clobber LVB (2819)
3746 - don't unlock pages twice when trigger_group_io returns error (2814)
3747 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3748 - don't write pages to disk if there was an error (1450)
3749 - don't ping imports that have recovery disabled (2676)
3750 - take buffered bytes into account when balancing socknal conn (2817)
3751 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3752 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3753 - remove little-endian swabbing of llog records (1987)
3754 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3755 - retry reposting request buffers if they fail (1191)
3756 - grow extent at grant time to avoid granting a revoked lock (2809)
3757 - lock revoke doesn't evict page if covered by a second lock (2765)
3758 - disable VM readahead to avoid reading outside lock extents (2805)
3760 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3761 - updated kernel patches for hp-2.4.20 kernel (2681)
3763 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3766 - fix truncated write corruption (2366)
3767 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3769 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3770 - ensure len(uuid) < 37 in lmc (1171)
3771 - fix ia64 OOPS in llog_test (2255)
3772 - zero end of page at obdfilter for partial page writes (2648)
3773 - don't leave stale dentries around after renames (bug 2428)
3774 - fix timeouts when evicting a client with a single lock held (2642)
3775 - set deadline for the initial HELLO message to drain (2634)
3776 - print out dotted-quad IP addresses in the socknal (2302)
3778 - additional debugging for MDS client eviction problem (2443)
3779 - fix mkfsoptions support for osts (2603, 2604)
3781 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3784 - add series for the vanilla 2.6.0 kernel
3785 - add series for the vanilla 2.4.24 kernel
3786 - add series for a cray x86/64 UL kernel drop
3787 - fix xattr patches for the vanilla 2.4.19 series
3789 - generate true UUIDs in lmc (1171)
3790 - have portals stack dumping break in UML (2466)
3791 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3792 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3793 - wake callback waiting threads on client eviction (2460)
3794 - Add --ptldebug and --subsystem to lmc (1719)
3795 - update assertion to allow safe interrupt allocation
3796 - set rq_no_resend for cancel requests (2432)
3797 - recalculate ptlrpcd timeout after resend (2494)
3798 - call vfs_rmdir when removing pending directories (2368)
3799 - fix renaming a file to itself (2429)
3800 - lmc creates a default one-stripe lov (2454)
3801 - expand procfs space to handle large clusters (2326)
3802 - increase UML stack to avoid overflow
3803 - update lconf's list of debug and subsystem masks
3804 - fix lfs find --obd (2510)
3805 - /proc tunable for disabling filter read caching (2591)
3806 - stop rpm packages from altering slapd.conf (2301)
3807 - disable nagle in the socknal under 0conf (2578)
3808 - choose mds inode size based on stripe count (2572)
3809 - fix kernel-source rpm problems (2516)
3811 - add --disable-doc to avoid pdf generation (2421)
3812 - update documentation, tests, type-os, comments
3813 - avoid format warnings on ia64
3814 - remove the TOE NAL
3815 - tiny code cleanups by removing unused fields
3817 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3820 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3821 - remove the most blatant lies from BUILDING file (2371)
3822 - change default debug level to reasonable production setting
3823 - reduce client side cache size to reduce cache flush time
3824 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3825 - make TCP zerocopy and pinger support enabled by default (2476)
3826 - sync writes completed after process exits caused crashes (2319)
3827 - maintain correct mount count on the MDS (2356)
3828 - backout 1557, because 2316 wasn't really fixed
3829 - better file I/O statistics gathering in /proc
3830 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3831 - another kernel patch to fix zero-copy TCP function export
3832 - don't take duplicate lock when processing re-sent getattr (2420)
3833 - lctl uses obd_self_export instead of creating new conn (2353)
3834 - MDS/OST recovery case which requires object creation asserted (2425)
3835 - move lfs from /usr/sbin to /usr/bin in packages
3836 - fix race between mds_client_add and mds_client_free (2417)
3837 - use kmalloc instead of slabs in portals (2430)
3838 - don't create duplicate records when a failover MDS is present (2442)
3839 - remove unnecessary mount age check (2332)
3840 - don't remove directory inodes from locks prematurely (2451)
3841 - don't break if MDS service name is the same as hostname (2103)
3842 - fix races in client write RPC generation when cache full (2482)
3844 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3847 - remove now-unused request->rq_obd (278)
3848 - if an allocation fails, print out how much memory we've used (1933)
3849 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3850 - add the "configurable stack size" patch to most series files (1256)
3851 - ability to write large log records, for 100+ OST configs (2306)
3852 - fix NULL deref when filter_prep fails (2314)
3853 - fix operator precedence error in filter_sync
3854 - dynamic allocation of socknal TX descriptors (2315)
3855 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3856 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3857 - ensure that CWARN messages in Portals always get to the syslog
3858 - __init/__exit are not for prototype decls (ldlm_init/exit)
3859 - x86-64 compile warning fixes
3860 - fix gateway LMC keyword conflict (2318)
3861 - fix MDS lock inversions in getattr/reint paths (1844)
3862 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3863 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3864 - DLM race condition prevented some lock evictions (2328)
3865 - ENOMEM detection and retry on socknal sends (2230)
3866 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3867 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3868 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3869 - better error messages when a client is rejected during recovery (1505)
3870 - avoid cancelling locks which were never granted, after failure (2330)
3871 - fix i_sem/journal inversion in mds_client_add (2333)
3872 - fix truncate/getattr lock cycle deadlock (2334)
3873 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3874 - fix two rare exit paths which could leak an l_lock() ref (2321)
3875 - fencepost error in MDS/OST orphan recovery (2226)
3876 - make log record alignment 8 bytes (1988)
3877 - lstripe now fails when requested offset > ost_count (2237)
3878 - ensure that all kernel series have a complete list.h (1607)
3879 - fix crashes in special-file operations (2316)
3880 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3882 - allow configurable automake binary, for testing new versions
3883 - small update to the lfs documentation
3885 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3887 * fix negative export reference count in fsfilt_sync (2312)
3889 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3890 * release candidate 0.9.1
3892 - orphans are moved into the PENDING directory for possible recovery
3893 - replayed opens now open by fid for orphan/rename safety (1042)
3894 - last close of an orphan inode generates a transno (683)
3895 - chdir() and mount() now pin the directory entry (1020)
3896 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3897 - discard very old requests without processing them (1502)
3898 - remove some common, well-understood CERRORs (1505)
3899 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3900 - clear "grant" flags in OST replies until OST grant code lands (1644)
3901 - fix read performance by not clobbering i_blksize on client (1598)
3902 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3903 - make LRU size a /proc tunable, clears locks when reduced (707)
3904 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3905 - server locks take references on exports now (1558)
3906 - build fixes for 2.4.20-rh trees (1663)
3907 - return an error from lov_create if all OSCs are inactive (1751)
3908 - fix import levels when a reconnect happens without a timeout (1597)
3909 - exit early from mds_open if we get a lookup error (1749)
3910 - partial page read at EOF wouldn't wait for disk before sending (1642)
3911 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3912 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3913 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3914 - write/truncate lock inversion (1639)
3915 - Don't auto-load obdclass, portals modules during cleanup (1495)
3916 - fix timestamps from jumping to "now" (1763)
3917 - extra journal assertions (1648)
3918 - add an extra multiunlink test (1771)
3919 - fix read_record/write_record API (1776)
3920 - fix leak of offset_extent, possible incorrect i_size later (1772)
3921 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3922 - Debugging for the kqswnal_get_idle_tx problems (1820)
3923 - Allow recovery to be attempted multiple times (1536)
3924 - Write out MDS last_rcvd file after it is first created (1600)
3925 - Fix tx_descriptor leak in failed transmit situations (1827)
3926 - ext3 journaling fixes for assertion failure after IO error (1871)
3927 - class_export_put() on freed export after completion AST error (1896)
3928 - Fix revalidate looping in VFS (1322)
3929 - Don't access a freed export during MDS_REINT timeout (1521)
3930 - Add open-unlink recovery support on the MDS (1673,1764)
3931 - Return an error if no MDS data was read from last_rcvd (1946)
3932 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3933 - Don't setup a disk device that doesn't match exported UUID (317)
3934 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3935 - avoid committing NULL handle in force close
3936 - local.sh is now a one-stripe LOV configuration
3937 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3938 - fix ext3 htree duplicate directory entry corruption (1516)
3939 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3940 - update to new LOV EA format (2097)
3941 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3943 2003-06-15 Phil Schwan <phil@clusterfs.com>
3946 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3947 - per-import/export recovery handling (958, 931, 959)
3948 - multiple last-rcvd slots, for serving multiple FSes (949)
3949 - connections are again shared between multiple imp/exports (963, 964)
3950 - "umount -f" would hang if any requests needed to be sent (393, 978)
3951 - avoid pinning large req buffer by copying for queued messages (989)
3952 - add "uuid" to "lctl device" command to help upcalls (991)
3953 - "open" RPCs with transnos would confuse recovery counters (1037)
3954 - do proper endian conversion of all wire messages (288, 340, 891)
3955 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3956 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3957 - fix ptlrpc_connection leak in target_handle_connect (1174)
3958 - fix import refcounting bug in OST and MDS cleanup (1134)
3959 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3960 - fix very unlikely obd_types race condition (501)
3961 - remove osc_open hack for echo_client (1187)
3962 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3963 - a failure in read_inode2 leads to deadlock (1139)
3964 - cancel ack-locks as soon as transaction is committed (1072)
3965 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3966 - make sure to commitrw after any preprw to avoid deadlock (1162)
3967 - failing to execute a file in a lustre FS would lock inode (1203)
3968 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3969 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3970 - obd_uuid2tgt would walk off the end of the list (1255)
3971 - on IA64 the osc would give portals incorrect bulk size (1258)
3972 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3973 - fix lock inversion caused by new llite matching code (1282)
3974 - limit the number of dirty pages on a client to 10MB (1286)
3975 - timed out locks were not being corrected cancelled (1289)
3976 - fix O_DIRECT above 4GB on IA-32 (1292)
3977 * major user-visible changes
3978 - fail out/fail over policy now controlled by the upcall (993)
3980 - add OBD_PING to check server availability and failure (954)
3981 - lustre messages are now sent in sending host order (288, 340, 891)
3982 - add eadatalen to MDS getattr reply (340)
3983 - OST read replies may contain second buffer, with per-page status (593)
3985 2003-03-11 Phil Schwan <phil@clusterfs.com>
3988 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3989 - failed open()s could cause deadlock; fixed (867, 869)
3990 - stop cancelling OST locks when files are closed (481)
3991 - overlapping XID spaces caused network corruption (851, 853)
3992 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3993 - setattr_raw added, to do single-RPC, server-side setattrs
3994 - lmc/lconf syntax change for OST UUIDs
3995 - fix crashy race condition between ptlrpc_free_req and osc_close
3996 - don't use request in mdc_enqueue if we hit a timeout (889)
3997 - don't set the inode i_size for regular files from the MDS (896)
3998 - handle out of order completion AST (842)
3999 - don't LBUG if a lock request times out after receiving AST (913)
4000 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4001 - if a bad lock AST arrives, send an error instead of dropping entirely
4002 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4003 - fix leak in bulk IO when only partially completed (899, 900, 926)
4004 - fix O_DIRECT for ia64 (55)
4005 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4006 - C-z after timeout could hang a process forever; fixed (977)
4008 - client-side I/O cache (678, 924, 929, 941, 970)
4010 - READPAGE and SETATTRs which don't take server-side locks get
4013 2003-02-11 Phil Schwan <phil@clusterfs.com>
4016 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4017 - Fix fsfilt_extN_readpage() to read a full page of directory
4018 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4019 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4020 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4021 - Most intent operations were removed, in favour of a new RPC mode
4022 that does a single RPC to the server and bypasses most of the VFS
4023 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4024 - Aggressively cancel local locks on DLM servers
4025 - mds_reint_unlink sends EA to the client if it's the last nlink.
4026 client uses that EA to unlink OST objects.
4027 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4028 - recursive symlinks were fixed (439)
4029 - fixed NULL deref in DEBUG_REQ
4030 - filter_update_lastobjid no longer calls sync, which annoyed extN
4031 - fixed multi-client small-writes to a single file problem (445)
4032 - fixed mtime updates during file writes (607)
4033 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4034 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4035 - fixed Linux OST size reporting problem (444, 656)
4036 - OST now updates object mtime with writes or setattr (607, 619)
4037 - client verifies file size before zeroing page past EOF (445)
4038 - OST now writes last allocated objid to disk with allocation (108)
4039 - LOV on echo now works (409)
4041 - mds_reint_unlink sends a new buffer, with the EA included. this
4042 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4043 set if a regular file was being unlinked, and it was the last link
4044 - use PtlGet from the target for bulk writes (315)
4045 - OST now updates object mtime with writes or setattr (607, 619)
4046 - LDLM now has a grant-time callback to revalidate locked items, if
4048 - Many MDS operations were reorganized to combat race conditions
4050 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4051 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4053 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4056 - Fully reactivate OST imports after reconnection (512, others)
4057 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4058 - More graceful error handling for truncating on dead OST (515)
4059 - Don't error out unless we're actually accessing dead stripes (474)
4060 - Fix garbage sizes when stripes are missing (410)
4061 - LRU counters were broken, causing constant lock purge (433, 432)
4062 - garbage on read from stripes with failed OSTs (441)
4063 - mark OSCs as active before reconnecting during recovery (438)
4064 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4065 - lfind did not preserve OST order in output (443)
4066 - symlinks cause hung clients, incorrect data (439)
4067 - stop dereferencing request after dropping refcount (457)
4068 - don't LASSERT(spin_is_locked) on non-SMP (455)
4069 - fixes for many rename() bugs
4070 - fstat didn't correctly synchronize attributes (399)
4071 - server must handle lock cancellation during blocking AST prep (487)
4072 - bulk descriptors were free()d too soon (511)
4073 - fix paths in lconf, which would load incorrect modules (451, 507)
4074 - fix confusing lconf 'host not found' error message (386)
4075 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4076 - fix race condition in mdc_blocking_ast() for inode access (526)
4077 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4078 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4079 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4080 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4081 - fix lstripe in several ways (kernel side) (527)
4082 - fix request leak in ldlm_cli_enqueue (262)
4083 - incorrect OSC was marked inactive after OST failure
4084 - call mds_fs_cleanup before unmounting filesystem (524)
4085 - fix races between taking ns_lock and ldlm_lock_change_resource
4086 - fix races updating LOV export open file list
4087 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4088 - fix recovery NULL deref in ldlm_cli_cancel_unused
4089 - fix some DLM races by using new hash table for lock handles (419)
4090 - permit the client to specify desired inodes, at replay
4091 - duplicate requests when we queue them for replay reintegration
4092 - fix last_rcvd offset calculation
4093 - sync after each recovered transaction, so we always make progress
4094 - never, not always, ERESTART requests without transnos
4095 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4097 - skip replay if the MDS says that the client is already connected
4098 - don't check for a recovery-enabled export to match lctl's UUID
4099 - don't INC_USE_COUNT for phantom exports
4100 - don't crash when cleaning up phantom exports (567)
4101 - don't double-finish or set replay data for errored mdc_open requests
4102 - abort requests when they time out, so we don't get old replies
4103 - send/receive replies for AST messages again
4104 - if the client says that it doesn't have the lock, cancel it on the
4106 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4107 mark it as destroyed, it will all work out when decref is called
4108 - fix module use counts (22, 581)
4110 - ASTs now expect a reply (server cancels lock on error reply)
4112 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4115 - fix many simultaneous client startup (392)
4116 - fix dentry->d_it clobbering
4117 - credentials weren't being shipped for readdir/getattr operations
4118 - remove invalid assertions triggered during some concurrent MD
4120 - proper Lustre versions added (336, 389)
4121 - fix memory leak for create error case (398)
4122 - fix LOV locking bug that would get cli/srv out of sync
4123 - fix echo client over LOV (409)
4124 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4125 - fix double-O_EXCL intent crash (424)
4126 - avoid sending multiple lock CANCELs (352)
4128 - MDS can do multi-client recovery (modulo bugs in new code)
4130 - many updates, edits, cleanups
4132 2002-11-18 Phil Schwan <phil@clusterfs.com>
4135 - fix null d_it dereference (346)
4136 - fix full OST/dbench hang (333)
4137 - fix permission problem with file removal (286)
4138 - fix removal of OSCs from LOV when they fail
4139 - fix NULL deref during bulk timeout (214)
4140 - fix problems related to multiple filesystems on one MDS (241)
4141 - fixed serious subtle metadata locking bugs
4142 - free locks on clients when inodes are removed due to memory
4144 - fix inode pointer in lock data (285)
4145 - partial support for multiple MDS on a single host (241)
4146 - data locks weren't cancelled at clear_inode time (290, 311)
4147 - intent locks could lead to unbounded lock growth (205)
4148 - added a maximum lock count, an LRU list, and a flusher
4149 - fix multiple rename (365)
4150 - properly abstracted the echo client
4151 - OSC locked 1 byte too many; fixed
4152 - rewrote brw callback code:
4153 - fixed recovery bugs related to LOVs (306)
4154 - fixed too-many-pages-in-one-write crash (191)
4155 - fixed (again) crash in sync_io_timeout (214)
4156 - probably fixed callback-related race (385)
4158 - Add capability to MDS protocol
4159 - LDLM cancellations and callbacks on different portals
4161 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4164 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4165 - defer open object destruction to close time (601981, 138)
4166 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4167 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4168 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4169 - don't call obd_setattr in ll_file_release for destroyed objects
4170 * protocol change to lustre_msg: move |version| and add |flags|
4171 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4172 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4173 * added replay of create, unlink, link and rename operations during
4174 MDS failover; recovery should be much more robust now
4175 * remove failed OSCs from LOVs (only lov_create uses this so far)
4176 * the lustre-HOWTO was brought (more) up to date (582544)
4178 2002-10-23 Phil Schwan <phil@clusterfs.com>
4181 - in-use dentries weren't being reused properly (617851)
4182 - prevent multiple LDLM setup (599178)
4183 - fix LOV size calculations for truncate (617853)
4184 - fix client handling of MDS intent errors (POSIX)
4185 - fix permission bug in lovstripe.c test (624321)
4186 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4187 - truncate past end of file could corrupt data
4188 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4189 - a race in recovery could return ETIMEDOUT to apps (623947)
4190 - building outside the source directory was fixed
4191 * the lustre-HOWTO was brought (more) up to date (582544)
4192 * major progress was made on recovery functionality
4194 2002-10-10 Phil Schwan <phil@clusterfs.com>
4197 - recovery deadlock fix
4198 - rm -rf causes LBUG fix (617817)
4199 - file open by multiple tasks fix (618962)
4200 - directory permissions bugs (602707 and 620007)
4201 - journal_stop fixed with locking (611313)
4202 - O_APPEND failures resolved (618273, perhaps 614459)
4203 - lconf PATH fix (619770)
4204 - IA64 build fix (621450)
4205 - RPC buffer sizes scale with amount of memory
4207 2002-10-01 Phil Schwan <phil@clusterfs.com>
4210 - locks would be cancelled without throwing away data pages,
4211 resulting in inconsistent data (605627)
4212 - inode attributes were not always being refreshed (605627, 612449)
4213 - lconf now continues to cleanup after lctl reports an error
4214 - MDS now enforces user permissions (602707)
4215 - lprocfs cleanup fixed, but not yet enabled (614157)
4216 - fixed infinite server hang, should a client not respond to an AST
4217 - avoid going into recovery if user calls readlink() with a buffer
4218 that's too small (613941)
4219 - AST RPCs no longer require replies (614867) -- this may be changed
4220 - don't crash server if client sends an IOV that's too big (611336)
4221 - fixed lock conversion deadlock (611892)
4222 - fixed the following of symlinks (614622)
4223 * recovery: the server can remove locks from a client that dies, other
4224 clients can make progress
4225 * more extN patch fixes
4226 * compile-time configurable ptlrpc buffer allocations
4228 - collaborative read cache document
4229 - Lustre Lite Performance CDR document-in-progress
4231 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4234 - fix typo in patch-2.4.18
4236 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4239 - clear ptlrpc request each time in handle_incoming_request()
4240 - unlink of files now destroys the object on the OST
4242 2002-09-19 Peter Braam <braam@clusterfs.com>
4244 * add hard link support
4245 * change obdfile creation method
4246 * kernel patch changed
4248 2002-09-19 Peter Braam <braam@clusterfs.com>
4251 - stack overflow bug in extN fixed
4253 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4255 * documentation updates
4256 - add man pages for config tools
4257 - update tests/README to describe testing with new config tools
4258 - finish metadata API descriptions
4259 * bug fixes and cleanups
4260 - statfs workaround for 16TB limit
4261 - LOV stripe allocation improved, can stripe on subset of OSTs
4262 - LOV file size/IO offset was wrong for files > 4GB in size
4263 - object EA data was being dropped, caused files to be unreadable
4264 - memory overflow with non-LOV OST caused memory corruption
4265 - fixed regression tests to work with new config tools, obdfilter
4266 - fixed bug when directory size became larger than 1 block
4267 - fixed bug (for single client case) when PWD was deleted
4268 - invalidate local directory pages when doing intent-based ops
4269 - avoid LDLM oops when lock callback contained bad data
4271 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4273 * documentation updates
4274 * bug fixes and cleanups
4275 - configuration tools
4278 - 64-bit compile warnings
4279 - 64-bit internal statfs data
4281 * test_brw on persistent OST devices
4283 * lprocfs (disabled)
4285 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4287 * documentation updates
4288 * bug fixes and cleanups
4289 * configuration tools
4291 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4294 * many small fixes to 0_5_4
4295 * io/network handling
4296 * thinkos in MDS operations
4298 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4301 * crucial basic fixes to 0.5.3
4302 * IOR, Iozone work over Elan
4305 2002-08-07 Phil Schwan <phil@clusterfs.com>
4306 * version 0_5_3, our first alpha
4307 * we use the new Portals iovs
4308 * documentation updates
4309 * bug fixes and cleanups
4310 * small changes in the DLM wire protocol
4312 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4313 * version 0_5_1 with some initial stability,
4314 * locking on MD and file I/O.
4315 * documentation updates
4316 * several bug fixes since 0.5.0
4317 * small changes in wire protocol
4319 2002-07-18 Phil Schwan <phil@clusterfs.com>
4321 * delivered as Lustre Light Alpha
4322 * fixed a crash after handling invalid MDS requests
4323 * fixed directory pages for architectures with non-4k pages sizes
4325 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4326 * release version v0_4_4
4327 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4329 * Fixes a number of interruption problems with OST operations.
4330 * Update documentation for portals header changes
4331 * Move all wire protocol structs/defines to lustre_idl.h
4332 * Fixes symlink length bug.
4333 * Add tcpdump to repository.
4335 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4336 * release version v0_4_3
4337 * Fixes statfs for inodes on extN.
4338 * Fixes bug in runtests which would delete /etc/hosts.
4339 * Use 64-bit object IDs wherever possible (not into VFS though)
4340 Remove ost_get_info, which is unused by lustre, and out of date.
4342 2002-07-03 Peter Braam <braam@clusterfs.com>
4343 * release version v0_4_2 Fixes a lookup error (type not passed)
4344 * move forward to head of Portals
4345 * move forward to latest Lustre kernel
4347 2002-06-25 Peter Braam <braam@clusterfs.com>
4348 * release version v0_4_1. Hopefully stable on single node use.