1 tbd Sun Microsystems, Inc.
4 2.6.9-67.0.1.EL (RHEL 4),
5 2.6.16.54-0.2.3 (SLES 10),
6 2.6.18-53.1.4.el5 (RHEL 5).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.40.4-cfs1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
17 Description: allow userland application know is lost one of stripes.
18 Details : fill lvb_blocks with error code on ost and return it to
19 application if error flag found.
23 Description: NULL lov_tgts causing MDS oops
24 Details : more safe checks for NULL lov_tgts for avoid oops.
26 Severity : enhancement
28 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
30 Severity : enhancement
32 Description: Add rhel5 support to HEAD.
36 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
37 Details : If an OST is down the MDS will hang indefinitely in
38 obd_statfs_rqset() waiting for the statfs data. While for
39 MDS QOS usage of statfs, it should not stuck in waiting.
41 Severity : enhancement
43 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
45 Severity : enhancement
47 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
51 Description: make mgs_setparam() handle fsname containing dash
52 Details : fsname containing a dash does not work with lctl conf_param
54 Severity : enhancement
56 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
58 Severity : enhancement
60 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
64 Frequency : rare, at shutdown
65 Description: access already free / zero obd_namespace.
66 Details : if client_disconnect_export was called without force flag set,
67 and exist connect request in flight, this can produce access to
68 NULL pointer (or already free pointer) when connect_interpret
69 store ocd flags in obd_namespace.
73 Frequency : only at startup
74 Description: not alloc memory with spinlock held.
75 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
78 Severity : enhancement
80 Description: make lustre randomly fail allocating memory
81 Details : Make lustre randomly failed allocating memory for testing purpose.
83 Severity : enhancement
85 Description: lost problems with lov objid file
86 Details : Fixes some scability and access to not inited memory problems
87 in work with lov objdid file.
92 Description: lfs find does not continue on file error
93 Details : Continue other files processing when a file/dir is absent.
97 Description: Inconsistent usage of lustre_pack_reply()
98 Details : Standardize the usage of lustre_pack_reply() such that it
99 always generate a CERROR on failure.
102 Frequency : occasional
104 Description: MDS hang or stay in waiting lock
105 Details : If client receive lock with CBPENDING flag ldlm need send lock
106 cancel as separate rpc, to avoid situation when cancel request
107 can't processed due all i/o threads stay in wait lock.
111 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
115 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
119 Description: Update to SLES9 kernel-2.6.5-7.287.3.
123 Description: Update to RHEL4 latest kernel.
125 Severity : enhancement
127 Description: Build SLES10 patchless client fails
128 Details : The configure was broken by run ./configure with
129 --with-linux-obj=.... argument for patchless client. When the
130 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
131 can't search header adequately. Use absolute path such as
132 -I($LINUX)/include instead.
136 Description: interrupt oig_wait produce painc on resend.
137 Details : brw_redo_request can be used for resend requests from ptlrpcd and
138 private set, and this produce situation when rq_ptlrpcd_data not
139 copyed to new allocated request and triggered LBUG on assert
140 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
141 ptlrpcd set if request is changed and can be safety changed to use
146 Description: LASSERT_{REQ,REP}SWAB macros are buggy
147 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
148 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
149 these macros and replace them with their difinitions instead.
153 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
154 Details : Update which_patch & target file for SLES10 latest kernel.
157 Description: Remove client patches from SLES 10 kernel.
158 Details : This causes SLES 10 clients to behave as patchless clients
159 even on a Lustre-patched (server) kernel.
161 Severity : enhancement
163 Description: self-adjustable client's lru lists
164 Details : use adaptive algorithm for managing client cached locks lru
165 lists according to current server load, other client's work
166 pattern, memory activities, etc. Both, server and client
167 side namespaces provide number of proc tunables for controlling
170 Severity : enhancement
172 Description: light-weight GSS support
173 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
174 and reduce performance overhead.
176 Severity : enhancement
178 Description: Linux keyring support
179 Details : Support using service of Linux keyring for Lustre GSS internal
180 context refresh/cache mechanism.
184 Description: Fix errors in lfs documentation
185 Details : Fixes man pages
189 Description: don't use GFP_* in generic Lustre code.
190 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
194 Description: obdclass is limited by single OBD_ALLOC(idarray)
195 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
199 Description: warnings with build patchless client with vanila 2.6.19 and up
200 Details : change old ctl_table style and replace ctl_table/ctl_table_header
201 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
205 Frequency : Only for SLES
206 Description: sanity_quota fail test_1
207 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
208 file, which makes TSTID[2] inunique.
212 Frequency : Only for SLES9
213 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
214 Details : Update target/ChangeLog/which_patch .
218 Frequency : Only for test_10 in sanity-quota.sh
219 Description: a bug in quota test code
220 Details : checking a test flag in a wrong place
224 Frequency : Only for test_13 in sanity-quota.sh
225 Description: a bug in quota test script
226 Details : shouldn't check overfull allocated quota
230 Description: jbd statistics
231 Details : Port older jbd statistics patch for sles10
233 Severity : enhancement
235 Description: Add printing inode info into message about error in writepage.
240 Description: reply_lock_interpret crash due to race with it and lock cancel.
241 Details : Do not replay locks that are being cancelled. Do not reference
242 locks by their address during replay, just by their handle.
246 Frequency : When flocks are used.
247 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
248 Details : Flock locks might destroy just granted lock if it could be merged
249 with another existing flock, this is done in completion handler,
250 so teach ldlm_cli_enquque_fini that this is a valid case for
256 Description: Oops in read and write path when failing to allocate lock.
257 Details : Check if lock allocation failed and return error back.
261 Description: lstripe command fails for valid OST index
262 Details : The stripe offset is compared to lov->desc.ld_tgt_count
263 instead of lov->desc.ld_active_tgt_count.
267 Description: sanity.sh failed test 103
268 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
273 Description: df doesn't work properly if diskfs blocksize != 4K
274 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
278 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
279 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
283 Description: obd_health_check_timeout too short
284 Details : set obd_health_check_timeout as 1.5x of obd_timeout
288 Description: llapi_file_create() does not allow some changes
289 Details : add llapi_file_open() that allows specifying the mode and
290 open flags, and also returns an open file handle.
294 Description: merge and cleanup kernel patches.
295 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
299 Description: Add journal checksum support.(Kernel part)
300 Details : The journal checksum feature adds two new flags i.e
301 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
302 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
303 indicates that the commit block contains the checksum for
304 the blocks described by the descriptor blocks. Now commit
305 record can be sent to disk without waiting for descriptor
306 blocks to be written to disk. This behavior is controlled
307 using JBD2_FEATURE_ASYNC_COMMIT flag.
311 Description: OSS needs mutliple precreate threads
312 Details : Add ability to start more than one create thread per OSS.
316 Description: Sanity.sh test_65k failed: setstripe should have succeeded
317 Details : execute lfs setstripe on client
321 Description: mds_obd_create error creating tmp object
322 Details : When the user sets quota on root, llog will be affected and can't
323 create files and write files.
326 Frequency : Always on ia64 patchless client, and possibly others.
328 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
329 Details : This allows the patchless client to be loaded on architectures
334 Description: RedHat Update kernel for RHEL5
335 Details : Add the kernel config file for RHEL5.
339 Description: RedHat Update kernel for RHEL5
340 Details : Modify the kernel config file more closer RHEL5.
344 Description: Build failure against Centos5 (RHEL5)
345 Details : Define PAGE_SIZE when it isn't present.
349 Description: test_7 on sanity-quota.sh can't be used on separate machines
350 Details : add facet to handle it
354 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
355 Details : with NFS, the anon dentry's parent was set to itself in
356 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
357 corresponding dentry other than use rec->ur_name.
359 Severity : enhancement
361 Description: lfs setstripe enhancement
362 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
366 Description: enable data checksumming by default
367 Details : enable checksum by default, allow --disable-checksum
368 configure option and "-o nochecksum" mount option. Checksums
369 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
373 Description: lustre support for RHEL5
374 Details : Add support for RHEL5.
378 Description: Client eviction due to failover config
379 Details : after a connection loss, the lustre client should attempt to
380 reconnect to the last active server first before trying the
381 other potential connections.
385 Description: when mds and osts use different quota unit(32bit and 64bit),
386 quota will be released repeatly.
387 Details : void sending multiple quota reqs to mds, which will keep the
388 status between the reqs.
392 Description: osts not allocated evenly to files
393 Details : change the condition to increase offset_idx
396 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
398 Description: Data corruption for OSTs that are formatted larger than 2TB
400 Details : When generating the bio request for lustre file writes the
401 sector number would overflow a temporary variable before being
402 used for the IO. The data reads correctly from Lustre (which
403 will overflow in a similar manner) but other file data or
404 filesystem metadata may be corrupted in some cases.
408 Description: Tune the kernel for good SCSI performance.
409 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
410 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
415 Description: rewrite ext2-derived code in obdclass/uuid.c
416 Details : rewrite inherited code (uuid parsing code from ext2 utils)
417 from scratch preserving functionality.
421 Description: "lfs find -obd UUID" prints directories
422 Details : "lfs find -obd UUID" will return all directory names instead
423 of just file names. It is incorrect because the directories
424 do not reside on the OSTs.
428 Description: use i_size_read and i_size_write in 2.6 port
429 Details : replace inode->i_size access with i_size_read/write()
432 Frequency : when removing large files
434 Description: scheduling issue during removal of large Lustre files
435 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
436 It causes scheduling issues when removing large files (17TB in the
440 Frequency : only with liblustre clients on XT3
442 Description: evictions taking too long
443 Details : allow llrd to evict clients directly on OSTs
446 Frequency : only on ppc
448 Description: /proc/fs/lustre/devices broken on ppc
449 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
450 We should make sure the type of 'index' is loff_t and then cast
451 explicitly as needed below. Do not assign an explicitly cast
455 Frequency : only for rhel5
457 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
458 Details : Modify the target file & which_kernel.
460 Severity : enhancement
462 Description: omit set fsid for export NFS
463 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
464 as FS_REQUIRES_DEV for avoid problems with generate fsid.
468 Frequency : Always, for kernels after 2.6.16
469 Description: Fix warning idr_remove called for id=.. which is not allocated.
470 Details : Last kernels save old s_dev before kill super and not allow
471 to restore from callback - restore it before call kill_anon_super.
475 Description: buffer overruns could theoretically occur
476 Details : llapi_semantic_traverse() modifies the "path" argument by
477 appending values to the end of the origin string, and a
478 overrun may occur. Adding buffer overrun check in liblustreapi.
482 Description: Fix error on 'ls .' at the top of the Lustre mount.
483 Details : Don't revalidate dentry if it is a root dentry.
487 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
488 Details : Modify vm-tunables-rhel4.patch.
492 Description: Kernel config for 2.6.18-vanilla.
493 Details : Modify targets/2.6-vanilla.target.in.
494 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
495 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
496 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
497 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
501 Description: improve handling recoverable errors
502 Details : if request processig with error which can be recoverable on server
503 request should be resend, otherwise page released from cache and
508 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
509 Details : Modify target file & which_patch.
510 A flaw was found in the IA32 system call emulation provided
511 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
512 value could be stored in the %RAX register, which could trigger an
513 out-of-bounds system call table access. An untrusted local user
514 could exploit this flaw to run code in the kernel
515 (ie a root privilege escalation). (CVE-2007-4573).
519 Description: O_DIRECT bypasses client statistics.
520 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
521 read_ahead_stats were updated. copied the stats section from
522 osc_send_oap_rpc() into async_internal().
526 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
530 Description: change order of libsysio includes
531 Details : '#include sysio.h' should always come before '#include xtio.h'
535 Description: handle "serious error: objid * already exists" more gracefully
536 Details : If LAST_ID value on disk is smaller than the objects existing in
537 the O/0/d* directories, it indicates disk corruption and causes an
538 LBUG(). If the object is 0-length, then we should use the existing
539 object. This will help to avoid a full fsck in most cases.
541 Severity : enhancement
543 Description: adapt the lustre_config script to support the upgrade case
544 Details : Add "-u" option for lustre_config script to support upgrading 1.4
545 server targets to 1.6 in parallel.
549 Description: To avoid grant space > avaible space when the disk is almost
550 full. Without this patch you might see the error "grant XXXX >
551 available" or some LBUG about grant, when the disk is almost
553 Details : In filter_check_grant, for non_grant cache write, we should
554 check the left space by if (*left > ungranted + bytes), instead
555 of (*left > ungranted), because only we are sure the left space
556 is enough for another "bytes", then the ungrant space should be
557 increase. In client, we should update cl_avail_grant only there
558 is OBD_MD_FLGRANT in the reply.
562 Description: Update RHEL 4 kernel to fix local root privilege escalation.
563 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
564 described in CVE-2007-4573. This problem could allow untrusted
565 local users to gain root access.
568 Frequency : when using O_DIRECT and quotas
570 Description: Incorrect file ownership on O_DIRECT output files
571 Details : block usage reported by 'lfs quota' does not take into account
572 files that have been written with O_DIRECT.
577 Description: touch file failed when fs is not full
578 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
579 otherwise we can get ENOSP while fs is not full.
583 Description: parallel lock callbacks
584 Details : Instead of sending blocking and completion callbacks as separated
585 requests, adding them to a set and sending in parallel.
588 Frequency : only for Cray XT3
589 Bugzilla : 12829/13455
590 Description: Changing primary group doesn't change the group lustre assigns to
592 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
593 overridden with the primary group provided by the group upcall,
594 whereas the supplied fsgid can be trusted if it is in the list of
595 supplementary groups returned by the group upcall.
597 Severity : enhancement
599 Description: Allow masking D_WARNING, D_ERROR messages from console
600 Details : Console messages can now be disabled via lnet.printk.
604 Description: User code with malformed file open parameter crashes client node
605 Details : Before packing join_file req, all the related reference should be
606 checked carefully in case some malformed flags cause fake
607 join_file req on client.
611 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
612 Details : ldlm_completion_ast() assumes that a lock is granted when the req
613 mode is equal to the granted mode. However, it should also check
614 that LDLM_FL_CP_REQD is not set.
616 --------------------------------------------------------------------------------
618 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
620 * Support for kernels:
621 2.6.5-7.283 (SLES 9),
622 2.6.9-55.EL (RHEL 4),
623 2.6.16.46-0.14 (SLES 10),
624 2.6.18.8 vanilla (kernel.org)
625 * Client support for unpatched kernels:
626 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
627 2.6.16 - 2.6.22 vanilla (kernel.org)
628 * Due to recently discovered recovery problems, we do not recommend
629 using patchless RHEL 4 clients with this or any earlier release.
630 * Recommended e2fsprogs version: 1.39.cfs8
631 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
632 * Starting with this release, the ldiskfs backing filesystem required
633 by Lustre is now in its own package, lustre-ldiskfs. This package
634 should be installed. It is versioned separately from Lustre and
635 may be released separately in future.
639 Description: block reactivating mgc import until all deactivates complete
640 Details : Fix race when failing back MDT/MGS to itself (testing)
643 Frequency : at statup only
645 Description: mds_lov_synchronize race leads to various problems
646 Details : simultaneous MDT->OST connections at startup can cause the
647 sync to abort, leaving the OSC in a bad state.
649 Severity : enhancement
651 Description: add optional extra BUILD_VERSION info
652 Details : add a new environment variable (namely LUSTRE_VERS) which allows
653 to override the lustre version.
656 Frequency : 2.6.18 servers only
658 Description: ll_kern_mount() doesn't release the module reference
659 Details : The ldiskfs module reference count never drops down to 0
660 because ll_kern_mount() doesn't release the module reference.
665 Description: server LBUG when using old ost_num_threads parameter
666 Details : Accept the old ost_num_threads parameter but warn that it
667 is deprecated, and fix an off-by-one error that caused an LBUG.
672 Description: Transient SCSI error results in persistent IO issue
673 Details : iobuf->dr_error is not reinitialized to 0 between two
677 Frequency : sometimes when underlying device returns I/O errors
679 Description: OSTs not going read-only during write failures
680 Details : OSTs are not remounted read-only when the journal commit threads
681 get I/O errors because fsfilt_ext3 calls journal_start/stop()
682 instead of the ext3 wrappers.
686 Description: poor connect scaling with increasing client count
687 Details : Don't run filter_grant_sanity_check for more than 100 exports
688 to improve scaling for large numbers of clients.
691 Frequency : SLES10 only
693 Description: sanity-quota.sh quotacheck failed: rc = -22
694 Details : Quotas cannot be enabled on SLES10.
697 Frequency : liblustre clients only
699 Description: getdirentries does not give error when run on compute nodes
700 Details : getdirentries does not fail when the size specified as an argument
701 is too small to contain at least one entry
703 Severity : enhancement
705 Description: Add LNET router traceability for debug purposes
706 Details : If a checksum failure occurs with a router as part of the
707 IO path, the NID of the last router that forwarded the bulk data
708 is printed so it can be identified.
713 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
714 Details : Due to a race condition, liblustre clients were occasionally
717 Severity : enhancement
719 Description: lfs setstripe use optional parameters instead of postional
722 Severity : enhancement
724 Description: Nanosecond timestamp support for ldiskfs
725 Details : The on-disk ldiskfs filesystem has added support for nanosecond
726 resolution timestamps. There is not yet support for this at
727 the Lustre filesystem level.
730 Frequency : during server recovery
732 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
733 Details : request with rq_no_resend flag not awake l_wait_event if they get a
737 Frequency : nfs export on patchless client
739 Description: connectathon hang when test nfs export over patchless client
740 Details : Disconnected dentry cannot be found with lookup, so we do not need
741 to unhash it or make it invalid
744 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
745 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
746 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
751 Description: open req refcounting wrong on reconnect
752 Details : If reconnect happened between getting open reply from server and
753 call to mdc_set_replay_data in ll_file_open, we will schedule
754 replay for unreferenced request that we are about to free.
755 Subsequent close will crash in variety of ways.
756 Check that request is still eligible for replay in
757 mdc_set_replay_data().
762 Description: disable writes to filesystem when reading health_check file
763 Details : the default for reading the health_check proc file has changed
764 to NOT do a journal transaction and write to disk, because this
765 can cause reads of the /proc file to hang and block HA state
766 checking on a healthy but otherwise heavily loaded system. It
767 is possible to return to the previous behaviour during configure
768 with --enable-health-write.
770 Severity : enhancement
772 Description: 64-bit inode version
773 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
774 to the inode. This will be required for version-based recovery.
779 Description: MDS fails to start if a duplicate client export is detected
780 Details : in some rare cases it was possible for a client to connect to
781 an MDS multiple times. Upon recovery the MDS would detect this
782 and fail during startup. Handle this more gracefully.
784 Severity : enhancement
786 Description: Add -o localflock option to simulate old noflock
788 Details : This will achieve local-only flock/fcntl locks
794 Description: log_commit_thread vs filter_destroy race leads to crash
795 Details : Take import reference before releasing llog record semaphore
800 Description: Wrong request locking in request set processing
801 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
802 imp_delayed_list, in this place should be used imp_lock.
805 Frequency : when reconnection
807 Description: Grant Leak when osc reconnect to OST
808 Details : When osc reconnect ost, OST(filter) should check whether it
809 should grant more space to client by comparing fed_grant and
810 cl_avail_grant, and return the granted space to client instead
811 of "new granted" space, because client will call osc_init_grant
812 to update the client grant space info.
815 Frequency : when client reconnect to OST
817 Description: Grant Leak when osc do resend and replay bulk write
818 Details : When osc reconnect to OST, OST(filter)should clear grant info of
819 bulk write request, because the grant info while be sync between
820 OSC and OST when reconnect, and we should ignore the grant info
821 these of resend/replay write req.
826 Description: Grant space more than avaiable left space sometimes.
827 Details : When then OST is about to be full, if two bulk writing from
828 different clients came to OST. Accord the avaliable space of the
829 OST, the first req should be permitted, and the second one
830 should be denied by ENOSPC. But if the seconde arrived before
831 the first one is commited. The OST might wrongly permit second
832 writing, which will cause grant space > avaiable space.
835 Frequency : when client is evicted
837 Description: Grant might be wrongly erased when osc is evicted by OST
838 Details : when the import is evicted by server, it will fork another
839 thread ptlrpc_invalidate_import_thread to invalidate the
840 import, where the grant will be set to 0. While the original
841 thread will update the grant it got when connecting. So if
842 the former happened latter, the grant will be wrongly errased
843 because of this race.
848 Description: Checking Stale with correct fid
849 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
850 to check whether it is stale, because sometimes, we want the
851 enqueue happened anyway, and op_data.fid2 will not be initialized.
853 Severity : enhancement
855 Description: update patchless client
856 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
859 Frequency : only with 2.4 kernel
861 Description: random memory corruption
862 Details : size of struct ll_inode_info is to big for union inode.u and this
863 can be cause of random memory corruption.
868 Description: Memory leak in recovery
869 Details : Lov_mds_md was not free in an error handler in mds_create_object.
870 It should also check obd_fail before fsfilt_start, otherwise if
871 fsfilt_start return -EROFS,(failover mds during mds recovery).
872 then the req will return with repmsg->transno = 0 and rc = EROFS.
873 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
874 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
875 be freed no matter whether fsfilt_commit success or not.
878 Frequency : only with huge count clients
880 Description: Prevents from taking the superblock lock in llap_from_page for
882 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
883 for llap_from_page call in ll_removepage prevents from taking the
884 superblock lock for a soon died page.
889 Description: Not check open intent error before release open handle
890 Details : in some rare cases, the open intent error is not checked before
891 release open handle, which may cause
892 ASSERTION(open_req->rq_transno != 0), because it tries to release
893 the failed open handle.
898 Description: Set cat log bitmap only after create log success.
899 Details : in some rare cases, the cat log bitmap is set too early. and it
900 should be set only after create log success.
904 Description: Accessing a block bevice can re-enable I/O when Lustre is
905 tearing down a device.
906 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
911 Description: service threads may hog cpus when there are a lot of requests
912 Details : Insert cond_resched to give other threads a chance to use some CPU
917 Description: the cat log was not initialized in recovery
918 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
919 unlink log on mds will not be initialized until mds post
920 recovery. And also in mds post recovery, the unlink log will
921 initialization will be done asynchronausly, so there will be race
922 between add unlink log and unlink log initialization.
926 Description: brw_stats were being printed incorrectly
927 Details : brw_stats were being printed as log2 but all of them were not
928 recorded as log2. Also remove some code duplication arising from
929 filter_tally_{read,write}.
933 Frequency : rare, only in recovery.
934 Description: ASSERTION(req->rq_type != LI_POISON) failed
935 Details : imp_lock should be held while iterating over imp_sending_list for
936 prevent destroy request after get timeout in ptlrpc_queue_wait.
940 Description: replay-single.sh test 52 fails
941 Details : A lock's skiplist need to be cleanup when it being unlinked
942 from its resource list.
946 Description: Short directio read returns full requested size rather than
948 Details : Direct I/O operations should return actual amount of bytes
949 transferred rather than requested size.
951 Severity : enhancement
953 Description: metadata RPC reduction (e.g. for rm performance)
954 Details : decrease the amount of synchronous RPC between clients and servers
955 by canceling conflicing lock before the operation on the client
956 and packing thier handles into the main operation RPC to server.
958 Severity : enhancement
960 Description: Async OSC create to avoid the blocking unnecessarily.
961 Details : If a OST has no remain object, system will block on the creating
962 when need to create a new object on this OST. Now, ways use
963 pre-created objects when available, instead of blocking on an
964 empty osc while others are not empty. If we must block, we block
965 for the shortest possible period of time.
967 Severity : enhancement
969 Description: refine locking for avoid write wrong info into lov_objid file
970 Details : fix possible races with add new target and write/update data in
973 --------------------------------------------------------------------------------
975 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
980 Frequency : on some architectures
982 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
983 Details : Uninitialized flags sometimes cause configuration commands to
987 Frequency : patchless clients only
989 Description: missing __iget() symbol export
990 Details : The __iget() symbol export is missing. To avoid the need for
991 this on patchless clients the deathrow inode reaper is turned
992 off, and we depend on the VM to clean up old inodes. This
993 dependency was during via the fix for bug 12181.
995 --------------------------------------------------------------------------------
997 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
999 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1000 INTEROPERATE with older versions automatically. In many cases a
1001 special upgrade step is needed. Please read the
1002 user documentation before upgrading any part of a live system.
1003 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
1004 version will not interoperate with 1.6 betas before beta5 (1.5.95).
1005 * WARNING: Lustre configuration and startup changes are required with
1006 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1012 Severity : enhancement
1014 Description: MountConf
1015 Details : Lustre configuration is now managed via mkfs and mount
1016 commands instead of lmc and lconf. New obd types (MGS, MGC)
1017 are added for dynamic configuration management. See
1018 https://mail.clusterfs.com/wikis/lustre/MountConf for
1021 Severity : enhancement
1023 Description: dynamic OST addition
1024 Details : OSTs can now be added to a live filesystem
1026 Severity : enhancement
1028 Description: startup order invariance
1029 Details : MDTs and OSTs can be started in any order. Clients only
1030 require the MDT to complete startup.
1032 Severity : enhancement
1034 Description: parallel, asynchronous orphan cleanup
1035 Details : orphan cleanup is now performed in separate threads for each
1036 OST, allowing parallel non-blocking operation.
1038 Severity : enhancement
1040 Description: optimized stripe assignment
1041 Details : stripe assignments are now made based on ost space available,
1042 ost previous usage, and OSS previous usage, in order to try
1043 to optimize storage space and networking resources.
1045 Severity : enhancement
1047 Description: Permanently set tunables
1048 Details : All writable /proc/fs/lustre tunables can now be permanently
1049 set on a per-server basis, at mkfs time or on a live system.
1051 Severity : enhancement
1053 Description: Lustre message v2
1054 Details : Add lustre message format v2.
1056 Severity : enhancement
1058 Description: client OST exclusion list
1059 Details : Clients can be started with a list of OSTs that should be
1060 declared "inactive" for known non-responsive OSTs.
1064 Description: SPEC SFS validation failure on NFS v2 over lustre.
1065 Details : Changes the blocksize for regular files to be 2x RPC size,
1066 and not depend on stripe size.
1068 Severity : enhancement
1070 Description: Multiple MD RPCs in flight.
1071 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1072 To avoid overly-overloading MDS, introduce a limit on number of
1073 MDS RPCs in flight for a single client and add /proc controls
1074 to adjust this limit.
1076 Severity : enhancement
1078 Description: client read/write statistics
1079 Details : Add client read/write call usage stats for performance
1080 analysis of user processes.
1081 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1082 file access. extents_stats shows chunk size distribution.
1083 extents_stats_per_process show chunk size distribution per
1086 Severity : enhancement
1088 Description: mds statistics
1089 Details : Add detailed mds operations statistics in
1090 /proc/fs/lustre/mds/*/stats.
1094 Description: Failure of copying files with lustre special EAs.
1095 Details : Client side always return success for setxattr call for lustre
1096 special xattr (currently only "trusted.lov").
1101 Description: Refcount LNET uuids
1102 Details : The global LNET uuid list grew linearly with every startup;
1103 refcount repeated list entries instead of always adding to
1106 Severity : enhancement
1108 Description: Dynamic service threads
1109 Details : Within a small range, start extra service threads
1110 automatically when the request queue builds up.
1113 Frequency : mixed-endian client/server environments
1115 Description: mixed-endian crashes
1116 Details : The new msg_v2 system had some failures in mixed-endian
1119 Severity : enhancement
1121 Description: Easy OST removal
1122 Details : OSTs can be permanently deactivated with e.g. 'lctl
1123 conf_param lustre-OST0001.osc.active=0'
1125 Severity : enhancement
1127 Description: MGS proc entries
1128 Details : Added basic proc entries for the MGS showing what filesystems
1131 Severity : enhancement
1133 Description: provide MGS failover
1134 Details : Added config lock reacquisition after MGS server failover.
1136 Severity : enhancement
1138 Description: add Linux 2.4 support
1139 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1143 Description: a large application tries to do I/O to the same resource and dies
1144 in the middle of it.
1145 Details : Check the req->rq_arrival time after the call to
1146 ost_brw_lock_get(), but before we do anything about
1147 processing it & sending the BULK transfer request. This
1148 should help move old stale pending locks off the queue as
1149 quickly as obd_timeout.
1152 Frequency : when an incorrect nid is specified during startup
1154 Description: ptlrpc connect to non-existant node causes kernel crash
1155 Details : LNET can't be re-entered from an event callback, which
1156 happened when we expire a message after the export has been
1157 cleaned up. Instead, hand the zombie cleanup off to another
1160 Severity : enhancement
1162 Description: plain/inodebits lock performance improvement
1163 Details : Grouping plain/inodebits in granted list by their request modes
1164 and bits policy, thus improving the performance of search through
1168 Frequency : only if OST filesystem is corrupted
1170 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1171 Details : for a short time RPCs with bulk IO are in the replay list,
1172 but replay of bulk IOs is unimplemented. If the OST filesystem
1173 is corrupted due to disk cache incoherency and then replay is
1174 started it is possible to trip an assertion. Avoid putting
1175 committed RPCs into the replay list at all to avoid this issue.
1178 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1180 Description: System hang on startup
1181 Details : This bug allowed the liblustre (e.g. catamount) client to
1182 return to the app before handling all startup RPCs. This
1183 could leave the node unresponsive to lustre network traffic
1184 and manifested as a server ptllnd timeout.
1186 Severity : enhancement
1188 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1189 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1190 boolean to control whether to print peer debug info when a
1191 client's RPC times out.
1194 Frequency : only for kernels with patches from Lustre below 1.4.3
1196 Description: Remove old rdonly API
1197 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1200 Frequency : only for devices with external journals
1202 Description: Set external device read-only also
1203 Details : During a commanded failover stop, we set the disk device
1204 read-only while the server shuts down. We now also set any
1205 external journal device read-only at the same time.
1208 Frequency : when upgrading from 1.4 while trying to change parameters
1210 Description: The wrong (new) MDC name was used when setting parameters for
1211 upgraded MDT's. Also allows changing of OSC (and MDC)
1212 parameters if --writeconf is specified at tunefs upgrade time.
1215 Frequency : when setting specific ost indicies
1217 Description: QOS code breaks on skipped indicies
1218 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1219 created with --index need not be sequential.
1221 Severity : enhancement
1223 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1224 Details : The uninit_groups feature works in conjunction with the kernel
1225 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1226 up the pass1 processing of e2fsck. This is a read-only feature
1227 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1228 mount filesystems that have had this feature enabled.
1230 Severity : enhancement
1232 Description: Improve multi-block allocation algorithm to avoid fragmentation
1233 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1234 allocation locality and avoid filesystem fragmentation.
1236 ------------------------------------------------------------------------------
1238 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1240 * Support for kernels:
1241 2.4.21-47.0.1.EL (RHEL 3)
1242 2.6.5-7.283 (SLES 9)
1243 2.6.9-42.0.10.EL (RHEL 4)
1244 2.6.12.6 vanilla (kernel.org)
1245 2.6.16.27-0.9 (SLES 10)
1246 * Recommended e2fsprogs version: 1.39.cfs5
1248 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1252 Frequency : occasional, depends on client load and configuration
1253 Bugzilla : 12181, 12203
1254 Description: data loss for recently-modified files
1256 Details : In some cases it is possible that recently written or created
1257 files may not be written to disk in a timely manner (this should
1258 normally be within 30s unless client IO load is very high).
1259 The problem appears as zero-length files or files that are a
1260 multiple of 1MB in size after a client crash or client eviction
1261 that are missing data at the end of the file.
1263 This problem is more likely to be hit on clients where files are
1264 repeatedly created and unlinked in the same directory, clients
1265 have a large amount of RAM, have many CPUs, the filesystem has
1266 many OSTs, the clients are rebooted frequently, and/or the files
1267 are not accessed by other nodes after being written.
1269 The presence of the problem can be detected by looking at
1270 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1271 smaller than the second (nr_unused) then dirty files will not
1272 be flushed automatically to disk. "sync; sleep 10" should be
1273 run several times on the node before unmounting it to update
1274 Lustre (this is also safe to run on nodes without this problem).
1276 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1277 that can cause this same problem, so customers using that kernel
1278 also need to update the kernel in addition to Lustre. In order
1279 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1281 It is normal that files written just before a client crash (less
1282 than 30s) may not yet have been flushed to disk, even for local
1286 Frequency : frequent on thin XT3 nodes
1288 Description: UUID collision on thin XT3 Linux nodes
1289 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1290 randomly, since we relied on an insufficiently-seeded PRNG.
1295 Description: OSS hangs after "All ost request buffers busy"
1296 Details : A deadlock between quota and journal operations caused OSS
1297 hangs after printing "All ost request buffers busy."
1300 Frequency : always on liblustre builds
1302 Description: Cleanup compiler warnings on liblustre
1305 Frequency : always on liblustre builds on XT3
1307 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1311 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1312 > 32000 subdirectories
1313 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1314 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1315 subdirectory count crosses 32000. This will aid e2fsck to
1316 correctly handle more than 32000 subdirectories.
1319 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1321 Description: System hang on startup
1322 Details : This bug allowed the liblustre (e.g. catamount) client to
1323 return to the app before handling all startup RPCs. This
1324 could leave the node unresponsive to lustre network traffic
1325 and manifested as a server ptllnd timeout.
1327 Severity : enhancement
1329 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1330 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1331 boolean to control whether to print peer debug info when a
1332 client's RPC times out.
1337 Description: make O_SYNC working on 2.6 kernels
1338 Details : 2.6 kernels use different method for mark pages for write,
1339 so need add a code to lustre for O_SYNC work.
1344 Description: Failure to close file and release space on NFS
1345 Details : Put inode details into lock acquired in ll_intent_file_open.
1346 Use mdc_intent_lock in ll_intent_open to properly
1347 detect all kind of errors unhandled by mdc_enqueue.
1352 Description: proc file read during shutdown sometimes raced obd removal,
1354 Details : Add lock to prevent obd access after proc file removal.
1357 Frequency : Only for files larger than 4GB on 32-bit clients.
1359 Description: improperly doing page alignment of locks
1360 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1361 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1366 Description: RPCs being resent when they shouldn't be
1367 Details : Some RPCs that should not be resent are being resent. This
1368 can cause inconsistencies in the RPC state machine. Do not
1369 resend such requests.
1372 Frequency : rare, only with NFS export
1374 Description: Crash on NFS re-export node
1375 Details : under very unusual load conditions an assertion is hit in
1376 ll_intent_file_open()
1379 Frequency : only if OST filesystem is corrupted
1381 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1382 Details : for a short time RPCs with bulk IO are in the replay list,
1383 but replay of bulk IOs is unimplemented. If the OST filesystem
1384 is corrupted due to disk cache incoherency and then replay is
1385 started it is possible to trip an assertion. Avoid putting
1386 committed RPCs into the replay list at all to avoid this issue.
1391 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1392 Details : Large single O_DIRECT read and write calls can fail to allocate
1393 a sufficiently large buffer to process the request. In case of
1394 allocation failure the allocation is retried with a smaller
1395 buffer and broken into smaller requests.
1397 Severity : enhancement
1399 Description: Add -o localflock option to simulate old noflock behaviour.
1400 Details : This will achieve local-only flock/fcntl locks coherentness.
1405 Description: versioning check is incomplete
1406 Details : Checking the version difference of client vs. server, report
1407 error if the gap is too big.
1412 Description: add support PG_writeback bit
1413 Details : add support for PG_writeback bit for Lustre, for more carefull
1414 work with page cache in 2.6 kernel. This also fix some deadlocks
1415 and remove hack for work O_SYNC with 2.6 kernel.
1417 Severity : enhancement
1419 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1420 Details : The uninit_groups feature works in conjunction with the kernel
1421 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1422 up the pass1 processing of e2fsck. This is a read-only feature
1423 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1424 mount filesystems that have had this feature enabled.
1426 Severity : enhancement
1428 Description: Improve multi-block allocation algorithm to avoid fragmentation
1429 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1430 allocation locality and avoid filesystem fragmentation.
1432 ------------------------------------------------------------------------------
1434 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1436 * Support for kernels:
1437 2.6.9-42.0.3.EL (RHEL 4)
1438 2.6.5-7.276 (SLES 9)
1439 2.4.21-47.0.1.EL (RHEL 3)
1440 2.6.12.6 vanilla (kernel.org)
1441 2.6.16.21-0.8 (SLES10)
1442 * Recommended e2fsprogs version: 1.39.cfs2-0
1444 * The backwards-compatible /proc/sys/portals symlink has been removed
1445 in this release. Before upgrading, please ensure that you change
1446 any configuration scripts or /etc/sysctl.conf files that access
1447 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1448 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1449 in advance of the upgrade on any system running Lustre 1.4.6 or
1450 newer, since /proc/sys/lnet was added in that version.
1451 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1455 Frequency : only when quota is used
1457 Description: avoid scanning export list for quota master
1458 Details : Change the algorithms to avoid scanning export list in order
1459 to improve the efficiency.
1462 Frequency : MDS failover only, very rarely
1464 Description: "went back in time" messages on mds failover
1465 Details : The greatest transno may be lost when the current operation
1466 finishes with an error (transno==0) and the client's last_rcvd
1467 record is over-written. Save the greatest transno in the
1468 mds_last_transno for this case.
1471 Frequency : always for specific kernels and striping counts
1473 Description: client may get "Matching packet too big" without ACL support
1474 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1475 when trying to access files in certain configurations. The
1476 clients should in fact be denied when mounting because they do
1477 not understand ACLs.
1480 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1482 Description: many clients connecting with IO in progress causes connect timeouts
1483 Details : Avoid synchronous journal commits to avoid delays caused by many
1484 clients connecting/disconnecting when bulk IO is in progress.
1485 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1486 OST_IO_PORTAL to avoid delays behind potentially many pending
1490 Frequency : occasionally with multiple writers to a single file
1492 Description: shared writes to file may result in wrong size reported by stat()
1493 Details : Allow growing of kms when extent lock is cancelled
1496 Frequency : always with random mmap IO to multi-striped file
1498 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1499 Details : Only if the hole is at the end of OST object so that kms is too
1500 small. Fix is to increase kms accordingly in ll_nopage.
1503 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1505 Description: writes to a missing object would leak memory on the OST
1506 Details : If there is an inconsistency between the MDS and OST filesystems,
1507 such that the MDS references an object that doesn't exist, writes
1508 to that object will leak memory due to incorrect cleanup in the
1509 error handling path, eventually running out of memory on the OST.
1514 Description: Creating too long symlink causes lustre errors
1515 Details : Check symlink and name lengths before sending requests to MDS.
1518 Frequency : only if flock is enabled (not on by default)
1520 Description: posix locks not released on fd closure on 2.6.9+
1521 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1522 kernels, this caused such locks not to be released on fd close and
1523 then assertions on fs unmount about still used locks.
1526 Frequency : MDS failover only, very rarely
1528 Description: clients may get ASSERTION(granted_lock != NULL)
1529 Details : When request was taking a long time, and a client was resending
1530 a getattr by name lock request. The were multiple lock requests
1531 with the same client lock handle and
1532 mds_getattr_name->fixup_handle_for_resent_request found one of the
1533 lock handles but later failed with ASSERTION(granted_lock != NULL).
1538 Description: handle->h_buffer_credits > 0, assertion failure
1539 Details : h_buffer_credits is zero after truncate, causing assertion
1540 failure. This patch extends the transaction or creates a new
1544 Frequency : NFS re-export or patchless client
1545 Bugzilla : 11179, 10796
1546 Description: Crash on NFS re-export node (__d_move)
1547 Details : We do not want to hash the dentry if we don't have a lock.
1548 But if this dentry is later used in d_move, we'd hit uninitialised
1549 list head d_hash, so we just do this to init d_hash field but
1550 leave dentry unhashed.
1553 Frequency : NFS re-export or patchless client
1555 Description: NFS exports has problem with symbolic link
1556 Details : lustre client didn't properly install dentry when re-exported
1557 to NFS or running patchless client.
1560 Frequency : NFS re-export or patchless client
1562 Description: Various nfs/patchless fixes.
1563 Details : fixes reuse disconected alias for lookup process - this fixes
1564 warning "find_exported_dentry: npd != pd",
1565 fix permission error with open files at nfs.
1566 fix apply umask when do revalidate.
1569 Frequency : occasional
1571 Description: Crash on NFS re-export node
1572 Details : calling clear_page() on the wrong pointer triggered oops in
1573 generic_mapping_read().
1576 Frequency : rarely, using O_DIRECT IO
1578 Description: unaligned directio crashes client with LASSERT
1579 Details : check for unaligned buffers before trying any requests.
1582 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1584 Description: stale data returned from RAID cache
1585 Details : If only a small amount of IO is done to the RAID device before
1586 reading it again it is possible to get stale data from the RAID
1587 cache instead of reading it from disk.
1590 Frequency : always for sles10 kernel
1592 Description: sles10 support
1593 Details : ll_follow_link: compile fixes and using of nd_set_link
1594 under newer kernels.
1597 Frequency : depends on arch, kernel and compiler version, always on sles10
1600 Description: recursive or deep enough symlinks cause stack overflow
1601 Details : getting rid of large stack-allocated variable in
1605 Frequency : depends on hardware
1607 Description: lustre write performance loss in the SLES10 kernel
1608 Details : the performance loss is caused by using of write barriers in the
1609 ext3 code. The SLES10 kernel turns barrier support on by
1610 default. The fix is to undo that change for ldiskfs.
1612 ------------------------------------------------------------------------------
1614 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1616 * Support for kernels:
1617 2.6.9-42.0.3EL (RHEL 4)
1618 2.6.5-7.276 (SLES 9)
1619 2.4.21-47.0.1.EL (RHEL 3)
1620 2.6.12.6 vanilla (kernel.org)
1624 Frequency : quota enabled and large files being deleted
1626 Description: releasing more than 4GB of quota at once hangs OST
1627 Details : If a user deletes more than 4GB of files on a single OST it
1628 will cause the OST to spin in an infinite loop. Release
1629 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1634 Description: statfs data retrieved from /proc may be stale or zero
1635 Details : When reading per-device statfs data from /proc, in the
1636 {kbytes,files}_{total,free,avail} files, it may appear
1637 as zero or be out of date.
1640 Frequency : systems with MD RAID1 external journal devices
1642 Description: lconf's call to blkid is confused by RAID1 journal devices
1643 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1644 of returning all block devices that match the journal UUID.
1647 Frequency : always, for aggregate stripe size over 4GB
1649 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1650 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1651 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1654 Frequency : always if "lfs find" used on a local file/directory
1656 Description: "lfs find" segfaults if used on a local file/directory
1657 Details : The case where a directory component was not specified wasn't
1658 handled correctly. Handle this properly.
1661 Frequency : always on ppc64
1663 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1664 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1665 perform endianness conversion.
1668 Frequency : rarely (truncate to non-zero file size after write under load)
1669 Bugzilla : 10730, 10687
1670 Description: Files padded with zeros to next 4K multiple
1671 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1672 it is possible that files that are truncated to a non-zero size
1673 immediately after being written are filled with zero bytes beyond
1674 the truncated size. No file data is lost.
1676 Severity : enhancement
1678 Description: Allow recovery/failover for liblustre clients.
1679 Details : liblustre clients were unaware of failover configurations until
1682 Severity : enhancement
1684 Description: user file locks should fail when not mounting with flock option
1685 Details : Set up an error-returning stub in ll_file_operations.lock field
1686 to prevent incorrect behaviour when client is mounted without
1687 flock option. Also, set up properly f_op->flock field for
1691 Frequency : always on ia64
1693 Description: "lfs df" loops on printing out MDS statfs information
1694 Details : The obd_ioctl_data was not initialized and in some systems
1695 this caused a failure during the ioctl that did not return
1696 an error. Initialize the struct and return an error on failure.
1699 Frequency : SLES 9 only
1701 Description: Error of copying files with lustre special EAs as root
1702 Details : Client side always return success for setxattr call for lustre
1703 special xattr (currently only "trusted.lov").
1706 Frequency : rarely on clusters with both ia64+i386 clients
1708 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1709 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1710 clients are doing concurrent writes to the same file it is
1711 possible that the ia64 clients may LASSERT because the OST
1712 extent locks are not PAGE_SIZE aligned. Ensure that grown
1713 locks are always aligned on the request boundary.
1716 Frequency : specific use, occasional
1718 Description: Overwriting in use executable truncates on-disk binary image
1719 Details : If one node attempts to overwrite an executable in use by
1720 another node, we now correctly return ETXTBSY instead of
1721 truncating the file.
1723 Severity : enhancement
1725 Description: Async OSC create to avoid the blocking unnecessarily.
1726 Details : If a OST has no remain object, system will block on the creating
1727 when need to create a new object on this OST. Now, ways use
1728 pre-created objects when available, instead of blocking on an
1729 empty osc while others are not empty. If we must block, we block
1730 for the shortest possible period of time.
1735 Description: chmod on Lustre root is propagated to other clients
1736 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1737 invalid by the follow_mount time.
1740 Frequency : liblustre clients only
1742 Description: Race in 'instant cancel' lock handling could lead to such locks
1743 never to be granted in case of SMP MDS
1744 Details : Do not destroy not yet granted but cbpending locks in
1748 Frequency : replay/resend of open
1750 Description: non null lock assetion failure in mds_intent_policy
1751 Details : Trying to replay/resend lockless open requests resulted in
1752 mds_open() returning 0 with no lock. Now it sets a flag if
1753 a lock is going to be returned.
1755 Severity : enhancement
1757 Description: Checksum enhancements
1758 Details : New checksum enhancements allow for resending RPCs that failed
1761 Severity : enhancement
1763 Description: Tunables on number of dirty pages in cacche
1764 Details : Allow to set limit on number of dirty pages cached.
1769 Description: client crash on unmount - lock still has references
1770 Details : In some error handling cases it was possible to leak a lock
1771 reference on a client while accessing a file. This was not
1772 harmful to the client during operation, but would cause the
1773 client to crash when the filesystem is unmounted.
1776 Frequency : specific case, rare
1778 Description: ETXTBSY on mds though file not in use by client
1779 Details : ETXTBSY is no longer incorrectly returned when attempting to
1780 chmod or chown a directory that the user previously tried to
1781 execute or a currently-executing binary.
1784 Frequency : extremely rare except on liblustre-based clients
1786 Description: Lustre space not freed when files are deleted
1787 Details : Clean up open-unlinked files after client eviction. Previously
1788 the unlink was skipped and the files remained as orphans.
1793 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1794 Details : In some cases it was possible to send a blocking callback to
1795 a client doing a glimpse, even though that client didn't get
1796 a lock granted. When the glimpse lock is cancelled on the OST
1797 the freed lock is left on the waiting list and corrupted the list.
1800 Frequency : all core dumps
1802 Description: Broke core dumps to lustre
1803 Details : Negative dentry may be unhashed if parent does not have UPDATE
1804 lock, but some callers, e.g. do_coredump, expect dentry to be
1805 hashed after successful create, hash it in ll_create_it.
1807 ------------------------------------------------------------------------------
1809 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1811 * Support for kernels:
1812 2.6.9-42.0.2.EL (RHEL 4)
1813 2.6.5-7.276 (SLES 9)
1814 2.4.21-40.EL (RHEL 3)
1815 2.6.12.6 vanilla (kernel.org)
1819 Frequency : always on RHEL 3
1821 Description: Number of open files grows over time
1822 Details : The number of open files grows over time, whether or not
1823 Lustre is started. This was due to a filp leak introduced
1824 by one of our kernel patches.
1826 ------------------------------------------------------------------------------
1828 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1830 * Support for kernels:
1831 2.6.9-42.EL (RHEL 4)
1832 2.6.5-7.276 (SLES 9)
1833 2.4.21-40.EL (RHEL 3)
1834 2.6.12.6 vanilla (kernel.org)
1839 Bugzilla : 5719, 9635, 9792, 9684
1840 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1841 Details : If a server is under heavy load and cannot reply to new
1842 connection requests before the client resends the (re)connect,
1843 the connection handling code can behave badly if two service
1844 threads are concurrently handing separate (re)connections from
1845 the same client. Add better locking to the connection handling
1846 code, and ensure that only a single connection will be processed
1847 for a given client UUID, even if the lock is dropped.
1849 Severity : enhancement
1851 Description: add TCP zero-copy support to kernel
1852 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1853 sends if the hardware supports scatter-gather and checksumming.
1854 This allows socklnd to do client-write and server-read more
1855 efficiently and reduce CPU utilization from skbuf copying.
1858 Frequency : only if NFS exporting from client
1860 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1861 Details : If setting attributes on a file created under NFS that had
1862 never been opened it would be possible to oops the client
1863 if the file had no objects.
1866 Frequency : always for liblustre
1868 Description: liblustre client does MDS+OSTs setattr RPC for each write
1869 Details : When doing a write from a liblustre client, the client
1870 incorrectly issued an RPC to the MDS and each OST the file was
1871 striped over in order to update the timestamps. When writing
1872 with small chunks and many clients this could overwhelm the MDS
1873 with RPCs. In all cases it would slow down the write because
1874 these RPCs are unnecessary.
1876 Severity : enhancement
1878 Description: allow number of MDS service threads to be changed at module load
1879 Details : It is now possible to change the number of MDS service threads
1880 running. Adding "options mds mds_num_threads={N}" to the MDS's
1881 /etc/modprobe.conf will set the number of threads for the next
1882 time Lustre is restarted (assuming the "mds" module is also
1883 reloaded at that time). The default number of threads will
1884 stay the same, 32 for most systems.
1889 Description: OST crash if filesystem is unformatted or corrupt
1890 Details : If an OST is started on a device that has never been formatted
1891 or if the filesystem is corrupt and cannot even mount then the
1892 error handling cleanup routines would dereference a NULL pointer.
1897 Description: NULL pointer deref in llap_from_page.
1898 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1899 incorrect) mapping if the page was truncated/reclaimed while it was
1900 searched for. Check for this condition and skip such pages when
1901 doing readahead. Introduce extra check to llap_from_page() to
1902 verify page->mapping->host is non-NULL (so page is not anonymous).
1905 Frequency : Sometimes when using sys_sendfile
1907 Description: "page not covered by a lock" warnings from ll_readpage
1908 Details : sendfile called ll_readpage without right page locks present.
1909 Now we introduced ll_file_sendfile that does necessary locking
1910 around call to generic_file_sendfile() much like we do in
1914 Frequency : with certain MDS communication failures at client mount time
1916 Description: NULL pointer deref after failed client mount
1917 Details : a client connection request may delayed by the network layer
1918 and not be sent until after the PTLRPC layer has timed out the
1919 request. If the client fails the mount immediately it will try
1920 to clean up before the network times out the request. Add a
1921 reference from the request import to the obd device and delay
1922 the cleanup until the network drops the request.
1925 Frequency : occasionally during client (re)connect
1927 Description: assertion failure during client (re)connect
1928 Details : processing a client connection request may be delayed by the
1929 client or server longer than the client connect timeout. This
1930 causes the client to resend the connection request. If the
1931 original connection request is replied in this interval, the
1932 client may trip an assertion failure in ptlrpc_connect_interpret()
1933 which thought it would be the only running connect process.
1936 Frequency : only with obd_echo servers and clients that are rebooted
1938 Description: kernel BUG accessing uninitialized data structure
1939 Details : When running an obd_echo server it did not start the ping_evictor
1940 thread, and when a client was evicted an uninitialized data
1941 structure was accessed. Start the ping_evictor in the RPC
1942 service startup instead of the OBD startup.
1944 Severity : enhancement
1945 Bugzilla : 10193 (patchless)
1946 Description: Remove dependency on various unexported kernel interfaces.
1947 Details : No longer need reparent_to_init, exit_mm, exit_files,
1948 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1951 Frequency : rare (only users of deprecated and unsupported LDAP config)
1953 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1954 Details : LDAP apparently contains 'lustreName' attributes instead of
1955 'name'. A simple remapping of the name is sufficient.
1958 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1960 Description: waiting_locks_callback trips kernel BUG if client is evicted
1961 Details : Running with the dump_on_timeout debug flag turned on makes
1962 it possible that the waiting_locks_callback() can try to dump
1963 the Lustre kernel debug logs from an interrupt handler. Defer
1964 this log dumping to the expired_lock_main() thread.
1966 Severity : enhancement
1968 Description: Support NFS exporting on 2.6 kernels.
1969 Details : Implement non-rawops metadata methods for NFS server to use without
1970 changing NFS server code.
1973 Frequency : very rare (synthetic metadata workload only)
1975 Description: two racing renames might cause an MDS thread to deadlock
1976 Details : Running the "racer" program may cause one MDS thread to rename
1977 a file from being the source of a rename to being the target of
1978 a rename at exactly the same time that another thread is doing
1979 so, and the second thread has already enqueued these locks after
1980 doing a lookup of the target and is trying to relock them in
1981 order. Ensure that we don't try to re-lock the same resource.
1984 Frequency : only very large systems with liblustre clients
1986 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1987 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1988 all OSTs in parallel. This allows the eviction of stale liblustre
1989 clients to proceed much faster than if they were done in series,
1990 and also offers similar improvements for other set_info RPCs.
1995 Description: excessive CPU usage during initial read phase on client
1996 Details : During the initial read phase on a client, it would agressively
1997 retry readahead on the file, consuming too much CPU and impacting
1998 performance (since 1.4.5.8). Improve the readahead algorithm
1999 to avoid this, and also improve some other common cases (read
2000 of small files in particular, where "small" is files smaller than
2001 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2006 Description: MDS crash when receiving packet with unknown intent.
2007 Details : Do not LBUG in unknown intent case, just return -EFAULT
2009 Severity : enhancement
2010 Bugzilla : 9293, 9385
2011 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2012 Details : Do not serialize getattr (non-intent version) and statfs.
2015 Frequency : occasional, when OST network is overloaded/intermittent
2017 Description: client evicted by OST after bulk IO timeout
2018 Details : If a client sends a bulk IO request (read or write) the OST
2019 may evict the client if it is unresposive to its data GET/PUT
2020 request. This is incorrect if the network is overloaded (takes
2021 too long to transfer the RPC data) or dropped the OST GET/PUT
2022 request. There is no need to evict the client at all, since
2023 the pinger and/or lock callbacks will handle this, and the
2024 client can restart the bulk request.
2027 Frequency : Always when mmapping file with no objects
2029 Description: client crashes when mmapping file with no objects
2030 Details : Check that we actually have objects in a file before doing any
2031 operations on objects in ll_vm_open, ll_vm_close and
2037 Description: Request leak when working with deleted CWD
2038 Details : Introduce advanced request refcount tracking for requests
2039 referenced from lustre intent.
2041 Severity : Enhancement
2043 Description: Cache open file handles on client.
2044 Details : MDS now will return special lock along with openhandle, if
2045 requested and client is allowed to hold openhandle, even if unused,
2046 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2047 closing files for every read/write openration.
2049 Severity : Enhancement
2051 Description: Cache open negative dentries on client when possible.
2052 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2053 negative dentries on lock revocation.
2058 Description: Remounting a client read-only wasn't possible with a zconf mount
2059 Details : It wasn't possible to remount a client read-only with llmount.
2061 Severity : enhancement
2062 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2063 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2064 MPICH is included to add Lustre-specific ADIO interfaces.
2065 This is based closely on the UFS ADIO layer and only differs
2066 in file creation, in order to allow the OST striping to be set.
2067 This is user-contributed code and not supported by CFS.
2072 Description: extended inode attributes (immutable, append-only) work improperly
2073 when 2.4 and 2.6 kernels are used on client/server or vice versa
2074 Details : Introduce kernel-independent values for these flags.
2076 Severity : enhancement
2079 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2080 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2081 granularity. Now, Fractional MB granularity is be supported,
2082 this is very useful for low memory system.
2084 Severity : enhancement
2086 Description: Getattr by fid
2087 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2088 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2089 have locks and avoids getattr rpc requests for it on every stat.
2092 Frequency : Always, for filesystems larger than 2TB
2094 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2095 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2096 filesystems larger than 2TB to be mounted with mballoc enabled.
2099 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2101 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2102 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2103 filesystems larger than 2TB is mounted then the kernel will
2104 silently corrupt the start of the filesystem. CONFIG_LBD is
2105 enabled for all CFS-supported kernels, but the possibility of
2106 this happening with a modified kernel config exists.
2108 Severity : enhancement
2110 Description: add client O_DIRECT support for 2.6 kernels
2111 Details : It is now possible to do O_DIRECT reads and writes to files
2112 in the Lustre client mountpoint on 2.6 kernel clients.
2114 Severity : enhancement
2116 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2117 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2118 parallel, not waiting for response from every OST before sending
2119 a rpc to the next OST.
2124 Description: setattr vs write race when updating file timestamps
2125 Details : Client processes that update a file timestamp into the past
2126 right after writing to the file (e.g. tar) it is possible that
2127 the updated file modification time can be reset to the current
2128 time due to a race between processing the setattr and write RPC.
2130 Severity : enhancement
2132 Description: Bring 'lfs find' closer in line with regular Linux find.
2133 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2134 -print0 options and obtains all the needed info through the lustre
2137 Severity : enhancement
2139 Description: support up to 1024 configured devices on one node
2140 Details : change obd_dev array from statically allocated to dynamically
2141 allocated structs as they are first used to reduce memory usage
2146 Description: Flush dirty partially truncated pages during truncate
2147 Details : Immediatelly flush partially truncated pages in filter_setattr,
2148 this way we completely avoid having any pages in page cache on OST
2149 and can retire ugly workarounds during writes to flush such pages.
2154 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2155 Details : protect inode from truncation within vfs_unlink() context
2156 just take a reference before calling vfs_unlink() and release it
2157 when parent's i_sem is free.
2162 Description: last_id value checked outside lock on OST caused LASSERT failure
2163 Details : If there were multiple MDS->OST object precreate requests in
2164 flight, it was possible that the OST's last object id was checked
2165 outside a lock and incorrectly tripped an assertion. Move checks
2166 inside locks, and discard old precreate requests.
2169 Frequency : always, if extents are used on OSTs
2171 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2172 Details : OSTs using the extents format would not zero the high 16 bits of
2173 the index physical block number. This is not a problem for any
2174 OST filesystems smaller than 16TB, and no kernels support ext3
2175 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2176 new/modified files) and can be fixed for existing filesystems
2177 with e2fsprogs-1.39-cfs1.
2182 Description: import connection selection may be incorrect if timer wraps
2183 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2184 import connections to be ignored if the 32-bit jiffies counter
2185 wraps. Use a 64-bit jiffies counter.
2188 Frequency : very large clusters immediately after boot
2190 Description: LNET request buffers exhausted under heavy short-term load
2191 Details : If a large number of client requests are generated on a service
2192 that has previously never seen so many requests it is possible
2193 that the request buffer growth cannot keep up with the spike in
2194 demand. Instead of dropping incoming requests, they are held in
2195 the LND until the RPC service can accept more requests.
2198 Frequency : Sometimes during replay
2200 Description: Assertion failure in ll_local_open after replay.
2201 Details : If replay happened on an open request reply before we were able
2202 to set replay handler, reply will become not swabbed tripping the
2203 assertion in ll_local_open. Now we set the handler right after
2204 recognising of open request
2207 Frequency : very rare
2209 Description: kernel reports "badness in vsnprintf"
2210 Details : Reading from the "recovery_status" /proc file in small chunks
2211 may cause a negative length in lprocfs_obd_rd_recovery_status()
2212 call to vsnprintf() (which is otherwise harmless). Exit early
2213 if there is no more space in the output buffer.
2215 Severity : enhancement
2217 Description: clear OBD RPC statistics by writing to them
2218 Details : It is now possible to clear the OBD RPC statistics by writing
2219 to the "stats" file.
2224 Description: Client mtime is not the same on different clients after utimes
2225 Details : In some cases, the client was using the utimes() syscall on
2226 a file cached on another node. The clients now validate the
2227 ctime from the MDS + OSTs to determine which one is right.
2232 Description: Inability to activate failout mode
2233 Details : lconf script incorrectly assumed that in pythong string's numeric
2234 value is used in comparisons.
2237 Frequency : always with multiple stripes per file
2239 Description: Inefficient object allocation for mutli-stripe files
2240 Details : When selecting which OSTs to stripe files over, for files with
2241 a stripe count that divides evenly into the number of OSTs,
2242 the MDS is always picking the same starting OST for each file.
2243 Return the OST selection heuristic to the original design.
2248 Description: mount failures may take full timeout to return an error
2249 Details : Under some heavy load conditions it is possible that a
2250 failed mount can wait for the full obd_timeout interval,
2251 possibly several minutes, before reporting an error.
2252 Instead return an error as soon as the status is known.
2254 Frequency : quota enabled and large files being deleted
2256 Description: releasing more than 4GB of quota at once hangs OST
2257 Details : If a user deletes more than 4GB of files on a single OST it
2258 will cause the OST to spin in an infinite loop. Release
2259 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2264 Description: statfs data retrieved from /proc may be stale or zero
2265 Details : When reading per-device statfs data from /proc, in the
2266 {kbytes,files}_{total,free,avail} files, it may appear
2267 as zero or be out of date.
2270 Frequency : systems with MD RAID1 external journal devices
2272 Description: lconf's call to blkid is confused by RAID1 journal devices
2273 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2274 of returning all block devices that match the journal UUID.
2277 Frequency : always, for aggregate stripe size over 4GB
2279 Description: assertion fails when trying to use 4GB stripe size
2280 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2281 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2284 Frequency : always on ppc64
2286 Description: the first write on an ext3 filesystem with mballoc got stuck
2287 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2288 perform endianness conversion.
2290 ------------------------------------------------------------------------------
2292 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2294 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2295 INTEROPERATE with older versions automatically. Please read the
2296 user documentation before upgrading any part of a live system.
2297 * WARNING: Lustre networking configuration changes are required with
2298 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2301 * Support for kernels:
2302 2.6.9-22.0.2.EL (RHEL 4)
2303 2.6.5-7.244 (SLES 9)
2304 2.6.12.6 vanilla (kernel.org)
2307 Severity : enhancement
2308 Bugzilla : 7981/8208
2309 Description: Introduced Lustre Networking (LNET)
2310 Details : LNET is new networking infrastructure for Lustre, it includes
2311 a reorganized network configuration mode (see the user
2312 documentation for full details) as well as support for routing
2313 between different network fabrics. Lustre Networking Devices
2314 (LNDs) for the supported network fabrics have also been
2315 created for this new infrastructure.
2317 Severity : enhancement
2318 Description: Introduced Access control lists
2319 Details : clients can set ACLs on files and directories in order to have
2320 more fine-grained permissions than the standard Unix UGO+RWX.
2321 The MDS must be started with the "-o acl" mount option.
2323 Severity : enhancement
2324 Description: Introduced filesystem quotas
2325 Details : Administrators may now establish per-user quotas on the
2328 Severity : enhancement
2330 Description: Configuration change for the XT3
2331 The PTLLND is now used to run Lustre over Portals on the XT3
2332 The configure option(s) --with-cray-portals are no longer used.
2333 Rather --with-portals=<path-to-portals-includes> is used to
2334 enable building on the XT3. In addition to enable XT3 specific
2335 features the option --enable-cray-xt3 must be used.
2340 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2341 Details : A race between allocating a new llcd and re-getting the llcd_lock
2342 allowed another thread to grab newly-allocated llcd.
2344 Severity : enhancement
2346 Description: 2.6 OST async journal commit and locking fix to improve performance
2347 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2348 2.6 kernels are now async as they already were in 2.4 kernels so
2349 that they can commit concurrently with the network bulk transfer.
2350 For block-allocated files the filter allocation semaphore is held
2351 to avoid filesystem fragmentation during allocation. BKL lock
2352 removed for 2.6 xattr operations where it is no longer needed.
2357 Description: lconf incorrectly determined whether two IP networks could talk
2358 Details : In some more complicated routing and multiple-network
2359 configurations, lconf will avoid trying to make a network
2360 connection to a disjoint part of the IP space. It was doing the
2361 math incorrectly for one set of cases.
2366 Description: Fix for potential infinite loop processing records in an llog.
2367 Details : If an llog record is corrupted/zeroed, it is possible to loop
2368 forever in llog_process(). Validate the llog record length
2369 and skip the remainder of the block on error.
2372 Frequency : occasional (liblustre only)
2374 Description: liblustre could not open files whose last component is a symlink
2375 Details : sysio_path_walk() would incorrectly pass the open intent to
2376 intermediate path components.
2379 Frequency : rare (liblustre only with non-standard tuning)
2380 Bugzilla : 7201 (7350)
2381 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2382 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2383 after it has been released, only do this for referenced locks
2385 Severity : enhancement
2387 Description: specifying an (invalid) directory default stripe_size of -1
2388 would reset the directory default striping
2389 Details : stripe_size -1 was used internally to signal directory stripe
2390 removal, now use "all default" to signal dir stripe removal
2391 as a directory striping of "all default" is not useful
2394 Frequency : common for large clusters running liblustre clients
2396 Description: doing an ls when liblustre clients are running is slow
2397 Details : sending a glimpse AST to a liblustre client waits for every AST
2398 to time out, as liblustre clients will not respond. Since they
2399 cannot cache data we refresh the OST lock LVB from disk instead.
2401 Severity : enhancement
2403 Description: doing an ls at the same time as file IO can be slow
2404 Details : enqueue and other "small" requests can be blocked behind many
2405 large IO requests. Create a new OST IO portal for non-IO
2406 requests so they can be processed faster.
2409 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2411 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2412 created file with mode 000
2413 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2414 allow an ftruncate() as a non-root user to a file with mode 000.
2415 Lustre now respects this flag to disable mode checks when
2416 truncating a file owned by the user
2419 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2422 Description: Revoking locks from clients that went dead or catatonic might take
2424 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2425 cancellation of such locks instant on servers without waiting for
2426 any reply from clients. Clients drops these locks when cancel
2427 notification from server is received without replying.
2430 Frequency : liblustre-only, when liblustre client dies or becomes busy
2432 Description: Doing ls on Linux clients can take a long time with active
2434 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2435 granting such locks to it in the first place if possible. Locks
2436 are taken by proxy on the OST during the read or write and
2437 dropped immediately afterward. Add connect flags handling, do
2438 not grant locks to liblustre clients for glimpse ASTs.
2440 Severity : enhancement
2442 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2443 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2444 sequential reads and behaves badly with applications doing random
2445 reads. Improve it by reading ahead at least read region, and
2446 avoiding excessive large RPC for small reads.
2448 Severity : enhancement
2450 Description: Creating more than 1000 files for a single job may cause a load
2451 imbalance on the OSTs if there are also a large number of OSTs.
2452 Details : qos_prep_create() uses an OST index reseed value that is an
2453 even multiple of the number of available OSTs so that if the
2454 reseed happens in the middle of the object allocation it will
2455 still utilize the OSTs as uniformly as possible.
2460 Description: OST or MDS may oops in ping_evictor_main()
2461 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2462 but doesn't restart at beginning of obd_exports_timed list
2465 Severity : enhancement
2467 Description: improve by-nid export eviction on the MDS and OST
2468 Details : allow multiple exports with the same NID to be evicted at one
2469 time without re-searching the exports list.
2472 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2474 Description: MDS may oops in groups_free()
2475 Details : in rare race conditions a newly allocated group_info struct is
2476 freed again, and this can be NULL. The 2.4 compatibility code
2477 for groups_free() checked for a NULL pointer, but 2.6 did not.
2480 Frequency : common for liblustre clients doing little filesystem IO
2481 Bugzilla : 9352, 7313
2482 Description: server may evict liblustre clients accessing contended locks
2483 Details : if a client is granted a lock or receives a completion AST
2484 with a blocking AST already set it would not reply to the AST
2485 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2488 Frequency : lfs setstripe, only systems with more than 160 OSTs
2490 Description: unable to set striping with a starting offset beyond OST 160
2491 Details : llapi_create_file() incorrectly limited the starting stripe
2492 index to the maximum single-file stripe count.
2495 Frequency : LDAP users only
2497 Description: lconf did not handle in-kernel recovery with LDAP properly
2498 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2500 Severity : enhancement
2502 Description: bind OST threads to NUMA nodes to improve performance
2503 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2504 node and do their allocations there to localize memory access
2506 Severity : enhancement
2508 Description: llmount can determine client NID directly from Myrinet (GM)
2509 Details : the client NID code from gmnalnid was moved directly into
2510 llmount, removing the need to use this or specifying the
2511 client NID explicitly when mounting GM clients with zeroconf
2514 Frequency : if client is started with down MDS
2516 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2517 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2518 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2519 ptlrpc_import_delayed_req() because we didn't check if the
2520 request was interrupted, and we also didn't break out of the
2521 event loop if there was no timeout
2526 Description: data loss during non-page-aligned writes to a single file from
2527 both multiple nodes and multiple threads on one node at same time
2528 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2529 inconsistency led to false short-reads, that were cached and later
2530 used by ->prepare_write() to fill in partially written page,
2531 leading to data loss.
2534 Frequency : always, if lconf --abort_recovery used
2536 Description: lconf --abort_recovery fails with 'Operation not supported'
2537 Details : lconf was attempting to abort recovery on the MDT device and not
2540 Severity : enhancement
2542 Description: remove cleanup logs
2543 Details : replace lconf-generated cleanup logs with lustre internal
2544 cleanup routines. Eliminates the need for client-cleanup and
2547 Severity : enhancement
2549 Description: add support for EAs (user and system) on lustre filesystems
2550 Details : it is now possible to store extended attributes in the Lustre
2551 client filesystem, and with the user_xattr mount option it
2552 is possible to allow users to store EAs on their files also
2554 Severity : enhancement
2556 Description: Add possibility (config option) to show minimal available OST free
2558 Details : When compiled with --enable-mindf configure option, statfs(2)
2559 (and so, df) will return least minimal free space available from
2560 all OSTs as amount of free space on FS, instead of summary of
2561 free spaces of all OSTs.
2563 Severity : enhancement
2565 Description: do not expand extent locks acquired on OST-side
2566 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2567 by server: they are not cached anyway.
2570 Frequency : when mmap is used/binaries executed from Lustre
2572 Description: Unmmap pages before throwing them away from read cache.
2573 Details : llap_shrink cache now attempts to unmap pages before discarding
2574 them (if unmapping failed - do not discard). SLES9 kernel has
2575 extra checks that trigger if this unmapping is not done first.
2580 Description: lconf didn't resolve symlinks before checking to see whether a
2581 given mountpoint was already in use
2584 Frequency : when migrating failover services
2585 Bugzilla : 6395, 9514
2586 Description: When migrating a subset of services from a node (e.g. failback
2587 from a failover service node) the remaining services would
2588 time out and evict clients.
2589 Details : lconf --force (implied by --failover) sets the global obd_timeout
2590 to 5 seconds in order to quickly disconnect, but this caused
2591 other RPCs to time out too quickly. Do not change the global
2592 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2594 Severity : enhancement
2595 Frequency : if MDS is started with down OST
2596 Bugzilla : 9439,5706
2597 Description: Allow startup/shutdown of an MDS without depending on the
2598 availability of the OSTs.
2599 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2600 Add appropriate locking and lov-osc refcounts for safe
2601 cleaning. Add osc abort_inflight calls in case the
2602 synchronize never started.
2605 Frequency : occasional (Cray XT3 only)
2607 Description: root not authorized to access files in CRAY_PORTALS environment
2608 Details : The client process capabilities were not honoured on the MDS in
2609 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2610 been accessed by an authorized user then root was able to access
2611 the file on the local client also. The root user capabilities
2612 are now allowed on the MDS, as this environment has secure UID.
2615 Frequency : occasional
2617 Description: ldiskfs "too long searching" message happens too often
2618 Details : A debugging message (otherwise harmless) prints too often on
2619 the OST console. This has been reduced to only happen when
2620 there are fragmentation problems on the filesystem.
2625 Description: Division by zero in statfs when all OSCs are inactive
2626 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2627 lov_statfs divides by value returned by lov_get_stripecnt().
2631 Bugzilla : 9489, 3273
2632 Description: First write from each client to each OST was only 4kB in size,
2633 to initialize client writeback cache, which caused sub-optimal
2634 RPCs and poor layout on disk for the first writen file.
2635 Details : Clients now request an initial cache grant at (re)connect time
2636 and so that they can start streaming writes to the cache right
2637 away and always do full-sized RPCs if there is enough data.
2638 If the OST is rebooted the client also re-establishes its grant
2639 so that client cached writes will be honoured under the grant.
2644 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2645 Details : Now I/O RPCs go to different portal number and (presumably) fast
2646 lock requests (and glimses) and other RPCs get their own service
2647 threads pool that should be able to service those RPCs
2650 Severity : enhancement
2652 Description: Ability to exchange lustre version between client and servers and
2653 issue warnings at client side if client is too old. Also for
2654 liblustre clients there is ability to refuse connection of too old
2656 Details : New 'version' field is added to connect data structure that is
2657 filled with version info. That info is later checked by server and
2661 Frequency : rare, liblustre only.
2662 Bugzilla : 9296, 9581
2663 Description: Two simultaneous writes from liblustre at offset within same page
2664 might proceed at the same time overwriting eachother with stale
2666 Details : I/O lock withing llu_file_prwv was released too early, before data
2667 actually was hitting the wire. Extended lock-holding time until
2668 server acknowledges receiving data.
2671 Frequency : extremely rare. Never observed in practice.
2673 Description: avoid generating lustre_handle cookie of 0.
2674 Details : class_handle_hash() generates handle cookies by incrementing
2675 global counter, and can hit 0 occasionaly (this is unlikely, but
2676 not impossible, because initial value of cookie counter is
2677 selected randonly). Value of 0 is used as a sentinel meaning
2678 "unassigned handle" --- avoid it. Also coalesce two critical
2679 sections in this function into one.
2681 Severity : enhancement
2683 Description: allow liblustre clients to delegate truncate locking to OST
2684 Details : To avoid overhead of locking, liblustre client instructs OST to
2685 take extent lock in ost_punch() on client's behalf. New connection
2686 flag is added to handle backward compatibility.
2688 Severity : enhancement
2689 Bugzilla : 4928, 7341, 9758
2690 Description: allow number of OST service threads to be specified
2691 Details : a module parameter allows the number of OST service threads
2692 to be specified via "options ost ost_num_threads={N}" in the
2693 OSS's /etc/modules.conf or /etc/modprobe.conf.
2697 Bugzilla : 6146, 9635, 9895
2698 Description: servers crash with bad pointer in target_handle_connect()
2699 Details : In rare cases when a client is reconnecting it was possible that
2700 the connection request was the last reference for that export.
2701 We would temporarily drop the export reference and get a new
2702 one, but this may have been the last reference and the export
2703 was just destroyed. Get new reference before dropping old one.
2705 Severity : enhancement
2706 Frequency : if client is started with failover MDS
2708 Description: Allow multiple MDS hostnames in the mount command
2709 Details : Try to read the configuration from all specified MDS
2710 hostnames during a client mount in case the "primary"
2713 Severity : enhancement
2715 Description: Stop sending data to evicted clients as soon as possible.
2716 Details : Check if the client we are about to send or are sending data to
2717 was evicted already. (Check is done every second of waiting,
2718 for which l_wait_event interface was extended to allow checking
2719 of exit condition at specified intervals).
2722 Frequency : rare, normally only when NFS exporting is done from client
2724 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2726 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2729 Frequency : always, if srand() is called before liblustre initialization
2731 Description: Liblustre uses system PRNG disturbing its usage by user application
2732 Details : Introduce internal to lustre fast and high-quality PRNG for
2733 lustre usage and make liblustre and some other places in generic
2734 lustre code to use it.
2736 Severity : enhancement
2737 Bugzilla : 9477, 9557, 9870
2738 Description: Verify that the MDS configuration logs are updated when xml is
2739 Details : Check if the .xml configuration logs are newer than the config
2740 logs stored on the MDS and report an error if this is the case.
2741 Request --write-conf, or allow starting with --old_conf.
2743 Severity : enhancement
2745 Description: Handle symlinks in the path when checking if Lustre is mounted.
2746 Details : Resolve intermediate symlinks when checking if a client has
2747 mounted a filesystem to avoid duplicate client mounts.
2752 Description: lconf can hit an error exception but still return success.
2753 Details : The lconf command catches the Command error exception at the top
2754 level script context and will exit with the associated exit
2755 status, but doesn't ensure that this exit status is non-zero.
2760 Description: failure of ptlrpc thread startup can cause oops
2761 Details : Starting a ptlrpc service thread can fail if there are a large
2762 number of threads or the server memory is very fragmented.
2763 Handle this without oopsing.
2766 Frequency : always, only if liblustre and non-default acceptor port was used
2768 Description: liblustre cannot connect to servers with non-default acceptor port
2769 Details : tcpnal_set_default_params() was not called and was therefore
2770 ignoring the environment varaible TCPNAL_PORT, as well as other
2771 TCPNAL_ environment variables
2776 Description: two objects could be created on the same OST for a single file
2777 Details : If an OST is down, in some cases it was possible to create two
2778 objects on a single OST for a single file. No problems other
2779 than potential performance impact and spurious error messages.
2783 Bugzilla : 5681, 9562
2784 Description: Client may oops in ll_unhash_aliases
2785 Details : Client dcache may become inconsistent in race condition.
2786 In some cases "getcwd" can fail if the current directory is
2792 Description: Inode refcounting problems in NFS export code
2793 Details : link_raw functions used to call d_instantiate without obtaining
2794 extra inode reference first.
2798 Bugzilla : 9942, 9903
2799 Description: Referencing freed requests leading to crash, memleaks with NFS.
2800 Details : We used to require that call to ll_revalidate_it was always
2801 followed by ll_lookup_it. Also with revalidate_special() it is
2802 possible to call ll_revalidate_it() twice for the same dentry
2803 even if first occurence returned success. This fix changes semantic
2804 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2805 reference on a request referred from the intent.
2806 ll_intent_release() then releases such a request.
2809 Frequency : rare, normally benchmark loads only
2811 Description: unlinked inodes were kept in memory on the client
2812 Details : If a client is repeatedly creating and unlinking files it
2813 can accumulate a lot of stale inodes in the inode slab cache.
2814 If there is no other client load running this can cause the
2815 client node to run out of memory. Instead flush old inodes
2816 from client cache that have the same inode number as a new inode.
2819 Frequency : SLES9 2.6.5 kernel and long filenames only
2820 Bugzilla : 9969, 10379
2821 Description: utime reports stale NFS file handle
2822 Details : SLES9 uses out-of-dentry names in some cases, which confused
2823 the lustre dentry revalidation. Change it to always use the
2827 Frequency : rare, unless heavy write-truncate concurrency is continuous
2828 Bugzilla : 4180, 6984, 7171, 9963, 9331
2829 Description: OST becomes very slow and/or deadlocked during object unlink
2830 Details : filter_destroy() was holding onto the parent directory lock
2831 while truncating+unlinking objects. For very large objects this
2832 may block other threads for a long time and slow overall OST
2833 responsiveness. It may also be possible to get a lock ordering
2834 deadlock in this case, or run out of journal credits because of
2835 the combined truncate+unlink. Solution is to do object truncate
2836 first in one transaction without parent lock, and then do the
2837 final unlink in a new transaction with the parent lock. This
2838 reduces the lock hold time dramatically.
2841 Frequency : rare, 2.4 kernels only
2843 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2844 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2845 before unmounting their filesystem in order to avoid deadlock.
2846 The kernel_locked() function in 2.4 kernels only checks whether
2847 the kernel lock is held, not whether it is this process that is
2848 holding it as 2.6 kernels do.
2853 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2854 Details : The client ptlrpc code may be trying to reconnect to a down
2855 server before a previous connection attempt has timed out.
2856 Increase the reconnect interval to be longer than the connection
2857 timeout interval to avoid sending duplicate connections to
2861 Frequency : echo_client brw_test command
2863 Description: fix echo_client to work with OST preallocated code
2864 Details : OST preallocation code (5137) didn't take echo_client IO path
2865 into account: echo_client calls filter methods outside of any
2866 OST thread and, hence, there is no per-thread preallocated
2867 pages and buffers to use. Solution: hijack pga pages for IO. As
2868 a byproduct, this avoids unnecessary data copying.
2872 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2873 Description: Client can oops in mdc_commit_close() after open replay
2874 Details : It was possible for the MDS to return an open request with no
2875 transaction number in mds_finish_transno() if the client was
2876 evicted, but without actually returning an error. Clients
2877 would later try to replay that open and may trip an assertion
2878 Simplify the client close codepath, and always return an error
2879 from the MDS in case the open is not successful.
2882 Frequency : rare, 2.6 OSTs only
2884 Description: OST may deadlock under high load on fragmented files
2885 Details : If there was a heavy load and highly-fragmented OST filesystems
2886 it was possible to have all the OST threads deadlock waiting on
2887 allocation of biovecs, because the biovecs were not released
2888 until the entire RPC IO was completed. Instead, release biovecs
2889 as soon as they are complete to ensure forward IO progress.
2891 Severity : enhancement
2893 Description: Support for specifying external journal device at mount
2894 Details : If an OST or MDS device is formatted with an external journal
2895 device, this device major/minor is stored in the ext3 superblock
2896 and may not be valid for failover. Allow detecting and
2897 specifying the external journal at mount time.
2902 Description: Mounting an MDS with pending unlinked files may cause oops
2903 Details : target_finish_recovery() calls mds_postrecov() which returned
2904 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2905 considers this an error and immediately begins cleaning up the
2906 lov, just after starting the mds_lov process
2908 Severity : enhancement
2910 Description: Implement 'lfs df' to report actual free space on per-OST basis
2911 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2912 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2913 usage of objects; '-h' to report in human readable format.
2915 ------------------------------------------------------------------------------
2917 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2924 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2925 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2926 at startup during orphan unlinking. Instead, skip buddy update
2927 before setup, as it will be regenerated after recovery is complete.
2932 Description: If an OST is inactive, its locks might reference stale inodes.
2933 Details : lov_change_cbdata() must iterate over all namespaces, even if
2934 they are inactive to clear inode references from the lock.
2936 Severity : enhancement
2937 Frequency : occasional, if non-standard max_dirty_mb used
2939 Description: Client will block write RPCs if not enough grant
2940 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2941 then the client will block writes while waiting for another RPC
2942 to complete instead of consuming its dirty limit. With change
2943 we get improved performance when max_dirty_mb is small.
2945 Severity : enhancement
2946 Bugzilla : 3389, 6253
2947 Description: Add support for supplementary groups on the MDS.
2948 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2949 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2950 lookups for user supplementary groups into a cache.
2954 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2955 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2956 for liblustre O_EXCL is enforced by MDS.
2960 Description: readdir never returns NULL in liblustre.
2961 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2962 dentry in struct dirent.
2966 Frequency : liblustre only, depends on application IO pattern
2967 Description: liblustre clients evicted if not contacting servers
2968 Details : Don't put liblustre clients into the ping_evictor list, so
2969 they will not be evicted by the pinger ever.
2971 Severity : enhancement
2973 Description: Add ability to evict clients by NID from MDS.
2974 Details : By echoing "nid:$NID" string into
2975 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2976 $NID would be instantly evicted from this MDS and from all active
2977 OSTs connected to it.
2981 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2982 Details : lookup_it_finish() used to query file size from OSTs that was not
2987 Description: service threads change working directory to that of init
2988 Details : Starting lustre service threads may pin the working directory
2989 of the parent thread, making that filesystem busy. Threads
2990 now change to the working directory of init to avoid this.
2994 Frequency : during shutdown only
2995 Description: shutdown with a failed MDS or OST can cause unmount to hang
2996 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3001 Frequency : 2.6 only
3002 Description: chmod/chown may include an extra supplementary group
3003 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3004 supplementary group and if none is specified this is used.
3007 Bugzilla : 5479 (6816)
3009 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3010 Details : If lookup is in progress on a file that is unlinked we might try
3011 to revalidate the inode and fail in revalidate after lookup is
3012 complete and ll_file_open() enqueues the open again but
3013 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3016 Frequency : always, if lconf --abort_recovery used
3018 Description: lconf --abort_recovery fails with 'Operation not supported'
3019 Details : lconf was attempting to abort recovery on the MDT device and not
3022 ------------------------------------------------------------------------------
3024 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3029 Frequency : rare (only unsupported configurations with a node running as an
3031 Bugzilla : 6514, 5137
3032 Description: Mounting a Lustre file system on a node running as an OST could
3034 Details : OSTs now preallocates memory needed to write out data at
3035 startup, instead of when needed, to avoid having to
3036 allocate memory in possibly low memory situations.
3037 Specifically, if the file system is mounted on on OST,
3038 memory pressure could force it to try to write out data,
3039 which it needed to allocate memory to do. Due to the low
3040 memory, it would be unable to do so and the node would
3041 become unresponsive.
3043 Severity : enhancement
3045 Description: Addition of lconf --service command line option
3046 Details : lconf now accepts a '--service <arg>' option, which is
3047 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3049 Severity : enhancement
3051 Description: Failover mode is now the default for OSTs.
3052 Details : By default, OSTs will now run in failover mode. To return to
3053 the old behaviour, add '--failout' to the lmc line for OSTs.
3055 Severity : enhancement
3057 Description: Health checks are now provided for MDS and OSTs
3058 Details : Additional detailed health check information on MSD and OSTs
3059 is now provided through the procfs health_check value.
3062 Frequency : occasional, depends on IO load
3064 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3065 after numerous create/delete cycles
3066 Details : The ext3 inode allocation policy would not allocate new inodes
3067 very well on the OSTs because there are no new directories
3068 being created. Instead we look for groups with free space if
3069 the parent directories are nearly full.
3074 Description: Network or server problems during mount may cause partially
3075 mounted clients instead of returning an error.
3076 Details : The config llog parsing code may overwrite the error return
3077 code during mount error handling, returning success instead
3083 Description: MDS can fail to allocate large reply buffers
3084 Details : After long uptimes the MDS can fail to allocate large reply
3085 buffers (e.g. zconf client mount config records) due to memory
3086 fragmentation or consumption by the buffer cache. Preallocate
3087 some large reply buffers so that these replies can be sent even
3088 under memory pressure.
3092 Frequency : rare (liblustre)
3093 Description: fsx running with liblustre complained that using truncate() to
3094 extend the file doesn't work. This patch corrects that issue.
3095 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3096 ATTR_SIZE and lsm use in llu_setattr_raw.
3100 Frequency : rare, only 2.6 kernels
3101 Description: Unusual file access patterns on the MDS may result in inode
3102 data being lost in very rare circumstances.
3103 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3104 code in the 2.6 kernel caused the inode and/or EA data not to
3105 be read from disk, causing single-file corruption.
3109 Frequency : rare, only 2.6 filesystems using extents
3110 Description: Heavy concurrent write and delete load may cause data corruption.
3111 Details : It was possible under high-load situations to have an extent
3112 metadata block in the block device cache from a just-unlinked
3113 file overwrite a newly-allocated data block. We now unmap any
3114 metadata buffers that alias just-allocated data blocks.
3118 Frequency : filesystems with default stripe_count larger than 77
3119 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3120 Details : lconf specifies an inode size of 4096 bytes when the default
3121 stripe_count is larger than 77. This conflicts with the default
3122 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3123 this case to avoid pinning too much memory for large EAs.
3125 ------------------------------------------------------------------------------
3127 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3132 Frequency : rare (extremely heavy IO load with hundreds of clients)
3134 Description: Client is evicted, gets IO error writing to file
3135 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3136 caused clients to be evicted for AST timeouts. The fixes in
3137 bug 5192 mean we no longer need to have such short AST timeouts
3138 so ldlm_timeout has been increased.
3141 Frequency : occasional during --force or --failover shutdown under load
3142 Bugzilla : 5949, 4834
3143 Description: Server oops/LBUG if stopped with --force or --failover under load
3144 Details : a collection of import/export refcount and cleanup ordering
3145 issues fixed for safer force cleanup
3148 Frequency : only filesystems larger than 120 OSTs
3149 Bugzilla : 5990, 6223
3150 Description: lfs getstripe would oops on a very large filesystem
3151 Details : lov_getconfig used kfree on vmalloc'd memory
3154 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3155 Bugzilla : 6242, 6243
3156 Description: reading from files that had been truncated to a non-zero size
3157 but never opened returned no data
3158 Details : ll_file_read() reads zeros from no-object files to EOF
3163 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3164 Details : There's little harm in aborting MDS/OSS recovery and letting it
3165 try again, so I removed the LASSERT and return an error instead.
3167 Severity : enhancement
3169 Description: New debugging infrastructure for tracking down data corruption
3170 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3171 (b) cover more of the client-side code path, and (c) try to narrow
3172 down where problems occurred
3176 Bugzilla : 3819, 4364, 4397, 6313
3177 Description: Racing close and eviction MDS could cause assertion in mds_close
3178 Details : It was possible to get multiple mfd references during close and
3179 client eviction, leading to one thread referencing a freed mfd.
3181 Severity: : enhancement
3182 Bugzilla : 3262, 6359
3183 Description: Attempts to reconnect to servers are now more aggressive.
3184 Details : This builds on the enhanced upcall-less recovery that was added
3185 in 1.4.2. When trying to reconnect to servers, clients will
3186 now try each server in the failover group every 10 seconds. By
3187 default, clients would previously try one server every 25 seconds.
3192 Description: After recovery, certain operations trigger a failed
3193 assertion on a client.
3194 Details : Failing over an mds, using lconf -d --failover, while a
3195 client was doing a readdir() call would cause the client to
3196 LBUG after recovery completed and the readdir() was resent.
3198 Severity : enhancement
3200 Description: Default groups are now added by lconf
3201 Details : You can now run lconf --group <servicename> without having to
3202 manually add groups with lmc.
3205 Frequency : occasional
3207 Description: Nodes with an elan id of 0 trigger a failed assertion
3210 Frequency : always when accessing e.g. tty/console device nodes
3212 Description: tty and some other devices nodes cannot be used on lustre
3213 Details : file's private_data field is used by device data and lustre
3214 values in there got lost. New field was added to struct file to
3215 store fs-specific private data.
3218 Frequency : when exporting Lustre via NFS
3220 Description: NFSD failed occasionally when looking up a path component
3221 Details : NFSD is looking up ".." which was broken in ext3 directories
3222 that had grown large enough to become hashed.
3225 Frequency : Clusters with multiple interfaces not on the same subnet
3227 Description: Nodes will repeatedly try to reconnect to an interface which it
3228 cannot reach and report an error to the log.
3229 Details : Extra peer list entries will be created by lconf with some peers
3230 unreachable. lconf now validates the peer before adding it.
3233 Frequency : Only if a default stripe is set on the filesystem root.
3235 Description: Setting a default stripe on the filesystem root prevented the
3236 filesystem from being remounted.
3237 Details : The client was sending extra request flags in the root getattr
3238 request and did not allocate a reply buffer for the dir EA.
3241 Frequency : occasional, higher if lots of files are accessed by one client
3242 Bugzilla : 6159, 6097
3243 Description: Client trips assertion regarding lsm mismatch/magic
3244 Details : While revalidating inodes the VFS looks up inodes with ifind()
3245 and in rare cases can find an inode that is being freed.
3246 The ll_test_inode() code will free the lsm during ifind()
3247 when it finds an existing inode and then the VFS later attaches
3248 this free lsm to a new inode.
3252 Bugzilla : 6422, 7030
3253 Description: MDS deadlock between mkdir and client eviction
3254 Details : Creating a new file via mkdir or mknod (starting a transaction
3255 and getting the ns lock) can deadlock with client eviction
3256 (gets ns lock and trying to finish a synchronous transaction).
3259 Frequency : occasional
3260 Description: While starting a server, the fsfilt_ext3 module could not be
3262 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3263 kernels. Previously, lconf would still use the ext3 name
3264 when trying to load modules. Now, it will correctly use
3265 ext3 on 2.4 and ldiskfs on 2.6.
3267 Severity : enhancement
3268 Description: The default stripe count has been changed to 1
3269 Details : The interpretation of the default stripe count (0, to lfs
3270 or lmc) has been changed to mean striping across a single
3271 OST, rather than all available. For general usage we have
3272 found a stripe count of 1 or 2 works best.
3274 Severity : enhancement
3275 Description: Add support for compiling against Cray portals.
3276 Details : Conditional compiling for some areas that are different
3280 Frequency : occasional
3281 Bugzilla : 6409, 6834
3282 Description: Creating files with an explicit stripe count may lead to
3283 a failed assertion on the MDS
3284 Details : If some OSTs are full or unavailable, creating files may
3285 trigger a failed assertion on the MDS. Now, Lustre will
3286 try to use other servers or return an error to the
3290 Frequency : occasional
3292 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3293 caused lock timeout during readahead (since 1.4.2).
3294 Details : Processes doing readahead might match a lock that hasn't been
3295 granted yet if there are overlapping and conflicting lock
3296 requests. The readahead process waits on ungranted lock
3297 (original lock is CBPENDING), while OST waits for that process
3298 to cancel CBPENDING read lock and eventually evicts client.
3300 Severity : enhancement
3302 Description: Initial enabling of flock support for clients
3303 Details : Implements fcntl advisory locking and file status functions.
3304 This feature is provided as an optional mount flag (default
3305 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3306 locking are implemented yet, and those that are are not guaranteed
3307 to be completely correct in production environments.
3308 mount -t lustre -o [flock|noflock] ...
3311 Frequency : occasional
3313 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3314 assertion in the ext3 JBD (journaling) layer.
3315 Details : The b_committed_data struct is protected by the big kernel lock
3316 in 2.4 kernels, serializing journal_commit_transaction() and
3317 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3318 access to this struct. In 2.6 kernels there is finer grained
3319 locking to improve SMP performance of the JBD layer.
3323 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3328 Description: merge and cleanup kernel patches.
3331 -----------------------------------------------------------------------------
3333 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3335 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3336 versions of Lustre. Please update all servers and clients to
3337 version 1.4.2 or later at the same time. You must also run
3338 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3339 to update the configuration logs.
3341 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3342 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3343 - swab llog records of type '0' so we get proper header size/idx (5861)
3344 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3345 - fix rename of one directory over another leaking an inode (5953)
3346 - avoid SetPageDirty on 2.6 (5981)
3347 - don't re-add just-being-destroyed locks to the waiting list (5653)
3348 - when creating new directories, inherit the parent's custom
3349 striping settings if present parent (3048)
3350 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3351 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3352 - don't hold client locks on temporary worklist from l_lru (5666)
3353 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3354 - automatically evict dead clients (5921)
3355 - Update file size properly in create+truncate+fstat case (6196)
3356 - Do not unhash mountpoint dentries, do not allow removal of
3358 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3359 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3360 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3361 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3362 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3363 - Correctly handle reads of files with no objects (6243)
3364 - lctl recover will also mark a device active if deactivate used (5933)
3366 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3367 - allow --write-conf on an MDS with different nettype than client (5619)
3368 - don't write config llogs to MDS for mounts not from that MDS (5617)
3369 - lconf should create multiple TCP connections from a client (5201)
3370 - init scripts are now turned off by default; run chkconfig --on
3371 lustre and chkconfig --on lustrefs to use them
3372 - upcalls are no longer needed for clients to recover to failover
3374 - add --abort-recovery option to lconf to abort recovery on device
3376 - add support for an arbitrary number of OSTs (3026)
3377 - Quota support protocol changes.
3378 - forward compatibility changes to wire structs (6007)
3379 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3380 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3381 - improved "lustre status" script
3382 - initialize blocksize for non-regular files (6062)
3383 - added --disable-server and --disable-client configure options (5782)
3384 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3385 - Vanilla 2.4.29 support
3386 - increase maximum number of obd devices to 520 (6242)
3387 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3388 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3389 SLES 9 2.6.5 kernels
3390 - limit stripes per file to 160 (the maximum EA size) (6093)
3392 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3395 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3396 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3397 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3398 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3399 - small ext3 extents cleanups and fixes (5733)
3400 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3401 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3402 - avoid needless client->OST connect, fix handle mismatch (5317)
3403 - fix DLM error path that led to out-of-sync client, long delays (5779)
3404 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3405 - fix several locking issues related to i_size (5492,5624,5654,5672)
3406 - don't move pending lock onto export if it is already evicted (5683)
3407 - fix kernel oops when creating .foo in unlinked directory (5548)
3408 - fix deadlock in obdfilter statistics vs. object create (5811)
3409 - use time_{before,after} to avoid timer jiffies wrap (5882)
3410 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3411 - Do not leak request if server was not able to process it (5154)
3412 - If mds_open unable to find parent dir, make that negative lookup(5154)
3413 - don't create new directories with extent-mapping (5909, 5936)
3415 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3416 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3417 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3418 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3419 - support for RHEL4 kernel on i686 (5773)
3420 - provide error messages when incompatible logs are encountered (5898)
3422 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3423 * version 1.4.0.10 (1.4.1 release candidate 1)
3425 - don't keep a lock reference when lock is not granted (4238)
3426 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3427 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3428 - drop import inflight refcount on signal_completed_replay error (5255)
3429 - unlock page after async write error during send (3677)
3430 - handle missing objects in filter_preprw_read properly (5265)
3431 - no transno return for symlink open, don't save no-trasno open (3440)
3432 - don't try to complete elan receive that already failed (4012)
3433 - free RPC server reply state on error (5406)
3434 - clean up thread from ptlrpc_start_thread() on error (5160)
3435 - readahead could read extra page into cache that wasn't ejected (5388)
3436 - prevent races in class_attach/setup/cleanup/detach (5260)
3437 - don't dereference de->d_inode after l_dput of de (5458)
3438 - use "int" for stripe value returned from lock_to_stripe (5544)
3439 - mballoc allocation and error-checking fixes in 2.6 (5504)
3440 - block device patches to fix I/O request sizes in 2.6 (5482)
3441 - look up hostnames for IB nals (5602)
3442 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3443 - don't start multiple acceptors for the same port (5277)
3444 - fix incorrect LASSERT in mds_getattr_name (5635)
3445 - export a proc file for general "ping" checking (5628)
3446 - fix "lfs check" to not block when the MDS is down (5628)
3448 - service request history (4965)
3449 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3450 - create an "evict_client" /proc entry on OSTs, like the MDS has
3451 - fix mount usage message, return errors per mount(8) (5168)
3452 - change grep [] to grep "[]" in tests so they work in more UMLs
3453 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3454 - remove ancient LOV_MAGIC_V0 EA support (5047)
3455 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3456 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3457 - allow readahead allocations to fail when low on memory (5383)
3458 - mmap locking landed again, after considerable improvement (2828)
3459 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3461 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3464 - send OST transaction number in read/write reply to free req (4966)
3465 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3466 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3467 - the watchdog thread now runs as interruptible (5246)
3468 - flock/lockf fixes (but it's still disabled, pending 5135)
3469 - don't use EXT3 constants in llite code (5094)
3470 - memory shortage at startup could cause assertion (5176)
3472 - reorganization of lov code
3473 - single portals codebase
3475 - add extents/mballoc support (5025)
3476 - direct I/O reads in the obdfilter (4048)
3477 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3479 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3482 - send OST transaction number in read/write reply to free req (4966)
3483 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3484 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3485 - the watchdog thread now runs as interruptible (5246)
3486 - handle missing objects in filter_preprw_read properly (5265)
3487 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3488 - drop import inflight refcount on signal_completed_replay error (5255)
3489 - unlock page after async write error during send (3677)
3490 - return original error code on reconstructed replies (3761)
3491 - no transno return for symlink open, don't save no-trasno open (3440)
3493 - add pid to ldlm debugging output (4922)
3494 - bump the watchdog timeouts -- we can't handle 30sec yet
3495 - extra debugging for orphan dentry/inode bug (5259)
3497 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3500 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3501 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3502 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3503 - fix NULL dereference in /proc/sys/portals/routes (4827)
3504 - allow failed mdc_close() operations to be interrupted (4561)
3505 - stop precreate on OST before MDS would time out on it (4778)
3506 - don't send partial-page writes before EOF from client (4410)
3507 - discard client grant for sub-page writes on large-page clients (4520)
3508 - don't free dentries not owned by NFS code, check generation (4806)
3509 - fix lsm leak if mds_create_objects() fails (4801)
3510 - limit debug_daemon file size, always print CERROR messages (4789)
3511 - use transno after validating reply (3892)
3512 - process timed out requests if import state changes (3754)
3513 - update mtime on OST during writes, return in glimpse (4829)
3514 - add mkfsoptions to LDAP (4679)
3515 - use ->max_readahead method instead of zapping global ra (5039)
3516 - don't interrupt __l_wait_event() during strace
3518 - add software watchdogs to catch hung threads quickly (4941)
3519 - make lustrefs init script start after nfs is mounted
3520 - fix CWARN/ERROR duplication (4930)
3521 - return async write errors to application if possible (2248)
3522 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3523 - print NAL number in %x format (4645)
3524 - update barely-supported suse-2.4.21-171 series (4842)
3525 - support for sles 9 %post scripts
3526 - support for building 2.6 kernel-source packages
3527 - support for sles km_* packages
3529 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3532 - ignore -ENOENT errors in osc_destroy (3639)
3533 - notify osc create thread that OSC is being cleaned up (4600)
3534 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3535 - reconstruct ost_handle() like mds_handle() (4657)
3536 - create a new thread to do import eviction to avoid deadlock (3969)
3537 - let lconf resolve symlinked-to devices (4629)
3538 - don't unlink "objects" from directory with default EA (4554)
3539 - hold socknal file ref over connect in case target is down (4394)
3540 - allow more than 32000 subdirectories in a single directory (3244)
3541 - fix blocks count for O_DIRECT writes (3751)
3542 - OST returns ENOSPC from object create when no space left (4539)
3543 - don't send truncate RPC if file size isn't changing (4410)
3544 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3545 - bind to privileged port in socknal and tcpnal (3689)
3547 - rate limit CERROR/CWARN console message to avoid overload (4519)
3548 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3549 - basic mmap support (3918)
3550 - kernel patch series update from b1_4 (4711)
3552 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3555 - avoid crash during MDS cleanup with OST shut down (2775)
3556 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3557 - don't use bad inodes on the MDS (3744)
3558 - dynamic object preallocation to improve recovery speed (4236)
3559 - don't hold spinlock over lock dumping or change debug flags (4401)
3560 - don't zero obd_dev when it is force cleaned (3651)
3561 - print grants to console if they go negative (4431)
3562 - "lctl deactivate" will stop automatic recovery attempts (3406)
3563 - look for existing locks in ldlm_handle_enqueue() (3764)
3564 - don't resolve lock handle twice in recovery avoiding race (4401)
3565 - revalidate should check working dir is a directory (4134)
3567 - don't always mark "slow" obdfilter messages as errors (4418)
3569 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3572 - don't close LustreDB during write_conf until it is done (3860)
3573 - fix typo in lconf for_each_profile (3821)
3574 - allow dumping logs from multiple threads at one time (3820)
3575 - don't allow multiple threads in OSC recovery (3812)
3576 - fix debug_size parameters (3864)
3577 - fix mds_postrecov to initialize import for llog ctxt (3121)
3578 - replace config semaphore with spinlock (3306)
3579 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3580 - don't allow enqueue to complete on a destroyed export (3822)
3581 - down write_lock before checking llog header bitmap (3825)
3582 - recover from lock replay timeout (3764)
3583 - up llog sem before sending rpc (3652)
3584 - reduce ns lock hold times when setting kms (3267)
3585 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3586 - fix NULL deref and obd_dev leak on setup error (3312)
3587 - replace some LBUG about llog ops with error handling (3841)
3588 - don't match INVALID dentries from d_lookup and spin (3784)
3589 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3590 - fix invalid assertion in ptlrpc_set_wait (3880)
3592 - add libwrap support for the TCP acceptor (3996)
3593 - add /proc/sys/portals/routes for non-root route listing (3994)
3594 - allow setting MDS UUID in .xml (2580)
3595 - print the stack of a process that LBUGs (4228)
3597 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3600 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3601 - make sure to unset replay flag from failed open requests (3440)
3602 - if default stripe count is 0, use OST count for inode size (3636)
3603 - update parent mtime/ctime on client for create/unlink (2611)
3604 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3605 - free recovery state on server during a forced cleanup (3571)
3606 - unregister_reply for resent reqs (3063)
3607 - loop back devices mounting and status check on 2.6 (3563)
3608 - fix resource-creation race that can provoke i_size == 0 (3513)
3609 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3610 - more debugging for page-accounting assertion (3746)
3611 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3612 - avoid deadlock after precreation failure (3758)
3613 - fix race and lock order deadlock in orphan handling (3450, 3750)
3614 - add validity checks when grabbing inodes from l_ast_data (3599)
3616 - add /proc/.../recovery_status to obdfilter (3428)
3617 - lightweight CDEBUG infrastructure, debug daemon (3668)
3618 - change default OSC RPC parameters to be better on small clusters
3619 - turn off OST read cache for files smaller than 32MB
3620 - install man pages and include them in rpms (3100)
3621 - add new init script for (un)mounting lustre filesystems (2593)
3622 - run chkconfig in %post for init scripts (3701)
3623 - drop scimac NAL (unmaintained)
3625 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3628 - clean kiobufs before and after use (3485)
3629 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3630 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3631 - change init script to fail more clearly if not run as root (1528)
3632 - allow clients to reconnect during replay (1742)
3633 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3634 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3635 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3636 - fix cond_resched (3554)
3637 - only evict unfinished clients after recovery (3515)
3638 - allow bulk resend, prevent data loss (3570)
3639 - dynamic ptlrpc request buffer allocation (2102)
3640 - don't allow unlinking open directory if it isn't empty (2904)
3641 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3642 - remove extraneous obd dereference causing LASSERT failure (3334)
3643 - don't use get_cycles() when creating temp. files on the mds (3156)
3644 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3645 - handle EEXIST for set-stripe, set proper directory name (3336)
3647 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3648 - fix ksocknal_fmb_callback() error messages (2918)
3650 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3653 - don't copy lvb into (possibly NULL) reply on error (2983)
3654 - don't deref dentry after dput, don't free lvb on error (2922)
3655 - use the kms to determine writeback rpc length (2947)
3656 - increment oti_logcookies when osc is inactive (2948)
3657 - update client's i_blocks count via lvb messages (2543)
3658 - handle intent open/close of special files properly (1557)
3659 - mount MDS with errors=remount-ro, like obdfilter (2009)
3660 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3661 - don't use cancelling-locks' kms values (2947)
3662 - use highest lock extent for kms, not last one (2925)
3663 - don't dereference ERR_PTR() dentry in error handling path (3107)
3664 - fix thread race in portals_debug_dumplog() (3122)
3665 - create lprocfs device entries at setup instead of at attach (1519)
3666 - common AST error handler, don't evict client on completion race (3145)
3667 - zero nameidata in detach_mnt in 2.6 (3118)
3668 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3669 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3670 - initialize RPC timeout timer earlier for 2.6 (3219)
3671 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3672 - print nal/nid for unknown nid (3258)
3673 - additional checks for oscc recovery before doing precreate (3284)
3674 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3675 - don't crash in mdc_close for bad permissions on open (3285)
3676 - zero i_rdev for non-device files (3147)
3677 - clear page->private before handing to FS, better assertion (3119)
3678 - tune the read pipeline (3236)
3679 - fix incorrect decref of invalidated dentry (2350)
3680 - provide read-ahead stats and refine rpc in flight stats (3328)
3681 - don't hold journal transaction open across create RPC (3313)
3682 - update atime on MDS at close time (3265)
3683 - close LDAP connection when recovering to avoid server load (3315)
3684 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3685 - don't leak open file on MDS after open resend (3325)
3686 - serialize filter_precreate and filter_destroy_precreated (3329)
3687 - loop device shouldn't call sync_dev() for nul device (3092)
3688 - clear page cache after eviction (2766)
3689 - resynchronize MDS->OST in background (2824)
3690 - refuse to mount the same filesystem twice on same mountpoint (3394)
3691 - allow llmount to create routes for mounting behind routers (3320)
3692 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3693 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3694 - fix rare problem with rename on htree directories (3417)
3696 - allow default OST striping configuration per directory (1414)
3697 - fix compilation for qswnal for 2.6 kernels (3125)
3698 - increase maximum number of MDS request buffers for large systems
3699 - change liblustreapi to be useful for external progs like lfsck (3098)
3700 - increase local configuration timeout for slow disks (3353)
3701 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3703 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3706 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3707 - don't overwrite extent policy data in reply if lock was blocked (2901)
3708 - drop filter export grants atomically with removal from device (2663)
3709 - del obd_self_export from work_list in class_disconnect_exports (2908)
3710 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3711 - swab reply message in mdc_close, other PPC fixes (2464)
3712 - fix destroying of named logs (2325)
3713 - overwrite old logs when running lconf --write_conf (2264)
3714 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3715 - fix race in target_handle_connect (2898)
3716 - mds_reint_create() should take same inode create lock (2926)
3717 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3718 - don't close files for self_export to avoid uninitialized obd (2936)
3719 - allow MDS with the same name as client node (2939)
3720 - hold dentry reference for closed log files for unlink (2325)
3721 - reserve space for all logs during transactions (2059)
3722 - don't evict page beyond end of stripe extent (2925)
3723 - don't oops on a deleted current working directory (2399)
3724 - handle hard links to targets without a parent properly (2517)
3725 - don't dereference NULL lock when racing during eviction (2867)
3726 - don't grow lock extents when lots of conflicting locks (2919)
3728 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3731 - account for cache space usage on clients to avoid data loss (974)
3732 - lfsck support in lustre kernel code (2349)
3733 - reduce journal credits needed for BRW writes (2370)
3734 - orphan handling to avoid losing space on client/server crashes
3735 - ptlrpcd can be blocked, stopping ALL progress (2477)
3736 - use lock value blocks to assist in proper KMS, faster stat (1021)
3737 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3738 - recovery for initial connections (2355)
3739 - fixes for mds_cleanup_orphans (1934)
3740 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3741 - block all file creations until orphan recovery completes (1901)
3742 - client remove rq_connection from request struct (2423)
3743 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3744 - recovery timer race (2670)
3745 - mdc_close recovey bug (2532)
3746 - ptlrpc cleanup bug (2710)
3747 - mds timeout on local locks (2588)
3748 - namespace lock held during RPCs (2431)
3749 - handle interrupted sync write properly (2503)
3750 - don't try to handle a message that hasn't been replied to (2699)
3751 - client assert failure during cleanup after abort recovery (2701)
3752 - leak mdc device after failed mount (2712)
3753 - ptlrpc_check_set allows timedout requests to complete (2714)
3754 - wait for inflight reqs when ptlrpcd finishes (2710)
3755 - make sure unregistered services are removed from the srv_list
3756 - reset bulk XID's when resending them (caught by 1138 test)
3757 - unregister_bulk after timeout
3758 - fix lconf error (2694)
3759 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3760 - readahead locks pages, leaves pending causing memory pressure (2673)
3761 - increase OST request buffers to 4096 on large machines (2729)
3762 - fix up permission of existing directories in simple_mkdir (2661)
3763 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3764 - don't assign transno to errored transactions (2742)
3765 - don't delete objects on OST if given a bogus objid from MDS (2751)
3766 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3767 - if rq_no_resend, then timeout request after recovery (2432)
3768 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3769 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3770 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3771 - MDS recovery shouldn't delete objects that it has given out (2730)
3772 - if enqueue arrives after completion, don't clobber LVB (2819)
3773 - don't unlock pages twice when trigger_group_io returns error (2814)
3774 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3775 - don't write pages to disk if there was an error (1450)
3776 - don't ping imports that have recovery disabled (2676)
3777 - take buffered bytes into account when balancing socknal conn (2817)
3778 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3779 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3780 - remove little-endian swabbing of llog records (1987)
3781 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3782 - retry reposting request buffers if they fail (1191)
3783 - grow extent at grant time to avoid granting a revoked lock (2809)
3784 - lock revoke doesn't evict page if covered by a second lock (2765)
3785 - disable VM readahead to avoid reading outside lock extents (2805)
3787 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3788 - updated kernel patches for hp-2.4.20 kernel (2681)
3790 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3793 - fix truncated write corruption (2366)
3794 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3796 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3797 - ensure len(uuid) < 37 in lmc (1171)
3798 - fix ia64 OOPS in llog_test (2255)
3799 - zero end of page at obdfilter for partial page writes (2648)
3800 - don't leave stale dentries around after renames (bug 2428)
3801 - fix timeouts when evicting a client with a single lock held (2642)
3802 - set deadline for the initial HELLO message to drain (2634)
3803 - print out dotted-quad IP addresses in the socknal (2302)
3805 - additional debugging for MDS client eviction problem (2443)
3806 - fix mkfsoptions support for osts (2603, 2604)
3808 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3811 - add series for the vanilla 2.6.0 kernel
3812 - add series for the vanilla 2.4.24 kernel
3813 - add series for a cray x86/64 UL kernel drop
3814 - fix xattr patches for the vanilla 2.4.19 series
3816 - generate true UUIDs in lmc (1171)
3817 - have portals stack dumping break in UML (2466)
3818 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3819 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3820 - wake callback waiting threads on client eviction (2460)
3821 - Add --ptldebug and --subsystem to lmc (1719)
3822 - update assertion to allow safe interrupt allocation
3823 - set rq_no_resend for cancel requests (2432)
3824 - recalculate ptlrpcd timeout after resend (2494)
3825 - call vfs_rmdir when removing pending directories (2368)
3826 - fix renaming a file to itself (2429)
3827 - lmc creates a default one-stripe lov (2454)
3828 - expand procfs space to handle large clusters (2326)
3829 - increase UML stack to avoid overflow
3830 - update lconf's list of debug and subsystem masks
3831 - fix lfs find --obd (2510)
3832 - /proc tunable for disabling filter read caching (2591)
3833 - stop rpm packages from altering slapd.conf (2301)
3834 - disable nagle in the socknal under 0conf (2578)
3835 - choose mds inode size based on stripe count (2572)
3836 - fix kernel-source rpm problems (2516)
3838 - add --disable-doc to avoid pdf generation (2421)
3839 - update documentation, tests, type-os, comments
3840 - avoid format warnings on ia64
3841 - remove the TOE NAL
3842 - tiny code cleanups by removing unused fields
3844 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3847 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3848 - remove the most blatant lies from BUILDING file (2371)
3849 - change default debug level to reasonable production setting
3850 - reduce client side cache size to reduce cache flush time
3851 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3852 - make TCP zerocopy and pinger support enabled by default (2476)
3853 - sync writes completed after process exits caused crashes (2319)
3854 - maintain correct mount count on the MDS (2356)
3855 - backout 1557, because 2316 wasn't really fixed
3856 - better file I/O statistics gathering in /proc
3857 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3858 - another kernel patch to fix zero-copy TCP function export
3859 - don't take duplicate lock when processing re-sent getattr (2420)
3860 - lctl uses obd_self_export instead of creating new conn (2353)
3861 - MDS/OST recovery case which requires object creation asserted (2425)
3862 - move lfs from /usr/sbin to /usr/bin in packages
3863 - fix race between mds_client_add and mds_client_free (2417)
3864 - use kmalloc instead of slabs in portals (2430)
3865 - don't create duplicate records when a failover MDS is present (2442)
3866 - remove unnecessary mount age check (2332)
3867 - don't remove directory inodes from locks prematurely (2451)
3868 - don't break if MDS service name is the same as hostname (2103)
3869 - fix races in client write RPC generation when cache full (2482)
3871 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3874 - remove now-unused request->rq_obd (278)
3875 - if an allocation fails, print out how much memory we've used (1933)
3876 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3877 - add the "configurable stack size" patch to most series files (1256)
3878 - ability to write large log records, for 100+ OST configs (2306)
3879 - fix NULL deref when filter_prep fails (2314)
3880 - fix operator precedence error in filter_sync
3881 - dynamic allocation of socknal TX descriptors (2315)
3882 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3883 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3884 - ensure that CWARN messages in Portals always get to the syslog
3885 - __init/__exit are not for prototype decls (ldlm_init/exit)
3886 - x86-64 compile warning fixes
3887 - fix gateway LMC keyword conflict (2318)
3888 - fix MDS lock inversions in getattr/reint paths (1844)
3889 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3890 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3891 - DLM race condition prevented some lock evictions (2328)
3892 - ENOMEM detection and retry on socknal sends (2230)
3893 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3894 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3895 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3896 - better error messages when a client is rejected during recovery (1505)
3897 - avoid cancelling locks which were never granted, after failure (2330)
3898 - fix i_sem/journal inversion in mds_client_add (2333)
3899 - fix truncate/getattr lock cycle deadlock (2334)
3900 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3901 - fix two rare exit paths which could leak an l_lock() ref (2321)
3902 - fencepost error in MDS/OST orphan recovery (2226)
3903 - make log record alignment 8 bytes (1988)
3904 - lstripe now fails when requested offset > ost_count (2237)
3905 - ensure that all kernel series have a complete list.h (1607)
3906 - fix crashes in special-file operations (2316)
3907 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3909 - allow configurable automake binary, for testing new versions
3910 - small update to the lfs documentation
3912 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3914 * fix negative export reference count in fsfilt_sync (2312)
3916 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3917 * release candidate 0.9.1
3919 - orphans are moved into the PENDING directory for possible recovery
3920 - replayed opens now open by fid for orphan/rename safety (1042)
3921 - last close of an orphan inode generates a transno (683)
3922 - chdir() and mount() now pin the directory entry (1020)
3923 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3924 - discard very old requests without processing them (1502)
3925 - remove some common, well-understood CERRORs (1505)
3926 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3927 - clear "grant" flags in OST replies until OST grant code lands (1644)
3928 - fix read performance by not clobbering i_blksize on client (1598)
3929 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3930 - make LRU size a /proc tunable, clears locks when reduced (707)
3931 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3932 - server locks take references on exports now (1558)
3933 - build fixes for 2.4.20-rh trees (1663)
3934 - return an error from lov_create if all OSCs are inactive (1751)
3935 - fix import levels when a reconnect happens without a timeout (1597)
3936 - exit early from mds_open if we get a lookup error (1749)
3937 - partial page read at EOF wouldn't wait for disk before sending (1642)
3938 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3939 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3940 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3941 - write/truncate lock inversion (1639)
3942 - Don't auto-load obdclass, portals modules during cleanup (1495)
3943 - fix timestamps from jumping to "now" (1763)
3944 - extra journal assertions (1648)
3945 - add an extra multiunlink test (1771)
3946 - fix read_record/write_record API (1776)
3947 - fix leak of offset_extent, possible incorrect i_size later (1772)
3948 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3949 - Debugging for the kqswnal_get_idle_tx problems (1820)
3950 - Allow recovery to be attempted multiple times (1536)
3951 - Write out MDS last_rcvd file after it is first created (1600)
3952 - Fix tx_descriptor leak in failed transmit situations (1827)
3953 - ext3 journaling fixes for assertion failure after IO error (1871)
3954 - class_export_put() on freed export after completion AST error (1896)
3955 - Fix revalidate looping in VFS (1322)
3956 - Don't access a freed export during MDS_REINT timeout (1521)
3957 - Add open-unlink recovery support on the MDS (1673,1764)
3958 - Return an error if no MDS data was read from last_rcvd (1946)
3959 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3960 - Don't setup a disk device that doesn't match exported UUID (317)
3961 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3962 - avoid committing NULL handle in force close
3963 - local.sh is now a one-stripe LOV configuration
3964 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3965 - fix ext3 htree duplicate directory entry corruption (1516)
3966 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3967 - update to new LOV EA format (2097)
3968 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3970 2003-06-15 Phil Schwan <phil@clusterfs.com>
3973 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3974 - per-import/export recovery handling (958, 931, 959)
3975 - multiple last-rcvd slots, for serving multiple FSes (949)
3976 - connections are again shared between multiple imp/exports (963, 964)
3977 - "umount -f" would hang if any requests needed to be sent (393, 978)
3978 - avoid pinning large req buffer by copying for queued messages (989)
3979 - add "uuid" to "lctl device" command to help upcalls (991)
3980 - "open" RPCs with transnos would confuse recovery counters (1037)
3981 - do proper endian conversion of all wire messages (288, 340, 891)
3982 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3983 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3984 - fix ptlrpc_connection leak in target_handle_connect (1174)
3985 - fix import refcounting bug in OST and MDS cleanup (1134)
3986 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3987 - fix very unlikely obd_types race condition (501)
3988 - remove osc_open hack for echo_client (1187)
3989 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3990 - a failure in read_inode2 leads to deadlock (1139)
3991 - cancel ack-locks as soon as transaction is committed (1072)
3992 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3993 - make sure to commitrw after any preprw to avoid deadlock (1162)
3994 - failing to execute a file in a lustre FS would lock inode (1203)
3995 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3996 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3997 - obd_uuid2tgt would walk off the end of the list (1255)
3998 - on IA64 the osc would give portals incorrect bulk size (1258)
3999 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4000 - fix lock inversion caused by new llite matching code (1282)
4001 - limit the number of dirty pages on a client to 10MB (1286)
4002 - timed out locks were not being corrected cancelled (1289)
4003 - fix O_DIRECT above 4GB on IA-32 (1292)
4004 * major user-visible changes
4005 - fail out/fail over policy now controlled by the upcall (993)
4007 - add OBD_PING to check server availability and failure (954)
4008 - lustre messages are now sent in sending host order (288, 340, 891)
4009 - add eadatalen to MDS getattr reply (340)
4010 - OST read replies may contain second buffer, with per-page status (593)
4012 2003-03-11 Phil Schwan <phil@clusterfs.com>
4015 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4016 - failed open()s could cause deadlock; fixed (867, 869)
4017 - stop cancelling OST locks when files are closed (481)
4018 - overlapping XID spaces caused network corruption (851, 853)
4019 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4020 - setattr_raw added, to do single-RPC, server-side setattrs
4021 - lmc/lconf syntax change for OST UUIDs
4022 - fix crashy race condition between ptlrpc_free_req and osc_close
4023 - don't use request in mdc_enqueue if we hit a timeout (889)
4024 - don't set the inode i_size for regular files from the MDS (896)
4025 - handle out of order completion AST (842)
4026 - don't LBUG if a lock request times out after receiving AST (913)
4027 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4028 - if a bad lock AST arrives, send an error instead of dropping entirely
4029 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4030 - fix leak in bulk IO when only partially completed (899, 900, 926)
4031 - fix O_DIRECT for ia64 (55)
4032 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4033 - C-z after timeout could hang a process forever; fixed (977)
4035 - client-side I/O cache (678, 924, 929, 941, 970)
4037 - READPAGE and SETATTRs which don't take server-side locks get
4040 2003-02-11 Phil Schwan <phil@clusterfs.com>
4043 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4044 - Fix fsfilt_extN_readpage() to read a full page of directory
4045 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4046 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4047 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4048 - Most intent operations were removed, in favour of a new RPC mode
4049 that does a single RPC to the server and bypasses most of the VFS
4050 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4051 - Aggressively cancel local locks on DLM servers
4052 - mds_reint_unlink sends EA to the client if it's the last nlink.
4053 client uses that EA to unlink OST objects.
4054 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4055 - recursive symlinks were fixed (439)
4056 - fixed NULL deref in DEBUG_REQ
4057 - filter_update_lastobjid no longer calls sync, which annoyed extN
4058 - fixed multi-client small-writes to a single file problem (445)
4059 - fixed mtime updates during file writes (607)
4060 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4061 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4062 - fixed Linux OST size reporting problem (444, 656)
4063 - OST now updates object mtime with writes or setattr (607, 619)
4064 - client verifies file size before zeroing page past EOF (445)
4065 - OST now writes last allocated objid to disk with allocation (108)
4066 - LOV on echo now works (409)
4068 - mds_reint_unlink sends a new buffer, with the EA included. this
4069 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4070 set if a regular file was being unlinked, and it was the last link
4071 - use PtlGet from the target for bulk writes (315)
4072 - OST now updates object mtime with writes or setattr (607, 619)
4073 - LDLM now has a grant-time callback to revalidate locked items, if
4075 - Many MDS operations were reorganized to combat race conditions
4077 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4078 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4080 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4083 - Fully reactivate OST imports after reconnection (512, others)
4084 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4085 - More graceful error handling for truncating on dead OST (515)
4086 - Don't error out unless we're actually accessing dead stripes (474)
4087 - Fix garbage sizes when stripes are missing (410)
4088 - LRU counters were broken, causing constant lock purge (433, 432)
4089 - garbage on read from stripes with failed OSTs (441)
4090 - mark OSCs as active before reconnecting during recovery (438)
4091 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4092 - lfind did not preserve OST order in output (443)
4093 - symlinks cause hung clients, incorrect data (439)
4094 - stop dereferencing request after dropping refcount (457)
4095 - don't LASSERT(spin_is_locked) on non-SMP (455)
4096 - fixes for many rename() bugs
4097 - fstat didn't correctly synchronize attributes (399)
4098 - server must handle lock cancellation during blocking AST prep (487)
4099 - bulk descriptors were free()d too soon (511)
4100 - fix paths in lconf, which would load incorrect modules (451, 507)
4101 - fix confusing lconf 'host not found' error message (386)
4102 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4103 - fix race condition in mdc_blocking_ast() for inode access (526)
4104 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4105 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4106 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4107 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4108 - fix lstripe in several ways (kernel side) (527)
4109 - fix request leak in ldlm_cli_enqueue (262)
4110 - incorrect OSC was marked inactive after OST failure
4111 - call mds_fs_cleanup before unmounting filesystem (524)
4112 - fix races between taking ns_lock and ldlm_lock_change_resource
4113 - fix races updating LOV export open file list
4114 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4115 - fix recovery NULL deref in ldlm_cli_cancel_unused
4116 - fix some DLM races by using new hash table for lock handles (419)
4117 - permit the client to specify desired inodes, at replay
4118 - duplicate requests when we queue them for replay reintegration
4119 - fix last_rcvd offset calculation
4120 - sync after each recovered transaction, so we always make progress
4121 - never, not always, ERESTART requests without transnos
4122 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4124 - skip replay if the MDS says that the client is already connected
4125 - don't check for a recovery-enabled export to match lctl's UUID
4126 - don't INC_USE_COUNT for phantom exports
4127 - don't crash when cleaning up phantom exports (567)
4128 - don't double-finish or set replay data for errored mdc_open requests
4129 - abort requests when they time out, so we don't get old replies
4130 - send/receive replies for AST messages again
4131 - if the client says that it doesn't have the lock, cancel it on the
4133 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4134 mark it as destroyed, it will all work out when decref is called
4135 - fix module use counts (22, 581)
4137 - ASTs now expect a reply (server cancels lock on error reply)
4139 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4142 - fix many simultaneous client startup (392)
4143 - fix dentry->d_it clobbering
4144 - credentials weren't being shipped for readdir/getattr operations
4145 - remove invalid assertions triggered during some concurrent MD
4147 - proper Lustre versions added (336, 389)
4148 - fix memory leak for create error case (398)
4149 - fix LOV locking bug that would get cli/srv out of sync
4150 - fix echo client over LOV (409)
4151 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4152 - fix double-O_EXCL intent crash (424)
4153 - avoid sending multiple lock CANCELs (352)
4155 - MDS can do multi-client recovery (modulo bugs in new code)
4157 - many updates, edits, cleanups
4159 2002-11-18 Phil Schwan <phil@clusterfs.com>
4162 - fix null d_it dereference (346)
4163 - fix full OST/dbench hang (333)
4164 - fix permission problem with file removal (286)
4165 - fix removal of OSCs from LOV when they fail
4166 - fix NULL deref during bulk timeout (214)
4167 - fix problems related to multiple filesystems on one MDS (241)
4168 - fixed serious subtle metadata locking bugs
4169 - free locks on clients when inodes are removed due to memory
4171 - fix inode pointer in lock data (285)
4172 - partial support for multiple MDS on a single host (241)
4173 - data locks weren't cancelled at clear_inode time (290, 311)
4174 - intent locks could lead to unbounded lock growth (205)
4175 - added a maximum lock count, an LRU list, and a flusher
4176 - fix multiple rename (365)
4177 - properly abstracted the echo client
4178 - OSC locked 1 byte too many; fixed
4179 - rewrote brw callback code:
4180 - fixed recovery bugs related to LOVs (306)
4181 - fixed too-many-pages-in-one-write crash (191)
4182 - fixed (again) crash in sync_io_timeout (214)
4183 - probably fixed callback-related race (385)
4185 - Add capability to MDS protocol
4186 - LDLM cancellations and callbacks on different portals
4188 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4191 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4192 - defer open object destruction to close time (601981, 138)
4193 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4194 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4195 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4196 - don't call obd_setattr in ll_file_release for destroyed objects
4197 * protocol change to lustre_msg: move |version| and add |flags|
4198 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4199 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4200 * added replay of create, unlink, link and rename operations during
4201 MDS failover; recovery should be much more robust now
4202 * remove failed OSCs from LOVs (only lov_create uses this so far)
4203 * the lustre-HOWTO was brought (more) up to date (582544)
4205 2002-10-23 Phil Schwan <phil@clusterfs.com>
4208 - in-use dentries weren't being reused properly (617851)
4209 - prevent multiple LDLM setup (599178)
4210 - fix LOV size calculations for truncate (617853)
4211 - fix client handling of MDS intent errors (POSIX)
4212 - fix permission bug in lovstripe.c test (624321)
4213 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4214 - truncate past end of file could corrupt data
4215 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4216 - a race in recovery could return ETIMEDOUT to apps (623947)
4217 - building outside the source directory was fixed
4218 * the lustre-HOWTO was brought (more) up to date (582544)
4219 * major progress was made on recovery functionality
4221 2002-10-10 Phil Schwan <phil@clusterfs.com>
4224 - recovery deadlock fix
4225 - rm -rf causes LBUG fix (617817)
4226 - file open by multiple tasks fix (618962)
4227 - directory permissions bugs (602707 and 620007)
4228 - journal_stop fixed with locking (611313)
4229 - O_APPEND failures resolved (618273, perhaps 614459)
4230 - lconf PATH fix (619770)
4231 - IA64 build fix (621450)
4232 - RPC buffer sizes scale with amount of memory
4234 2002-10-01 Phil Schwan <phil@clusterfs.com>
4237 - locks would be cancelled without throwing away data pages,
4238 resulting in inconsistent data (605627)
4239 - inode attributes were not always being refreshed (605627, 612449)
4240 - lconf now continues to cleanup after lctl reports an error
4241 - MDS now enforces user permissions (602707)
4242 - lprocfs cleanup fixed, but not yet enabled (614157)
4243 - fixed infinite server hang, should a client not respond to an AST
4244 - avoid going into recovery if user calls readlink() with a buffer
4245 that's too small (613941)
4246 - AST RPCs no longer require replies (614867) -- this may be changed
4247 - don't crash server if client sends an IOV that's too big (611336)
4248 - fixed lock conversion deadlock (611892)
4249 - fixed the following of symlinks (614622)
4250 * recovery: the server can remove locks from a client that dies, other
4251 clients can make progress
4252 * more extN patch fixes
4253 * compile-time configurable ptlrpc buffer allocations
4255 - collaborative read cache document
4256 - Lustre Lite Performance CDR document-in-progress
4258 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4261 - fix typo in patch-2.4.18
4263 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4266 - clear ptlrpc request each time in handle_incoming_request()
4267 - unlink of files now destroys the object on the OST
4269 2002-09-19 Peter Braam <braam@clusterfs.com>
4271 * add hard link support
4272 * change obdfile creation method
4273 * kernel patch changed
4275 2002-09-19 Peter Braam <braam@clusterfs.com>
4278 - stack overflow bug in extN fixed
4280 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4282 * documentation updates
4283 - add man pages for config tools
4284 - update tests/README to describe testing with new config tools
4285 - finish metadata API descriptions
4286 * bug fixes and cleanups
4287 - statfs workaround for 16TB limit
4288 - LOV stripe allocation improved, can stripe on subset of OSTs
4289 - LOV file size/IO offset was wrong for files > 4GB in size
4290 - object EA data was being dropped, caused files to be unreadable
4291 - memory overflow with non-LOV OST caused memory corruption
4292 - fixed regression tests to work with new config tools, obdfilter
4293 - fixed bug when directory size became larger than 1 block
4294 - fixed bug (for single client case) when PWD was deleted
4295 - invalidate local directory pages when doing intent-based ops
4296 - avoid LDLM oops when lock callback contained bad data
4298 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4300 * documentation updates
4301 * bug fixes and cleanups
4302 - configuration tools
4305 - 64-bit compile warnings
4306 - 64-bit internal statfs data
4308 * test_brw on persistent OST devices
4310 * lprocfs (disabled)
4312 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4314 * documentation updates
4315 * bug fixes and cleanups
4316 * configuration tools
4318 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4321 * many small fixes to 0_5_4
4322 * io/network handling
4323 * thinkos in MDS operations
4325 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4328 * crucial basic fixes to 0.5.3
4329 * IOR, Iozone work over Elan
4332 2002-08-07 Phil Schwan <phil@clusterfs.com>
4333 * version 0_5_3, our first alpha
4334 * we use the new Portals iovs
4335 * documentation updates
4336 * bug fixes and cleanups
4337 * small changes in the DLM wire protocol
4339 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4340 * version 0_5_1 with some initial stability,
4341 * locking on MD and file I/O.
4342 * documentation updates
4343 * several bug fixes since 0.5.0
4344 * small changes in wire protocol
4346 2002-07-18 Phil Schwan <phil@clusterfs.com>
4348 * delivered as Lustre Light Alpha
4349 * fixed a crash after handling invalid MDS requests
4350 * fixed directory pages for architectures with non-4k pages sizes
4352 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4353 * release version v0_4_4
4354 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4356 * Fixes a number of interruption problems with OST operations.
4357 * Update documentation for portals header changes
4358 * Move all wire protocol structs/defines to lustre_idl.h
4359 * Fixes symlink length bug.
4360 * Add tcpdump to repository.
4362 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4363 * release version v0_4_3
4364 * Fixes statfs for inodes on extN.
4365 * Fixes bug in runtests which would delete /etc/hosts.
4366 * Use 64-bit object IDs wherever possible (not into VFS though)
4367 Remove ost_get_info, which is unused by lustre, and out of date.
4369 2002-07-03 Peter Braam <braam@clusterfs.com>
4370 * release version v0_4_2 Fixes a lookup error (type not passed)
4371 * move forward to head of Portals
4372 * move forward to latest Lustre kernel
4374 2002-06-25 Peter Braam <braam@clusterfs.com>
4375 * release version v0_4_1. Hopefully stable on single node use.