1 tbd Sun Microsystems, Inc.
4 2.6.9-67.0.4.EL (RHEL 4),
5 2.6.16.54-0.2.5 (SLES 10),
6 2.6.18-53.1.13.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).
16 Frequency : occasional
18 Description: Correctly check stale fid, not start epoch if ost not support SOM
19 Details : open with flag O_CREATE need set old fid in op_fid3 because op_fid2
20 overwrited with new generated fid, but mds can anwer with one of these
21 two fids and both is not stale. setattr incorectly start epoch and
22 assume will be called done_writeting, but without SOM done_writing
26 Frequency : rare, depends on device drivers and load
28 Description: MDS or OSS nodes crash due to stack overflow
29 Details : Code changes in 1.8.0 increased the stack usage of some functions.
30 In some cases, in conjunction with device drivers that use a lot
31 of stack the MDS (or possibly OSS) service threads could overflow
32 the stack. One change which was identified to consume additional
33 stack has been reworked to avoid the extra stack usage.
36 Frequency : occasional
38 Description: Do not fail import if osc_interpret_create gets -EAGAIN
39 Details : If osc_interpret_create got -EAGAIN it immediately exits and
40 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
41 oscc_has_objects and see OSC has no objests and call
42 oscc_internal_create to resend create request.
44 Severity : enhancement
46 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
48 Severity : enhancement
50 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
55 Description: Fix replay if there is an un-replied request and open
56 Details : In some cases, older replay request will revert the
57 mcd->mcd_last_xid on MDS which is used to record the client's
60 Severity : enhancement
62 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
64 Severity : enhancement
66 Description: Add rhel5 support to HEAD.
68 Serverity : enhancement
70 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
75 Description: Don't allow skipping OSTs if index has been specified.
76 Details : Don't allow skipping OSTs if index has been specified, make locking
77 in internal create lots better.
81 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
82 Details : access to bitfield in structure is always rounded to long
83 and this produce problem with not atomic change any bit.
87 Description: Lustre make rpms failed.
88 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
89 builds Lustre rpms from tarball.
92 Frequency : rare on shutdown ost
94 Description: If llog cancel was not send before clean_exports phase, this can
95 produce deadlock in llog code.
96 Details : If llog thread has last reference to obd and call class_import_put
97 this produce deadlock because llog_cleanup_commit_master wait when
98 last llog_commit_thread exited, but this never success because was
99 called from llog_commit_thread.
103 Description: allow userland application know is lost one of stripes.
104 Details : fill lvb_blocks with error code on ost and return it to
105 application if error flag found.
109 Description: NULL lov_tgts causing MDS oops
110 Details : more safe checks for NULL lov_tgts for avoid oops.
112 Severity : enhancement
114 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
118 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
119 Details : If an OST is down the MDS will hang indefinitely in
120 obd_statfs_rqset() waiting for the statfs data. While for
121 MDS QOS usage of statfs, it should not stuck in waiting.
123 Severity : enhancement
125 Description: remote_acl support
126 Details : Support ACL-based permission check for remote user.
127 Support setfacl/getfacl for remote user with the utils
128 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
129 the system "{s,g}etfacl" utils.
131 Severity : enhancement
133 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
135 Severity : enhancement
137 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
141 Description: make mgs_setparam() handle fsname containing dash
142 Details : fsname containing a dash does not work with lctl conf_param
144 Severity : enhancement
146 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
148 Severity : enhancement
150 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
154 Frequency : rare, at shutdown
155 Description: access already free / zero obd_namespace.
156 Details : if client_disconnect_export was called without force flag set,
157 and exist connect request in flight, this can produce access to
158 NULL pointer (or already free pointer) when connect_interpret
159 store ocd flags in obd_namespace.
163 Frequency : only at startup
164 Description: not alloc memory with spinlock held.
165 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
166 if any spinlock held.
168 Severity : enhancement
170 Description: make lustre randomly fail allocating memory
171 Details : Make lustre randomly failed allocating memory for testing purpose.
173 Severity : enhancement
175 Description: lost problems with lov objid file
176 Details : Fixes some scability and access to not inited memory problems
177 in work with lov objdid file.
182 Description: lfs find does not continue on file error
183 Details : Continue other files processing when a file/dir is absent.
187 Description: Inconsistent usage of lustre_pack_reply()
188 Details : Standardize the usage of lustre_pack_reply() such that it
189 always generate a CERROR on failure.
192 Frequency : occasional
194 Description: MDS hang or stay in waiting lock
195 Details : If client receive lock with CBPENDING flag ldlm need send lock
196 cancel as separate rpc, to avoid situation when cancel request
197 can't processed due all i/o threads stay in wait lock.
201 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
205 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
209 Description: Update to SLES9 kernel-2.6.5-7.287.3.
213 Description: Update to RHEL4 latest kernel.
215 Severity : enhancement
217 Description: Build SLES10 patchless client fails
218 Details : The configure was broken by run ./configure with
219 --with-linux-obj=.... argument for patchless client. When the
220 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
221 can't search header adequately. Use absolute path such as
222 -I($LINUX)/include instead.
226 Description: interrupt oig_wait produce painc on resend.
227 Details : brw_redo_request can be used for resend requests from ptlrpcd and
228 private set, and this produce situation when rq_ptlrpcd_data not
229 copyed to new allocated request and triggered LBUG on assert
230 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
231 ptlrpcd set if request is changed and can be safety changed to use
236 Description: LASSERT_{REQ,REP}SWAB macros are buggy
237 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
238 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
239 these macros and replace them with their difinitions instead.
243 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
244 Details : Update which_patch & target file for SLES10 latest kernel.
247 Description: Remove client patches from SLES 10 kernel.
248 Details : This causes SLES 10 clients to behave as patchless clients
249 even on a Lustre-patched (server) kernel.
251 Severity : enhancement
253 Description: self-adjustable client's lru lists
254 Details : use adaptive algorithm for managing client cached locks lru
255 lists according to current server load, other client's work
256 pattern, memory activities, etc. Both, server and client
257 side namespaces provide number of proc tunables for controlling
260 Severity : enhancement
262 Description: light-weight GSS support
263 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
264 and reduce performance overhead.
266 Severity : enhancement
268 Description: Linux keyring support
269 Details : Support using service of Linux keyring for Lustre GSS internal
270 context refresh/cache mechanism.
274 Description: Fix errors in lfs documentation
275 Details : Fixes man pages
279 Description: don't use GFP_* in generic Lustre code.
280 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
284 Description: obdclass is limited by single OBD_ALLOC(idarray)
285 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
289 Description: warnings with build patchless client with vanila 2.6.19 and up
290 Details : change old ctl_table style and replace ctl_table/ctl_table_header
291 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
295 Frequency : Only for SLES
296 Description: sanity_quota fail test_1
297 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
298 file, which makes TSTID[2] inunique.
302 Frequency : Only for SLES9
303 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
304 Details : Update target/ChangeLog/which_patch .
308 Frequency : Only for test_10 in sanity-quota.sh
309 Description: a bug in quota test code
310 Details : checking a test flag in a wrong place
314 Frequency : Only for test_13 in sanity-quota.sh
315 Description: a bug in quota test script
316 Details : shouldn't check overfull allocated quota
320 Description: jbd statistics
321 Details : Port older jbd statistics patch for sles10
323 Severity : enhancement
325 Description: Add printing inode info into message about error in writepage.
330 Description: reply_lock_interpret crash due to race with it and lock cancel.
331 Details : Do not replay locks that are being cancelled. Do not reference
332 locks by their address during replay, just by their handle.
336 Frequency : When flocks are used.
337 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
338 Details : Flock locks might destroy just granted lock if it could be merged
339 with another existing flock, this is done in completion handler,
340 so teach ldlm_cli_enquque_fini that this is a valid case for
346 Description: Oops in read and write path when failing to allocate lock.
347 Details : Check if lock allocation failed and return error back.
351 Description: lstripe command fails for valid OST index
352 Details : The stripe offset is compared to lov->desc.ld_tgt_count
353 instead of lov->desc.ld_active_tgt_count.
357 Description: sanity.sh failed test 103
358 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
363 Description: df doesn't work properly if diskfs blocksize != 4K
364 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
368 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
369 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
373 Description: obd_health_check_timeout too short
374 Details : set obd_health_check_timeout as 1.5x of obd_timeout
378 Description: llapi_file_create() does not allow some changes
379 Details : add llapi_file_open() that allows specifying the mode and
380 open flags, and also returns an open file handle.
384 Description: merge and cleanup kernel patches.
385 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
389 Description: Add journal checksum support.(Kernel part)
390 Details : The journal checksum feature adds two new flags i.e
391 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
392 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
393 indicates that the commit block contains the checksum for
394 the blocks described by the descriptor blocks. Now commit
395 record can be sent to disk without waiting for descriptor
396 blocks to be written to disk. This behavior is controlled
397 using JBD2_FEATURE_ASYNC_COMMIT flag.
401 Description: OSS needs mutliple precreate threads
402 Details : Add ability to start more than one create thread per OSS.
406 Description: Sanity.sh test_65k failed: setstripe should have succeeded
407 Details : execute lfs setstripe on client
411 Description: mds_obd_create error creating tmp object
412 Details : When the user sets quota on root, llog will be affected and can't
413 create files and write files.
416 Frequency : Always on ia64 patchless client, and possibly others.
418 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
419 Details : This allows the patchless client to be loaded on architectures
424 Description: RedHat Update kernel for RHEL5
425 Details : Add the kernel config file for RHEL5.
429 Description: RedHat Update kernel for RHEL5
430 Details : Modify the kernel config file more closer RHEL5.
434 Description: Build failure against Centos5 (RHEL5)
435 Details : Define PAGE_SIZE when it isn't present.
439 Description: test_7 on sanity-quota.sh can't be used on separate machines
440 Details : add facet to handle it
444 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
445 Details : with NFS, the anon dentry's parent was set to itself in
446 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
447 corresponding dentry other than use rec->ur_name.
449 Severity : enhancement
451 Description: lfs setstripe enhancement
452 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
456 Description: enable data checksumming by default
457 Details : enable checksum by default, allow --disable-checksum
458 configure option and "-o nochecksum" mount option. Checksums
459 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
463 Description: lustre support for RHEL5
464 Details : Add support for RHEL5.
468 Description: Client eviction due to failover config
469 Details : after a connection loss, the lustre client should attempt to
470 reconnect to the last active server first before trying the
471 other potential connections.
475 Description: when mds and osts use different quota unit(32bit and 64bit),
476 quota will be released repeatly.
477 Details : void sending multiple quota reqs to mds, which will keep the
478 status between the reqs.
482 Description: osts not allocated evenly to files
483 Details : change the condition to increase offset_idx
486 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
488 Description: Data corruption for OSTs that are formatted larger than 2TB
490 Details : When generating the bio request for lustre file writes the
491 sector number would overflow a temporary variable before being
492 used for the IO. The data reads correctly from Lustre (which
493 will overflow in a similar manner) but other file data or
494 filesystem metadata may be corrupted in some cases.
498 Description: Tune the kernel for good SCSI performance.
499 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
500 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
505 Description: rewrite ext2-derived code in obdclass/uuid.c
506 Details : rewrite inherited code (uuid parsing code from ext2 utils)
507 from scratch preserving functionality.
511 Description: "lfs find -obd UUID" prints directories
512 Details : "lfs find -obd UUID" will return all directory names instead
513 of just file names. It is incorrect because the directories
514 do not reside on the OSTs.
518 Description: use i_size_read and i_size_write in 2.6 port
519 Details : replace inode->i_size access with i_size_read/write()
522 Frequency : when removing large files
524 Description: scheduling issue during removal of large Lustre files
525 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
526 It causes scheduling issues when removing large files (17TB in the
530 Frequency : only with liblustre clients on XT3
532 Description: evictions taking too long
533 Details : allow llrd to evict clients directly on OSTs
536 Frequency : only on ppc
538 Description: /proc/fs/lustre/devices broken on ppc
539 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
540 We should make sure the type of 'index' is loff_t and then cast
541 explicitly as needed below. Do not assign an explicitly cast
545 Frequency : only for rhel5
547 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
548 Details : Modify the target file & which_kernel.
550 Severity : enhancement
552 Description: omit set fsid for export NFS
553 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
554 as FS_REQUIRES_DEV for avoid problems with generate fsid.
558 Frequency : Always, for kernels after 2.6.16
559 Description: Fix warning idr_remove called for id=.. which is not allocated.
560 Details : Last kernels save old s_dev before kill super and not allow
561 to restore from callback - restore it before call kill_anon_super.
565 Description: buffer overruns could theoretically occur
566 Details : llapi_semantic_traverse() modifies the "path" argument by
567 appending values to the end of the origin string, and a
568 overrun may occur. Adding buffer overrun check in liblustreapi.
572 Description: Fix error on 'ls .' at the top of the Lustre mount.
573 Details : Don't revalidate dentry if it is a root dentry.
577 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
578 Details : Modify vm-tunables-rhel4.patch.
582 Description: Kernel config for 2.6.18-vanilla.
583 Details : Modify targets/2.6-vanilla.target.in.
584 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
585 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
586 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
587 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
591 Description: improve handling recoverable errors
592 Details : if request processig with error which can be recoverable on server
593 request should be resend, otherwise page released from cache and
598 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
599 Details : Modify target file & which_patch.
600 A flaw was found in the IA32 system call emulation provided
601 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
602 value could be stored in the %RAX register, which could trigger an
603 out-of-bounds system call table access. An untrusted local user
604 could exploit this flaw to run code in the kernel
605 (ie a root privilege escalation). (CVE-2007-4573).
609 Description: O_DIRECT bypasses client statistics.
610 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
611 read_ahead_stats were updated. copied the stats section from
612 osc_send_oap_rpc() into async_internal().
616 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
620 Description: change order of libsysio includes
621 Details : '#include sysio.h' should always come before '#include xtio.h'
625 Description: handle "serious error: objid * already exists" more gracefully
626 Details : If LAST_ID value on disk is smaller than the objects existing in
627 the O/0/d* directories, it indicates disk corruption and causes an
628 LBUG(). If the object is 0-length, then we should use the existing
629 object. This will help to avoid a full fsck in most cases.
631 Severity : enhancement
633 Description: adapt the lustre_config script to support the upgrade case
634 Details : Add "-u" option for lustre_config script to support upgrading 1.4
635 server targets to 1.6 in parallel.
639 Description: To avoid grant space > avaible space when the disk is almost
640 full. Without this patch you might see the error "grant XXXX >
641 available" or some LBUG about grant, when the disk is almost
643 Details : In filter_check_grant, for non_grant cache write, we should
644 check the left space by if (*left > ungranted + bytes), instead
645 of (*left > ungranted), because only we are sure the left space
646 is enough for another "bytes", then the ungrant space should be
647 increase. In client, we should update cl_avail_grant only there
648 is OBD_MD_FLGRANT in the reply.
652 Description: Update RHEL 4 kernel to fix local root privilege escalation.
653 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
654 described in CVE-2007-4573. This problem could allow untrusted
655 local users to gain root access.
658 Frequency : when using O_DIRECT and quotas
660 Description: Incorrect file ownership on O_DIRECT output files
661 Details : block usage reported by 'lfs quota' does not take into account
662 files that have been written with O_DIRECT.
667 Description: touch file failed when fs is not full
668 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
669 otherwise we can get ENOSP while fs is not full.
673 Description: parallel lock callbacks
674 Details : Instead of sending blocking and completion callbacks as separated
675 requests, adding them to a set and sending in parallel.
678 Frequency : only for Cray XT3
679 Bugzilla : 12829/13455
680 Description: Changing primary group doesn't change the group lustre assigns to
682 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
683 overridden with the primary group provided by the group upcall,
684 whereas the supplied fsgid can be trusted if it is in the list of
685 supplementary groups returned by the group upcall.
687 Severity : enhancement
689 Description: Allow masking D_WARNING, D_ERROR messages from console
690 Details : Console messages can now be disabled via lnet.printk.
694 Description: User code with malformed file open parameter crashes client node
695 Details : Before packing join_file req, all the related reference should be
696 checked carefully in case some malformed flags cause fake
697 join_file req on client.
701 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
702 Details : ldlm_completion_ast() assumes that a lock is granted when the req
703 mode is equal to the granted mode. However, it should also check
704 that LDLM_FL_CP_REQD is not set.
708 Description: Heavy nfs access might result in deadlocks
709 Details : After ELC code landed, it is now improper to enqueue any mds
710 locks under och_sem, because enqueue might want to decide to
711 cancel open locks for same inode we are holding och_sem for.
715 Description: Client eviction while running blogbench
716 Details : A lot of unlink operations with concurrent I/O can lead to a
717 deadlock causing evictions. To address the problem, the number of
718 oustanding OST_DESTROY requests is now throttled to
719 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
720 callbacks are processed in priority.
724 Description: enable ACLs on MDS by default
725 Details : ACLs must be enabled on MDS by default.
728 Frequency : PPC/PPC64 only
730 Description: conflicts between asm-ppc64/types.h and lustre_types.h
731 Details : fix duplicated definitions between asm-ppc64/types.h and
732 lustre_types.h on PPC.
735 Frequency : PPC/PPC64 only
737 Description: asm-ppc/segment.h does not exist
738 Details : fix compile issue on PPC.
742 Description: better handle error messages in extents code
744 --------------------------------------------------------------------------------
746 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
748 * Support for kernels:
749 2.6.5-7.283 (SLES 9),
750 2.6.9-55.EL (RHEL 4),
751 2.6.16.46-0.14 (SLES 10),
752 2.6.18.8 vanilla (kernel.org)
753 * Client support for unpatched kernels:
754 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
755 2.6.16 - 2.6.22 vanilla (kernel.org)
756 * Due to recently discovered recovery problems, we do not recommend
757 using patchless RHEL 4 clients with this or any earlier release.
758 * Recommended e2fsprogs version: 1.39.cfs8
759 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
760 * Starting with this release, the ldiskfs backing filesystem required
761 by Lustre is now in its own package, lustre-ldiskfs. This package
762 should be installed. It is versioned separately from Lustre and
763 may be released separately in future.
767 Description: block reactivating mgc import until all deactivates complete
768 Details : Fix race when failing back MDT/MGS to itself (testing)
771 Frequency : at statup only
773 Description: mds_lov_synchronize race leads to various problems
774 Details : simultaneous MDT->OST connections at startup can cause the
775 sync to abort, leaving the OSC in a bad state.
777 Severity : enhancement
779 Description: add optional extra BUILD_VERSION info
780 Details : add a new environment variable (namely LUSTRE_VERS) which allows
781 to override the lustre version.
784 Frequency : 2.6.18 servers only
786 Description: ll_kern_mount() doesn't release the module reference
787 Details : The ldiskfs module reference count never drops down to 0
788 because ll_kern_mount() doesn't release the module reference.
793 Description: server LBUG when using old ost_num_threads parameter
794 Details : Accept the old ost_num_threads parameter but warn that it
795 is deprecated, and fix an off-by-one error that caused an LBUG.
800 Description: Transient SCSI error results in persistent IO issue
801 Details : iobuf->dr_error is not reinitialized to 0 between two
805 Frequency : sometimes when underlying device returns I/O errors
807 Description: OSTs not going read-only during write failures
808 Details : OSTs are not remounted read-only when the journal commit threads
809 get I/O errors because fsfilt_ext3 calls journal_start/stop()
810 instead of the ext3 wrappers.
814 Description: poor connect scaling with increasing client count
815 Details : Don't run filter_grant_sanity_check for more than 100 exports
816 to improve scaling for large numbers of clients.
819 Frequency : SLES10 only
821 Description: sanity-quota.sh quotacheck failed: rc = -22
822 Details : Quotas cannot be enabled on SLES10.
825 Frequency : liblustre clients only
827 Description: getdirentries does not give error when run on compute nodes
828 Details : getdirentries does not fail when the size specified as an argument
829 is too small to contain at least one entry
831 Severity : enhancement
833 Description: Add LNET router traceability for debug purposes
834 Details : If a checksum failure occurs with a router as part of the
835 IO path, the NID of the last router that forwarded the bulk data
836 is printed so it can be identified.
841 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
842 Details : Due to a race condition, liblustre clients were occasionally
845 Severity : enhancement
847 Description: lfs setstripe use optional parameters instead of postional
850 Severity : enhancement
852 Description: Nanosecond timestamp support for ldiskfs
853 Details : The on-disk ldiskfs filesystem has added support for nanosecond
854 resolution timestamps. There is not yet support for this at
855 the Lustre filesystem level.
858 Frequency : during server recovery
860 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
861 Details : request with rq_no_resend flag not awake l_wait_event if they get a
865 Frequency : nfs export on patchless client
867 Description: connectathon hang when test nfs export over patchless client
868 Details : Disconnected dentry cannot be found with lookup, so we do not need
869 to unhash it or make it invalid
872 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
873 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
874 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
879 Description: open req refcounting wrong on reconnect
880 Details : If reconnect happened between getting open reply from server and
881 call to mdc_set_replay_data in ll_file_open, we will schedule
882 replay for unreferenced request that we are about to free.
883 Subsequent close will crash in variety of ways.
884 Check that request is still eligible for replay in
885 mdc_set_replay_data().
890 Description: disable writes to filesystem when reading health_check file
891 Details : the default for reading the health_check proc file has changed
892 to NOT do a journal transaction and write to disk, because this
893 can cause reads of the /proc file to hang and block HA state
894 checking on a healthy but otherwise heavily loaded system. It
895 is possible to return to the previous behaviour during configure
896 with --enable-health-write.
898 Severity : enhancement
900 Description: 64-bit inode version
901 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
902 to the inode. This will be required for version-based recovery.
907 Description: MDS fails to start if a duplicate client export is detected
908 Details : in some rare cases it was possible for a client to connect to
909 an MDS multiple times. Upon recovery the MDS would detect this
910 and fail during startup. Handle this more gracefully.
912 Severity : enhancement
914 Description: Add -o localflock option to simulate old noflock
916 Details : This will achieve local-only flock/fcntl locks
922 Description: log_commit_thread vs filter_destroy race leads to crash
923 Details : Take import reference before releasing llog record semaphore
928 Description: Wrong request locking in request set processing
929 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
930 imp_delayed_list, in this place should be used imp_lock.
933 Frequency : when reconnection
935 Description: Grant Leak when osc reconnect to OST
936 Details : When osc reconnect ost, OST(filter) should check whether it
937 should grant more space to client by comparing fed_grant and
938 cl_avail_grant, and return the granted space to client instead
939 of "new granted" space, because client will call osc_init_grant
940 to update the client grant space info.
943 Frequency : when client reconnect to OST
945 Description: Grant Leak when osc do resend and replay bulk write
946 Details : When osc reconnect to OST, OST(filter)should clear grant info of
947 bulk write request, because the grant info while be sync between
948 OSC and OST when reconnect, and we should ignore the grant info
949 these of resend/replay write req.
954 Description: Grant space more than avaiable left space sometimes.
955 Details : When then OST is about to be full, if two bulk writing from
956 different clients came to OST. Accord the avaliable space of the
957 OST, the first req should be permitted, and the second one
958 should be denied by ENOSPC. But if the seconde arrived before
959 the first one is commited. The OST might wrongly permit second
960 writing, which will cause grant space > avaiable space.
963 Frequency : when client is evicted
965 Description: Grant might be wrongly erased when osc is evicted by OST
966 Details : when the import is evicted by server, it will fork another
967 thread ptlrpc_invalidate_import_thread to invalidate the
968 import, where the grant will be set to 0. While the original
969 thread will update the grant it got when connecting. So if
970 the former happened latter, the grant will be wrongly errased
971 because of this race.
976 Description: Checking Stale with correct fid
977 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
978 to check whether it is stale, because sometimes, we want the
979 enqueue happened anyway, and op_data.fid2 will not be initialized.
981 Severity : enhancement
983 Description: update patchless client
984 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
987 Frequency : only with 2.4 kernel
989 Description: random memory corruption
990 Details : size of struct ll_inode_info is to big for union inode.u and this
991 can be cause of random memory corruption.
996 Description: Memory leak in recovery
997 Details : Lov_mds_md was not free in an error handler in mds_create_object.
998 It should also check obd_fail before fsfilt_start, otherwise if
999 fsfilt_start return -EROFS,(failover mds during mds recovery).
1000 then the req will return with repmsg->transno = 0 and rc = EROFS.
1001 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1002 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1003 be freed no matter whether fsfilt_commit success or not.
1006 Frequency : only with huge count clients
1008 Description: Prevents from taking the superblock lock in llap_from_page for
1010 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1011 for llap_from_page call in ll_removepage prevents from taking the
1012 superblock lock for a soon died page.
1017 Description: Not check open intent error before release open handle
1018 Details : in some rare cases, the open intent error is not checked before
1019 release open handle, which may cause
1020 ASSERTION(open_req->rq_transno != 0), because it tries to release
1021 the failed open handle.
1026 Description: Set cat log bitmap only after create log success.
1027 Details : in some rare cases, the cat log bitmap is set too early. and it
1028 should be set only after create log success.
1032 Description: Accessing a block bevice can re-enable I/O when Lustre is
1033 tearing down a device.
1034 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1039 Description: service threads may hog cpus when there are a lot of requests
1040 Details : Insert cond_resched to give other threads a chance to use some CPU
1045 Description: the cat log was not initialized in recovery
1046 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1047 unlink log on mds will not be initialized until mds post
1048 recovery. And also in mds post recovery, the unlink log will
1049 initialization will be done asynchronausly, so there will be race
1050 between add unlink log and unlink log initialization.
1054 Description: brw_stats were being printed incorrectly
1055 Details : brw_stats were being printed as log2 but all of them were not
1056 recorded as log2. Also remove some code duplication arising from
1057 filter_tally_{read,write}.
1061 Frequency : rare, only in recovery.
1062 Description: ASSERTION(req->rq_type != LI_POISON) failed
1063 Details : imp_lock should be held while iterating over imp_sending_list for
1064 prevent destroy request after get timeout in ptlrpc_queue_wait.
1068 Description: replay-single.sh test 52 fails
1069 Details : A lock's skiplist need to be cleanup when it being unlinked
1070 from its resource list.
1074 Description: Short directio read returns full requested size rather than
1076 Details : Direct I/O operations should return actual amount of bytes
1077 transferred rather than requested size.
1079 Severity : enhancement
1081 Description: metadata RPC reduction (e.g. for rm performance)
1082 Details : decrease the amount of synchronous RPC between clients and servers
1083 by canceling conflicing lock before the operation on the client
1084 and packing thier handles into the main operation RPC to server.
1086 Severity : enhancement
1088 Description: Async OSC create to avoid the blocking unnecessarily.
1089 Details : If a OST has no remain object, system will block on the creating
1090 when need to create a new object on this OST. Now, ways use
1091 pre-created objects when available, instead of blocking on an
1092 empty osc while others are not empty. If we must block, we block
1093 for the shortest possible period of time.
1095 Severity : enhancement
1097 Description: refine locking for avoid write wrong info into lov_objid file
1098 Details : fix possible races with add new target and write/update data in
1101 --------------------------------------------------------------------------------
1103 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1108 Frequency : on some architectures
1110 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1111 Details : Uninitialized flags sometimes cause configuration commands to
1115 Frequency : patchless clients only
1117 Description: missing __iget() symbol export
1118 Details : The __iget() symbol export is missing. To avoid the need for
1119 this on patchless clients the deathrow inode reaper is turned
1120 off, and we depend on the VM to clean up old inodes. This
1121 dependency was during via the fix for bug 12181.
1123 --------------------------------------------------------------------------------
1125 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1127 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1128 INTEROPERATE with older versions automatically. In many cases a
1129 special upgrade step is needed. Please read the
1130 user documentation before upgrading any part of a live system.
1131 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
1132 version will not interoperate with 1.6 betas before beta5 (1.5.95).
1133 * WARNING: Lustre configuration and startup changes are required with
1134 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1140 Severity : enhancement
1142 Description: MountConf
1143 Details : Lustre configuration is now managed via mkfs and mount
1144 commands instead of lmc and lconf. New obd types (MGS, MGC)
1145 are added for dynamic configuration management. See
1146 https://mail.clusterfs.com/wikis/lustre/MountConf for
1149 Severity : enhancement
1151 Description: dynamic OST addition
1152 Details : OSTs can now be added to a live filesystem
1154 Severity : enhancement
1156 Description: startup order invariance
1157 Details : MDTs and OSTs can be started in any order. Clients only
1158 require the MDT to complete startup.
1160 Severity : enhancement
1162 Description: parallel, asynchronous orphan cleanup
1163 Details : orphan cleanup is now performed in separate threads for each
1164 OST, allowing parallel non-blocking operation.
1166 Severity : enhancement
1168 Description: optimized stripe assignment
1169 Details : stripe assignments are now made based on ost space available,
1170 ost previous usage, and OSS previous usage, in order to try
1171 to optimize storage space and networking resources.
1173 Severity : enhancement
1175 Description: Permanently set tunables
1176 Details : All writable /proc/fs/lustre tunables can now be permanently
1177 set on a per-server basis, at mkfs time or on a live system.
1179 Severity : enhancement
1181 Description: Lustre message v2
1182 Details : Add lustre message format v2.
1184 Severity : enhancement
1186 Description: client OST exclusion list
1187 Details : Clients can be started with a list of OSTs that should be
1188 declared "inactive" for known non-responsive OSTs.
1192 Description: SPEC SFS validation failure on NFS v2 over lustre.
1193 Details : Changes the blocksize for regular files to be 2x RPC size,
1194 and not depend on stripe size.
1196 Severity : enhancement
1198 Description: Multiple MD RPCs in flight.
1199 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1200 To avoid overly-overloading MDS, introduce a limit on number of
1201 MDS RPCs in flight for a single client and add /proc controls
1202 to adjust this limit.
1204 Severity : enhancement
1206 Description: client read/write statistics
1207 Details : Add client read/write call usage stats for performance
1208 analysis of user processes.
1209 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1210 file access. extents_stats shows chunk size distribution.
1211 extents_stats_per_process show chunk size distribution per
1214 Severity : enhancement
1216 Description: mds statistics
1217 Details : Add detailed mds operations statistics in
1218 /proc/fs/lustre/mds/*/stats.
1222 Description: Failure of copying files with lustre special EAs.
1223 Details : Client side always return success for setxattr call for lustre
1224 special xattr (currently only "trusted.lov").
1229 Description: Refcount LNET uuids
1230 Details : The global LNET uuid list grew linearly with every startup;
1231 refcount repeated list entries instead of always adding to
1234 Severity : enhancement
1236 Description: Dynamic service threads
1237 Details : Within a small range, start extra service threads
1238 automatically when the request queue builds up.
1241 Frequency : mixed-endian client/server environments
1243 Description: mixed-endian crashes
1244 Details : The new msg_v2 system had some failures in mixed-endian
1247 Severity : enhancement
1249 Description: Easy OST removal
1250 Details : OSTs can be permanently deactivated with e.g. 'lctl
1251 conf_param lustre-OST0001.osc.active=0'
1253 Severity : enhancement
1255 Description: MGS proc entries
1256 Details : Added basic proc entries for the MGS showing what filesystems
1259 Severity : enhancement
1261 Description: provide MGS failover
1262 Details : Added config lock reacquisition after MGS server failover.
1264 Severity : enhancement
1266 Description: add Linux 2.4 support
1267 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1271 Description: a large application tries to do I/O to the same resource and dies
1272 in the middle of it.
1273 Details : Check the req->rq_arrival time after the call to
1274 ost_brw_lock_get(), but before we do anything about
1275 processing it & sending the BULK transfer request. This
1276 should help move old stale pending locks off the queue as
1277 quickly as obd_timeout.
1280 Frequency : when an incorrect nid is specified during startup
1282 Description: ptlrpc connect to non-existant node causes kernel crash
1283 Details : LNET can't be re-entered from an event callback, which
1284 happened when we expire a message after the export has been
1285 cleaned up. Instead, hand the zombie cleanup off to another
1288 Severity : enhancement
1290 Description: plain/inodebits lock performance improvement
1291 Details : Grouping plain/inodebits in granted list by their request modes
1292 and bits policy, thus improving the performance of search through
1296 Frequency : only if OST filesystem is corrupted
1298 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1299 Details : for a short time RPCs with bulk IO are in the replay list,
1300 but replay of bulk IOs is unimplemented. If the OST filesystem
1301 is corrupted due to disk cache incoherency and then replay is
1302 started it is possible to trip an assertion. Avoid putting
1303 committed RPCs into the replay list at all to avoid this issue.
1306 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1308 Description: System hang on startup
1309 Details : This bug allowed the liblustre (e.g. catamount) client to
1310 return to the app before handling all startup RPCs. This
1311 could leave the node unresponsive to lustre network traffic
1312 and manifested as a server ptllnd timeout.
1314 Severity : enhancement
1316 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1317 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1318 boolean to control whether to print peer debug info when a
1319 client's RPC times out.
1322 Frequency : only for kernels with patches from Lustre below 1.4.3
1324 Description: Remove old rdonly API
1325 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1328 Frequency : only for devices with external journals
1330 Description: Set external device read-only also
1331 Details : During a commanded failover stop, we set the disk device
1332 read-only while the server shuts down. We now also set any
1333 external journal device read-only at the same time.
1336 Frequency : when upgrading from 1.4 while trying to change parameters
1338 Description: The wrong (new) MDC name was used when setting parameters for
1339 upgraded MDT's. Also allows changing of OSC (and MDC)
1340 parameters if --writeconf is specified at tunefs upgrade time.
1343 Frequency : when setting specific ost indicies
1345 Description: QOS code breaks on skipped indicies
1346 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1347 created with --index need not be sequential.
1349 Severity : enhancement
1351 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1352 Details : The uninit_groups feature works in conjunction with the kernel
1353 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1354 up the pass1 processing of e2fsck. This is a read-only feature
1355 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1356 mount filesystems that have had this feature enabled.
1358 Severity : enhancement
1360 Description: Improve multi-block allocation algorithm to avoid fragmentation
1361 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1362 allocation locality and avoid filesystem fragmentation.
1364 ------------------------------------------------------------------------------
1366 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1368 * Support for kernels:
1369 2.4.21-47.0.1.EL (RHEL 3)
1370 2.6.5-7.283 (SLES 9)
1371 2.6.9-42.0.10.EL (RHEL 4)
1372 2.6.12.6 vanilla (kernel.org)
1373 2.6.16.27-0.9 (SLES 10)
1374 * Recommended e2fsprogs version: 1.39.cfs5
1376 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1380 Frequency : occasional, depends on client load and configuration
1381 Bugzilla : 12181, 12203
1382 Description: data loss for recently-modified files
1384 Details : In some cases it is possible that recently written or created
1385 files may not be written to disk in a timely manner (this should
1386 normally be within 30s unless client IO load is very high).
1387 The problem appears as zero-length files or files that are a
1388 multiple of 1MB in size after a client crash or client eviction
1389 that are missing data at the end of the file.
1391 This problem is more likely to be hit on clients where files are
1392 repeatedly created and unlinked in the same directory, clients
1393 have a large amount of RAM, have many CPUs, the filesystem has
1394 many OSTs, the clients are rebooted frequently, and/or the files
1395 are not accessed by other nodes after being written.
1397 The presence of the problem can be detected by looking at
1398 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1399 smaller than the second (nr_unused) then dirty files will not
1400 be flushed automatically to disk. "sync; sleep 10" should be
1401 run several times on the node before unmounting it to update
1402 Lustre (this is also safe to run on nodes without this problem).
1404 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1405 that can cause this same problem, so customers using that kernel
1406 also need to update the kernel in addition to Lustre. In order
1407 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1409 It is normal that files written just before a client crash (less
1410 than 30s) may not yet have been flushed to disk, even for local
1414 Frequency : frequent on thin XT3 nodes
1416 Description: UUID collision on thin XT3 Linux nodes
1417 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1418 randomly, since we relied on an insufficiently-seeded PRNG.
1423 Description: OSS hangs after "All ost request buffers busy"
1424 Details : A deadlock between quota and journal operations caused OSS
1425 hangs after printing "All ost request buffers busy."
1428 Frequency : always on liblustre builds
1430 Description: Cleanup compiler warnings on liblustre
1433 Frequency : always on liblustre builds on XT3
1435 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1439 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1440 > 32000 subdirectories
1441 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1442 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1443 subdirectory count crosses 32000. This will aid e2fsck to
1444 correctly handle more than 32000 subdirectories.
1447 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1449 Description: System hang on startup
1450 Details : This bug allowed the liblustre (e.g. catamount) client to
1451 return to the app before handling all startup RPCs. This
1452 could leave the node unresponsive to lustre network traffic
1453 and manifested as a server ptllnd timeout.
1455 Severity : enhancement
1457 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1458 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1459 boolean to control whether to print peer debug info when a
1460 client's RPC times out.
1465 Description: make O_SYNC working on 2.6 kernels
1466 Details : 2.6 kernels use different method for mark pages for write,
1467 so need add a code to lustre for O_SYNC work.
1472 Description: Failure to close file and release space on NFS
1473 Details : Put inode details into lock acquired in ll_intent_file_open.
1474 Use mdc_intent_lock in ll_intent_open to properly
1475 detect all kind of errors unhandled by mdc_enqueue.
1480 Description: proc file read during shutdown sometimes raced obd removal,
1482 Details : Add lock to prevent obd access after proc file removal.
1485 Frequency : Only for files larger than 4GB on 32-bit clients.
1487 Description: improperly doing page alignment of locks
1488 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1489 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1494 Description: RPCs being resent when they shouldn't be
1495 Details : Some RPCs that should not be resent are being resent. This
1496 can cause inconsistencies in the RPC state machine. Do not
1497 resend such requests.
1500 Frequency : rare, only with NFS export
1502 Description: Crash on NFS re-export node
1503 Details : under very unusual load conditions an assertion is hit in
1504 ll_intent_file_open()
1507 Frequency : only if OST filesystem is corrupted
1509 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1510 Details : for a short time RPCs with bulk IO are in the replay list,
1511 but replay of bulk IOs is unimplemented. If the OST filesystem
1512 is corrupted due to disk cache incoherency and then replay is
1513 started it is possible to trip an assertion. Avoid putting
1514 committed RPCs into the replay list at all to avoid this issue.
1519 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1520 Details : Large single O_DIRECT read and write calls can fail to allocate
1521 a sufficiently large buffer to process the request. In case of
1522 allocation failure the allocation is retried with a smaller
1523 buffer and broken into smaller requests.
1525 Severity : enhancement
1527 Description: Add -o localflock option to simulate old noflock behaviour.
1528 Details : This will achieve local-only flock/fcntl locks coherentness.
1533 Description: versioning check is incomplete
1534 Details : Checking the version difference of client vs. server, report
1535 error if the gap is too big.
1540 Description: add support PG_writeback bit
1541 Details : add support for PG_writeback bit for Lustre, for more carefull
1542 work with page cache in 2.6 kernel. This also fix some deadlocks
1543 and remove hack for work O_SYNC with 2.6 kernel.
1545 Severity : enhancement
1547 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1548 Details : The uninit_groups feature works in conjunction with the kernel
1549 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1550 up the pass1 processing of e2fsck. This is a read-only feature
1551 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1552 mount filesystems that have had this feature enabled.
1554 Severity : enhancement
1556 Description: Improve multi-block allocation algorithm to avoid fragmentation
1557 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1558 allocation locality and avoid filesystem fragmentation.
1560 ------------------------------------------------------------------------------
1562 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1564 * Support for kernels:
1565 2.6.9-42.0.3.EL (RHEL 4)
1566 2.6.5-7.276 (SLES 9)
1567 2.4.21-47.0.1.EL (RHEL 3)
1568 2.6.12.6 vanilla (kernel.org)
1569 2.6.16.21-0.8 (SLES10)
1570 * Recommended e2fsprogs version: 1.39.cfs2-0
1572 * The backwards-compatible /proc/sys/portals symlink has been removed
1573 in this release. Before upgrading, please ensure that you change
1574 any configuration scripts or /etc/sysctl.conf files that access
1575 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1576 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1577 in advance of the upgrade on any system running Lustre 1.4.6 or
1578 newer, since /proc/sys/lnet was added in that version.
1579 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1583 Frequency : only when quota is used
1585 Description: avoid scanning export list for quota master
1586 Details : Change the algorithms to avoid scanning export list in order
1587 to improve the efficiency.
1590 Frequency : MDS failover only, very rarely
1592 Description: "went back in time" messages on mds failover
1593 Details : The greatest transno may be lost when the current operation
1594 finishes with an error (transno==0) and the client's last_rcvd
1595 record is over-written. Save the greatest transno in the
1596 mds_last_transno for this case.
1599 Frequency : always for specific kernels and striping counts
1601 Description: client may get "Matching packet too big" without ACL support
1602 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1603 when trying to access files in certain configurations. The
1604 clients should in fact be denied when mounting because they do
1605 not understand ACLs.
1608 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1610 Description: many clients connecting with IO in progress causes connect timeouts
1611 Details : Avoid synchronous journal commits to avoid delays caused by many
1612 clients connecting/disconnecting when bulk IO is in progress.
1613 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1614 OST_IO_PORTAL to avoid delays behind potentially many pending
1618 Frequency : occasionally with multiple writers to a single file
1620 Description: shared writes to file may result in wrong size reported by stat()
1621 Details : Allow growing of kms when extent lock is cancelled
1624 Frequency : always with random mmap IO to multi-striped file
1626 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1627 Details : Only if the hole is at the end of OST object so that kms is too
1628 small. Fix is to increase kms accordingly in ll_nopage.
1631 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1633 Description: writes to a missing object would leak memory on the OST
1634 Details : If there is an inconsistency between the MDS and OST filesystems,
1635 such that the MDS references an object that doesn't exist, writes
1636 to that object will leak memory due to incorrect cleanup in the
1637 error handling path, eventually running out of memory on the OST.
1642 Description: Creating too long symlink causes lustre errors
1643 Details : Check symlink and name lengths before sending requests to MDS.
1646 Frequency : only if flock is enabled (not on by default)
1648 Description: posix locks not released on fd closure on 2.6.9+
1649 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1650 kernels, this caused such locks not to be released on fd close and
1651 then assertions on fs unmount about still used locks.
1654 Frequency : MDS failover only, very rarely
1656 Description: clients may get ASSERTION(granted_lock != NULL)
1657 Details : When request was taking a long time, and a client was resending
1658 a getattr by name lock request. The were multiple lock requests
1659 with the same client lock handle and
1660 mds_getattr_name->fixup_handle_for_resent_request found one of the
1661 lock handles but later failed with ASSERTION(granted_lock != NULL).
1666 Description: handle->h_buffer_credits > 0, assertion failure
1667 Details : h_buffer_credits is zero after truncate, causing assertion
1668 failure. This patch extends the transaction or creates a new
1672 Frequency : NFS re-export or patchless client
1673 Bugzilla : 11179, 10796
1674 Description: Crash on NFS re-export node (__d_move)
1675 Details : We do not want to hash the dentry if we don't have a lock.
1676 But if this dentry is later used in d_move, we'd hit uninitialised
1677 list head d_hash, so we just do this to init d_hash field but
1678 leave dentry unhashed.
1681 Frequency : NFS re-export or patchless client
1683 Description: NFS exports has problem with symbolic link
1684 Details : lustre client didn't properly install dentry when re-exported
1685 to NFS or running patchless client.
1688 Frequency : NFS re-export or patchless client
1690 Description: Various nfs/patchless fixes.
1691 Details : fixes reuse disconected alias for lookup process - this fixes
1692 warning "find_exported_dentry: npd != pd",
1693 fix permission error with open files at nfs.
1694 fix apply umask when do revalidate.
1697 Frequency : occasional
1699 Description: Crash on NFS re-export node
1700 Details : calling clear_page() on the wrong pointer triggered oops in
1701 generic_mapping_read().
1704 Frequency : rarely, using O_DIRECT IO
1706 Description: unaligned directio crashes client with LASSERT
1707 Details : check for unaligned buffers before trying any requests.
1710 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1712 Description: stale data returned from RAID cache
1713 Details : If only a small amount of IO is done to the RAID device before
1714 reading it again it is possible to get stale data from the RAID
1715 cache instead of reading it from disk.
1718 Frequency : always for sles10 kernel
1720 Description: sles10 support
1721 Details : ll_follow_link: compile fixes and using of nd_set_link
1722 under newer kernels.
1725 Frequency : depends on arch, kernel and compiler version, always on sles10
1728 Description: recursive or deep enough symlinks cause stack overflow
1729 Details : getting rid of large stack-allocated variable in
1733 Frequency : depends on hardware
1735 Description: lustre write performance loss in the SLES10 kernel
1736 Details : the performance loss is caused by using of write barriers in the
1737 ext3 code. The SLES10 kernel turns barrier support on by
1738 default. The fix is to undo that change for ldiskfs.
1740 ------------------------------------------------------------------------------
1742 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1744 * Support for kernels:
1745 2.6.9-42.0.3EL (RHEL 4)
1746 2.6.5-7.276 (SLES 9)
1747 2.4.21-47.0.1.EL (RHEL 3)
1748 2.6.12.6 vanilla (kernel.org)
1752 Frequency : quota enabled and large files being deleted
1754 Description: releasing more than 4GB of quota at once hangs OST
1755 Details : If a user deletes more than 4GB of files on a single OST it
1756 will cause the OST to spin in an infinite loop. Release
1757 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1762 Description: statfs data retrieved from /proc may be stale or zero
1763 Details : When reading per-device statfs data from /proc, in the
1764 {kbytes,files}_{total,free,avail} files, it may appear
1765 as zero or be out of date.
1768 Frequency : systems with MD RAID1 external journal devices
1770 Description: lconf's call to blkid is confused by RAID1 journal devices
1771 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1772 of returning all block devices that match the journal UUID.
1775 Frequency : always, for aggregate stripe size over 4GB
1777 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1778 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1779 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1782 Frequency : always if "lfs find" used on a local file/directory
1784 Description: "lfs find" segfaults if used on a local file/directory
1785 Details : The case where a directory component was not specified wasn't
1786 handled correctly. Handle this properly.
1789 Frequency : always on ppc64
1791 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1792 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1793 perform endianness conversion.
1796 Frequency : rarely (truncate to non-zero file size after write under load)
1797 Bugzilla : 10730, 10687
1798 Description: Files padded with zeros to next 4K multiple
1799 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1800 it is possible that files that are truncated to a non-zero size
1801 immediately after being written are filled with zero bytes beyond
1802 the truncated size. No file data is lost.
1804 Severity : enhancement
1806 Description: Allow recovery/failover for liblustre clients.
1807 Details : liblustre clients were unaware of failover configurations until
1810 Severity : enhancement
1812 Description: user file locks should fail when not mounting with flock option
1813 Details : Set up an error-returning stub in ll_file_operations.lock field
1814 to prevent incorrect behaviour when client is mounted without
1815 flock option. Also, set up properly f_op->flock field for
1819 Frequency : always on ia64
1821 Description: "lfs df" loops on printing out MDS statfs information
1822 Details : The obd_ioctl_data was not initialized and in some systems
1823 this caused a failure during the ioctl that did not return
1824 an error. Initialize the struct and return an error on failure.
1827 Frequency : SLES 9 only
1829 Description: Error of copying files with lustre special EAs as root
1830 Details : Client side always return success for setxattr call for lustre
1831 special xattr (currently only "trusted.lov").
1834 Frequency : rarely on clusters with both ia64+i386 clients
1836 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1837 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1838 clients are doing concurrent writes to the same file it is
1839 possible that the ia64 clients may LASSERT because the OST
1840 extent locks are not PAGE_SIZE aligned. Ensure that grown
1841 locks are always aligned on the request boundary.
1844 Frequency : specific use, occasional
1846 Description: Overwriting in use executable truncates on-disk binary image
1847 Details : If one node attempts to overwrite an executable in use by
1848 another node, we now correctly return ETXTBSY instead of
1849 truncating the file.
1851 Severity : enhancement
1853 Description: Async OSC create to avoid the blocking unnecessarily.
1854 Details : If a OST has no remain object, system will block on the creating
1855 when need to create a new object on this OST. Now, ways use
1856 pre-created objects when available, instead of blocking on an
1857 empty osc while others are not empty. If we must block, we block
1858 for the shortest possible period of time.
1863 Description: chmod on Lustre root is propagated to other clients
1864 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1865 invalid by the follow_mount time.
1868 Frequency : liblustre clients only
1870 Description: Race in 'instant cancel' lock handling could lead to such locks
1871 never to be granted in case of SMP MDS
1872 Details : Do not destroy not yet granted but cbpending locks in
1876 Frequency : replay/resend of open
1878 Description: non null lock assetion failure in mds_intent_policy
1879 Details : Trying to replay/resend lockless open requests resulted in
1880 mds_open() returning 0 with no lock. Now it sets a flag if
1881 a lock is going to be returned.
1883 Severity : enhancement
1885 Description: Checksum enhancements
1886 Details : New checksum enhancements allow for resending RPCs that failed
1889 Severity : enhancement
1891 Description: Tunables on number of dirty pages in cacche
1892 Details : Allow to set limit on number of dirty pages cached.
1897 Description: client crash on unmount - lock still has references
1898 Details : In some error handling cases it was possible to leak a lock
1899 reference on a client while accessing a file. This was not
1900 harmful to the client during operation, but would cause the
1901 client to crash when the filesystem is unmounted.
1904 Frequency : specific case, rare
1906 Description: ETXTBSY on mds though file not in use by client
1907 Details : ETXTBSY is no longer incorrectly returned when attempting to
1908 chmod or chown a directory that the user previously tried to
1909 execute or a currently-executing binary.
1912 Frequency : extremely rare except on liblustre-based clients
1914 Description: Lustre space not freed when files are deleted
1915 Details : Clean up open-unlinked files after client eviction. Previously
1916 the unlink was skipped and the files remained as orphans.
1921 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1922 Details : In some cases it was possible to send a blocking callback to
1923 a client doing a glimpse, even though that client didn't get
1924 a lock granted. When the glimpse lock is cancelled on the OST
1925 the freed lock is left on the waiting list and corrupted the list.
1928 Frequency : all core dumps
1930 Description: Broke core dumps to lustre
1931 Details : Negative dentry may be unhashed if parent does not have UPDATE
1932 lock, but some callers, e.g. do_coredump, expect dentry to be
1933 hashed after successful create, hash it in ll_create_it.
1935 ------------------------------------------------------------------------------
1937 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1939 * Support for kernels:
1940 2.6.9-42.0.2.EL (RHEL 4)
1941 2.6.5-7.276 (SLES 9)
1942 2.4.21-40.EL (RHEL 3)
1943 2.6.12.6 vanilla (kernel.org)
1947 Frequency : always on RHEL 3
1949 Description: Number of open files grows over time
1950 Details : The number of open files grows over time, whether or not
1951 Lustre is started. This was due to a filp leak introduced
1952 by one of our kernel patches.
1954 ------------------------------------------------------------------------------
1956 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1958 * Support for kernels:
1959 2.6.9-42.EL (RHEL 4)
1960 2.6.5-7.276 (SLES 9)
1961 2.4.21-40.EL (RHEL 3)
1962 2.6.12.6 vanilla (kernel.org)
1967 Bugzilla : 5719, 9635, 9792, 9684
1968 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1969 Details : If a server is under heavy load and cannot reply to new
1970 connection requests before the client resends the (re)connect,
1971 the connection handling code can behave badly if two service
1972 threads are concurrently handing separate (re)connections from
1973 the same client. Add better locking to the connection handling
1974 code, and ensure that only a single connection will be processed
1975 for a given client UUID, even if the lock is dropped.
1977 Severity : enhancement
1979 Description: add TCP zero-copy support to kernel
1980 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1981 sends if the hardware supports scatter-gather and checksumming.
1982 This allows socklnd to do client-write and server-read more
1983 efficiently and reduce CPU utilization from skbuf copying.
1986 Frequency : only if NFS exporting from client
1988 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1989 Details : If setting attributes on a file created under NFS that had
1990 never been opened it would be possible to oops the client
1991 if the file had no objects.
1994 Frequency : always for liblustre
1996 Description: liblustre client does MDS+OSTs setattr RPC for each write
1997 Details : When doing a write from a liblustre client, the client
1998 incorrectly issued an RPC to the MDS and each OST the file was
1999 striped over in order to update the timestamps. When writing
2000 with small chunks and many clients this could overwhelm the MDS
2001 with RPCs. In all cases it would slow down the write because
2002 these RPCs are unnecessary.
2004 Severity : enhancement
2006 Description: allow number of MDS service threads to be changed at module load
2007 Details : It is now possible to change the number of MDS service threads
2008 running. Adding "options mds mds_num_threads={N}" to the MDS's
2009 /etc/modprobe.conf will set the number of threads for the next
2010 time Lustre is restarted (assuming the "mds" module is also
2011 reloaded at that time). The default number of threads will
2012 stay the same, 32 for most systems.
2017 Description: OST crash if filesystem is unformatted or corrupt
2018 Details : If an OST is started on a device that has never been formatted
2019 or if the filesystem is corrupt and cannot even mount then the
2020 error handling cleanup routines would dereference a NULL pointer.
2025 Description: NULL pointer deref in llap_from_page.
2026 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2027 incorrect) mapping if the page was truncated/reclaimed while it was
2028 searched for. Check for this condition and skip such pages when
2029 doing readahead. Introduce extra check to llap_from_page() to
2030 verify page->mapping->host is non-NULL (so page is not anonymous).
2033 Frequency : Sometimes when using sys_sendfile
2035 Description: "page not covered by a lock" warnings from ll_readpage
2036 Details : sendfile called ll_readpage without right page locks present.
2037 Now we introduced ll_file_sendfile that does necessary locking
2038 around call to generic_file_sendfile() much like we do in
2042 Frequency : with certain MDS communication failures at client mount time
2044 Description: NULL pointer deref after failed client mount
2045 Details : a client connection request may delayed by the network layer
2046 and not be sent until after the PTLRPC layer has timed out the
2047 request. If the client fails the mount immediately it will try
2048 to clean up before the network times out the request. Add a
2049 reference from the request import to the obd device and delay
2050 the cleanup until the network drops the request.
2053 Frequency : occasionally during client (re)connect
2055 Description: assertion failure during client (re)connect
2056 Details : processing a client connection request may be delayed by the
2057 client or server longer than the client connect timeout. This
2058 causes the client to resend the connection request. If the
2059 original connection request is replied in this interval, the
2060 client may trip an assertion failure in ptlrpc_connect_interpret()
2061 which thought it would be the only running connect process.
2064 Frequency : only with obd_echo servers and clients that are rebooted
2066 Description: kernel BUG accessing uninitialized data structure
2067 Details : When running an obd_echo server it did not start the ping_evictor
2068 thread, and when a client was evicted an uninitialized data
2069 structure was accessed. Start the ping_evictor in the RPC
2070 service startup instead of the OBD startup.
2072 Severity : enhancement
2073 Bugzilla : 10193 (patchless)
2074 Description: Remove dependency on various unexported kernel interfaces.
2075 Details : No longer need reparent_to_init, exit_mm, exit_files,
2076 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2079 Frequency : rare (only users of deprecated and unsupported LDAP config)
2081 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2082 Details : LDAP apparently contains 'lustreName' attributes instead of
2083 'name'. A simple remapping of the name is sufficient.
2086 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2088 Description: waiting_locks_callback trips kernel BUG if client is evicted
2089 Details : Running with the dump_on_timeout debug flag turned on makes
2090 it possible that the waiting_locks_callback() can try to dump
2091 the Lustre kernel debug logs from an interrupt handler. Defer
2092 this log dumping to the expired_lock_main() thread.
2094 Severity : enhancement
2096 Description: Support NFS exporting on 2.6 kernels.
2097 Details : Implement non-rawops metadata methods for NFS server to use without
2098 changing NFS server code.
2101 Frequency : very rare (synthetic metadata workload only)
2103 Description: two racing renames might cause an MDS thread to deadlock
2104 Details : Running the "racer" program may cause one MDS thread to rename
2105 a file from being the source of a rename to being the target of
2106 a rename at exactly the same time that another thread is doing
2107 so, and the second thread has already enqueued these locks after
2108 doing a lookup of the target and is trying to relock them in
2109 order. Ensure that we don't try to re-lock the same resource.
2112 Frequency : only very large systems with liblustre clients
2114 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2115 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2116 all OSTs in parallel. This allows the eviction of stale liblustre
2117 clients to proceed much faster than if they were done in series,
2118 and also offers similar improvements for other set_info RPCs.
2123 Description: excessive CPU usage during initial read phase on client
2124 Details : During the initial read phase on a client, it would agressively
2125 retry readahead on the file, consuming too much CPU and impacting
2126 performance (since 1.4.5.8). Improve the readahead algorithm
2127 to avoid this, and also improve some other common cases (read
2128 of small files in particular, where "small" is files smaller than
2129 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2134 Description: MDS crash when receiving packet with unknown intent.
2135 Details : Do not LBUG in unknown intent case, just return -EFAULT
2137 Severity : enhancement
2138 Bugzilla : 9293, 9385
2139 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2140 Details : Do not serialize getattr (non-intent version) and statfs.
2143 Frequency : occasional, when OST network is overloaded/intermittent
2145 Description: client evicted by OST after bulk IO timeout
2146 Details : If a client sends a bulk IO request (read or write) the OST
2147 may evict the client if it is unresposive to its data GET/PUT
2148 request. This is incorrect if the network is overloaded (takes
2149 too long to transfer the RPC data) or dropped the OST GET/PUT
2150 request. There is no need to evict the client at all, since
2151 the pinger and/or lock callbacks will handle this, and the
2152 client can restart the bulk request.
2155 Frequency : Always when mmapping file with no objects
2157 Description: client crashes when mmapping file with no objects
2158 Details : Check that we actually have objects in a file before doing any
2159 operations on objects in ll_vm_open, ll_vm_close and
2165 Description: Request leak when working with deleted CWD
2166 Details : Introduce advanced request refcount tracking for requests
2167 referenced from lustre intent.
2169 Severity : Enhancement
2171 Description: Cache open file handles on client.
2172 Details : MDS now will return special lock along with openhandle, if
2173 requested and client is allowed to hold openhandle, even if unused,
2174 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2175 closing files for every read/write openration.
2177 Severity : Enhancement
2179 Description: Cache open negative dentries on client when possible.
2180 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2181 negative dentries on lock revocation.
2186 Description: Remounting a client read-only wasn't possible with a zconf mount
2187 Details : It wasn't possible to remount a client read-only with llmount.
2189 Severity : enhancement
2190 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2191 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2192 MPICH is included to add Lustre-specific ADIO interfaces.
2193 This is based closely on the UFS ADIO layer and only differs
2194 in file creation, in order to allow the OST striping to be set.
2195 This is user-contributed code and not supported by CFS.
2200 Description: extended inode attributes (immutable, append-only) work improperly
2201 when 2.4 and 2.6 kernels are used on client/server or vice versa
2202 Details : Introduce kernel-independent values for these flags.
2204 Severity : enhancement
2207 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2208 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2209 granularity. Now, Fractional MB granularity is be supported,
2210 this is very useful for low memory system.
2212 Severity : enhancement
2214 Description: Getattr by fid
2215 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2216 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2217 have locks and avoids getattr rpc requests for it on every stat.
2220 Frequency : Always, for filesystems larger than 2TB
2222 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2223 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2224 filesystems larger than 2TB to be mounted with mballoc enabled.
2227 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2229 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2230 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2231 filesystems larger than 2TB is mounted then the kernel will
2232 silently corrupt the start of the filesystem. CONFIG_LBD is
2233 enabled for all CFS-supported kernels, but the possibility of
2234 this happening with a modified kernel config exists.
2236 Severity : enhancement
2238 Description: add client O_DIRECT support for 2.6 kernels
2239 Details : It is now possible to do O_DIRECT reads and writes to files
2240 in the Lustre client mountpoint on 2.6 kernel clients.
2242 Severity : enhancement
2244 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2245 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2246 parallel, not waiting for response from every OST before sending
2247 a rpc to the next OST.
2252 Description: setattr vs write race when updating file timestamps
2253 Details : Client processes that update a file timestamp into the past
2254 right after writing to the file (e.g. tar) it is possible that
2255 the updated file modification time can be reset to the current
2256 time due to a race between processing the setattr and write RPC.
2258 Severity : enhancement
2260 Description: Bring 'lfs find' closer in line with regular Linux find.
2261 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2262 -print0 options and obtains all the needed info through the lustre
2265 Severity : enhancement
2267 Description: support up to 1024 configured devices on one node
2268 Details : change obd_dev array from statically allocated to dynamically
2269 allocated structs as they are first used to reduce memory usage
2274 Description: Flush dirty partially truncated pages during truncate
2275 Details : Immediatelly flush partially truncated pages in filter_setattr,
2276 this way we completely avoid having any pages in page cache on OST
2277 and can retire ugly workarounds during writes to flush such pages.
2282 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2283 Details : protect inode from truncation within vfs_unlink() context
2284 just take a reference before calling vfs_unlink() and release it
2285 when parent's i_sem is free.
2290 Description: last_id value checked outside lock on OST caused LASSERT failure
2291 Details : If there were multiple MDS->OST object precreate requests in
2292 flight, it was possible that the OST's last object id was checked
2293 outside a lock and incorrectly tripped an assertion. Move checks
2294 inside locks, and discard old precreate requests.
2297 Frequency : always, if extents are used on OSTs
2299 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2300 Details : OSTs using the extents format would not zero the high 16 bits of
2301 the index physical block number. This is not a problem for any
2302 OST filesystems smaller than 16TB, and no kernels support ext3
2303 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2304 new/modified files) and can be fixed for existing filesystems
2305 with e2fsprogs-1.39-cfs1.
2310 Description: import connection selection may be incorrect if timer wraps
2311 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2312 import connections to be ignored if the 32-bit jiffies counter
2313 wraps. Use a 64-bit jiffies counter.
2316 Frequency : very large clusters immediately after boot
2318 Description: LNET request buffers exhausted under heavy short-term load
2319 Details : If a large number of client requests are generated on a service
2320 that has previously never seen so many requests it is possible
2321 that the request buffer growth cannot keep up with the spike in
2322 demand. Instead of dropping incoming requests, they are held in
2323 the LND until the RPC service can accept more requests.
2326 Frequency : Sometimes during replay
2328 Description: Assertion failure in ll_local_open after replay.
2329 Details : If replay happened on an open request reply before we were able
2330 to set replay handler, reply will become not swabbed tripping the
2331 assertion in ll_local_open. Now we set the handler right after
2332 recognising of open request
2335 Frequency : very rare
2337 Description: kernel reports "badness in vsnprintf"
2338 Details : Reading from the "recovery_status" /proc file in small chunks
2339 may cause a negative length in lprocfs_obd_rd_recovery_status()
2340 call to vsnprintf() (which is otherwise harmless). Exit early
2341 if there is no more space in the output buffer.
2343 Severity : enhancement
2345 Description: clear OBD RPC statistics by writing to them
2346 Details : It is now possible to clear the OBD RPC statistics by writing
2347 to the "stats" file.
2352 Description: Client mtime is not the same on different clients after utimes
2353 Details : In some cases, the client was using the utimes() syscall on
2354 a file cached on another node. The clients now validate the
2355 ctime from the MDS + OSTs to determine which one is right.
2360 Description: Inability to activate failout mode
2361 Details : lconf script incorrectly assumed that in pythong string's numeric
2362 value is used in comparisons.
2365 Frequency : always with multiple stripes per file
2367 Description: Inefficient object allocation for mutli-stripe files
2368 Details : When selecting which OSTs to stripe files over, for files with
2369 a stripe count that divides evenly into the number of OSTs,
2370 the MDS is always picking the same starting OST for each file.
2371 Return the OST selection heuristic to the original design.
2376 Description: mount failures may take full timeout to return an error
2377 Details : Under some heavy load conditions it is possible that a
2378 failed mount can wait for the full obd_timeout interval,
2379 possibly several minutes, before reporting an error.
2380 Instead return an error as soon as the status is known.
2382 Frequency : quota enabled and large files being deleted
2384 Description: releasing more than 4GB of quota at once hangs OST
2385 Details : If a user deletes more than 4GB of files on a single OST it
2386 will cause the OST to spin in an infinite loop. Release
2387 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2392 Description: statfs data retrieved from /proc may be stale or zero
2393 Details : When reading per-device statfs data from /proc, in the
2394 {kbytes,files}_{total,free,avail} files, it may appear
2395 as zero or be out of date.
2398 Frequency : systems with MD RAID1 external journal devices
2400 Description: lconf's call to blkid is confused by RAID1 journal devices
2401 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2402 of returning all block devices that match the journal UUID.
2405 Frequency : always, for aggregate stripe size over 4GB
2407 Description: assertion fails when trying to use 4GB stripe size
2408 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2409 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2412 Frequency : always on ppc64
2414 Description: the first write on an ext3 filesystem with mballoc got stuck
2415 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2416 perform endianness conversion.
2418 ------------------------------------------------------------------------------
2420 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2422 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2423 INTEROPERATE with older versions automatically. Please read the
2424 user documentation before upgrading any part of a live system.
2425 * WARNING: Lustre networking configuration changes are required with
2426 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2429 * Support for kernels:
2430 2.6.9-22.0.2.EL (RHEL 4)
2431 2.6.5-7.244 (SLES 9)
2432 2.6.12.6 vanilla (kernel.org)
2435 Severity : enhancement
2436 Bugzilla : 7981/8208
2437 Description: Introduced Lustre Networking (LNET)
2438 Details : LNET is new networking infrastructure for Lustre, it includes
2439 a reorganized network configuration mode (see the user
2440 documentation for full details) as well as support for routing
2441 between different network fabrics. Lustre Networking Devices
2442 (LNDs) for the supported network fabrics have also been
2443 created for this new infrastructure.
2445 Severity : enhancement
2446 Description: Introduced Access control lists
2447 Details : clients can set ACLs on files and directories in order to have
2448 more fine-grained permissions than the standard Unix UGO+RWX.
2449 The MDS must be started with the "-o acl" mount option.
2451 Severity : enhancement
2452 Description: Introduced filesystem quotas
2453 Details : Administrators may now establish per-user quotas on the
2456 Severity : enhancement
2458 Description: Configuration change for the XT3
2459 The PTLLND is now used to run Lustre over Portals on the XT3
2460 The configure option(s) --with-cray-portals are no longer used.
2461 Rather --with-portals=<path-to-portals-includes> is used to
2462 enable building on the XT3. In addition to enable XT3 specific
2463 features the option --enable-cray-xt3 must be used.
2468 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2469 Details : A race between allocating a new llcd and re-getting the llcd_lock
2470 allowed another thread to grab newly-allocated llcd.
2472 Severity : enhancement
2474 Description: 2.6 OST async journal commit and locking fix to improve performance
2475 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2476 2.6 kernels are now async as they already were in 2.4 kernels so
2477 that they can commit concurrently with the network bulk transfer.
2478 For block-allocated files the filter allocation semaphore is held
2479 to avoid filesystem fragmentation during allocation. BKL lock
2480 removed for 2.6 xattr operations where it is no longer needed.
2485 Description: lconf incorrectly determined whether two IP networks could talk
2486 Details : In some more complicated routing and multiple-network
2487 configurations, lconf will avoid trying to make a network
2488 connection to a disjoint part of the IP space. It was doing the
2489 math incorrectly for one set of cases.
2494 Description: Fix for potential infinite loop processing records in an llog.
2495 Details : If an llog record is corrupted/zeroed, it is possible to loop
2496 forever in llog_process(). Validate the llog record length
2497 and skip the remainder of the block on error.
2500 Frequency : occasional (liblustre only)
2502 Description: liblustre could not open files whose last component is a symlink
2503 Details : sysio_path_walk() would incorrectly pass the open intent to
2504 intermediate path components.
2507 Frequency : rare (liblustre only with non-standard tuning)
2508 Bugzilla : 7201 (7350)
2509 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2510 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2511 after it has been released, only do this for referenced locks
2513 Severity : enhancement
2515 Description: specifying an (invalid) directory default stripe_size of -1
2516 would reset the directory default striping
2517 Details : stripe_size -1 was used internally to signal directory stripe
2518 removal, now use "all default" to signal dir stripe removal
2519 as a directory striping of "all default" is not useful
2522 Frequency : common for large clusters running liblustre clients
2524 Description: doing an ls when liblustre clients are running is slow
2525 Details : sending a glimpse AST to a liblustre client waits for every AST
2526 to time out, as liblustre clients will not respond. Since they
2527 cannot cache data we refresh the OST lock LVB from disk instead.
2529 Severity : enhancement
2531 Description: doing an ls at the same time as file IO can be slow
2532 Details : enqueue and other "small" requests can be blocked behind many
2533 large IO requests. Create a new OST IO portal for non-IO
2534 requests so they can be processed faster.
2537 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2539 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2540 created file with mode 000
2541 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2542 allow an ftruncate() as a non-root user to a file with mode 000.
2543 Lustre now respects this flag to disable mode checks when
2544 truncating a file owned by the user
2547 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2550 Description: Revoking locks from clients that went dead or catatonic might take
2552 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2553 cancellation of such locks instant on servers without waiting for
2554 any reply from clients. Clients drops these locks when cancel
2555 notification from server is received without replying.
2558 Frequency : liblustre-only, when liblustre client dies or becomes busy
2560 Description: Doing ls on Linux clients can take a long time with active
2562 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2563 granting such locks to it in the first place if possible. Locks
2564 are taken by proxy on the OST during the read or write and
2565 dropped immediately afterward. Add connect flags handling, do
2566 not grant locks to liblustre clients for glimpse ASTs.
2568 Severity : enhancement
2570 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2571 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2572 sequential reads and behaves badly with applications doing random
2573 reads. Improve it by reading ahead at least read region, and
2574 avoiding excessive large RPC for small reads.
2576 Severity : enhancement
2578 Description: Creating more than 1000 files for a single job may cause a load
2579 imbalance on the OSTs if there are also a large number of OSTs.
2580 Details : qos_prep_create() uses an OST index reseed value that is an
2581 even multiple of the number of available OSTs so that if the
2582 reseed happens in the middle of the object allocation it will
2583 still utilize the OSTs as uniformly as possible.
2588 Description: OST or MDS may oops in ping_evictor_main()
2589 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2590 but doesn't restart at beginning of obd_exports_timed list
2593 Severity : enhancement
2595 Description: improve by-nid export eviction on the MDS and OST
2596 Details : allow multiple exports with the same NID to be evicted at one
2597 time without re-searching the exports list.
2600 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2602 Description: MDS may oops in groups_free()
2603 Details : in rare race conditions a newly allocated group_info struct is
2604 freed again, and this can be NULL. The 2.4 compatibility code
2605 for groups_free() checked for a NULL pointer, but 2.6 did not.
2608 Frequency : common for liblustre clients doing little filesystem IO
2609 Bugzilla : 9352, 7313
2610 Description: server may evict liblustre clients accessing contended locks
2611 Details : if a client is granted a lock or receives a completion AST
2612 with a blocking AST already set it would not reply to the AST
2613 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2616 Frequency : lfs setstripe, only systems with more than 160 OSTs
2618 Description: unable to set striping with a starting offset beyond OST 160
2619 Details : llapi_create_file() incorrectly limited the starting stripe
2620 index to the maximum single-file stripe count.
2623 Frequency : LDAP users only
2625 Description: lconf did not handle in-kernel recovery with LDAP properly
2626 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2628 Severity : enhancement
2630 Description: bind OST threads to NUMA nodes to improve performance
2631 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2632 node and do their allocations there to localize memory access
2634 Severity : enhancement
2636 Description: llmount can determine client NID directly from Myrinet (GM)
2637 Details : the client NID code from gmnalnid was moved directly into
2638 llmount, removing the need to use this or specifying the
2639 client NID explicitly when mounting GM clients with zeroconf
2642 Frequency : if client is started with down MDS
2644 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2645 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2646 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2647 ptlrpc_import_delayed_req() because we didn't check if the
2648 request was interrupted, and we also didn't break out of the
2649 event loop if there was no timeout
2654 Description: data loss during non-page-aligned writes to a single file from
2655 both multiple nodes and multiple threads on one node at same time
2656 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2657 inconsistency led to false short-reads, that were cached and later
2658 used by ->prepare_write() to fill in partially written page,
2659 leading to data loss.
2662 Frequency : always, if lconf --abort_recovery used
2664 Description: lconf --abort_recovery fails with 'Operation not supported'
2665 Details : lconf was attempting to abort recovery on the MDT device and not
2668 Severity : enhancement
2670 Description: remove cleanup logs
2671 Details : replace lconf-generated cleanup logs with lustre internal
2672 cleanup routines. Eliminates the need for client-cleanup and
2675 Severity : enhancement
2677 Description: add support for EAs (user and system) on lustre filesystems
2678 Details : it is now possible to store extended attributes in the Lustre
2679 client filesystem, and with the user_xattr mount option it
2680 is possible to allow users to store EAs on their files also
2682 Severity : enhancement
2684 Description: Add possibility (config option) to show minimal available OST free
2686 Details : When compiled with --enable-mindf configure option, statfs(2)
2687 (and so, df) will return least minimal free space available from
2688 all OSTs as amount of free space on FS, instead of summary of
2689 free spaces of all OSTs.
2691 Severity : enhancement
2693 Description: do not expand extent locks acquired on OST-side
2694 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2695 by server: they are not cached anyway.
2698 Frequency : when mmap is used/binaries executed from Lustre
2700 Description: Unmmap pages before throwing them away from read cache.
2701 Details : llap_shrink cache now attempts to unmap pages before discarding
2702 them (if unmapping failed - do not discard). SLES9 kernel has
2703 extra checks that trigger if this unmapping is not done first.
2708 Description: lconf didn't resolve symlinks before checking to see whether a
2709 given mountpoint was already in use
2712 Frequency : when migrating failover services
2713 Bugzilla : 6395, 9514
2714 Description: When migrating a subset of services from a node (e.g. failback
2715 from a failover service node) the remaining services would
2716 time out and evict clients.
2717 Details : lconf --force (implied by --failover) sets the global obd_timeout
2718 to 5 seconds in order to quickly disconnect, but this caused
2719 other RPCs to time out too quickly. Do not change the global
2720 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2722 Severity : enhancement
2723 Frequency : if MDS is started with down OST
2724 Bugzilla : 9439,5706
2725 Description: Allow startup/shutdown of an MDS without depending on the
2726 availability of the OSTs.
2727 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2728 Add appropriate locking and lov-osc refcounts for safe
2729 cleaning. Add osc abort_inflight calls in case the
2730 synchronize never started.
2733 Frequency : occasional (Cray XT3 only)
2735 Description: root not authorized to access files in CRAY_PORTALS environment
2736 Details : The client process capabilities were not honoured on the MDS in
2737 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2738 been accessed by an authorized user then root was able to access
2739 the file on the local client also. The root user capabilities
2740 are now allowed on the MDS, as this environment has secure UID.
2743 Frequency : occasional
2745 Description: ldiskfs "too long searching" message happens too often
2746 Details : A debugging message (otherwise harmless) prints too often on
2747 the OST console. This has been reduced to only happen when
2748 there are fragmentation problems on the filesystem.
2753 Description: Division by zero in statfs when all OSCs are inactive
2754 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2755 lov_statfs divides by value returned by lov_get_stripecnt().
2759 Bugzilla : 9489, 3273
2760 Description: First write from each client to each OST was only 4kB in size,
2761 to initialize client writeback cache, which caused sub-optimal
2762 RPCs and poor layout on disk for the first writen file.
2763 Details : Clients now request an initial cache grant at (re)connect time
2764 and so that they can start streaming writes to the cache right
2765 away and always do full-sized RPCs if there is enough data.
2766 If the OST is rebooted the client also re-establishes its grant
2767 so that client cached writes will be honoured under the grant.
2772 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2773 Details : Now I/O RPCs go to different portal number and (presumably) fast
2774 lock requests (and glimses) and other RPCs get their own service
2775 threads pool that should be able to service those RPCs
2778 Severity : enhancement
2780 Description: Ability to exchange lustre version between client and servers and
2781 issue warnings at client side if client is too old. Also for
2782 liblustre clients there is ability to refuse connection of too old
2784 Details : New 'version' field is added to connect data structure that is
2785 filled with version info. That info is later checked by server and
2789 Frequency : rare, liblustre only.
2790 Bugzilla : 9296, 9581
2791 Description: Two simultaneous writes from liblustre at offset within same page
2792 might proceed at the same time overwriting eachother with stale
2794 Details : I/O lock withing llu_file_prwv was released too early, before data
2795 actually was hitting the wire. Extended lock-holding time until
2796 server acknowledges receiving data.
2799 Frequency : extremely rare. Never observed in practice.
2801 Description: avoid generating lustre_handle cookie of 0.
2802 Details : class_handle_hash() generates handle cookies by incrementing
2803 global counter, and can hit 0 occasionaly (this is unlikely, but
2804 not impossible, because initial value of cookie counter is
2805 selected randonly). Value of 0 is used as a sentinel meaning
2806 "unassigned handle" --- avoid it. Also coalesce two critical
2807 sections in this function into one.
2809 Severity : enhancement
2811 Description: allow liblustre clients to delegate truncate locking to OST
2812 Details : To avoid overhead of locking, liblustre client instructs OST to
2813 take extent lock in ost_punch() on client's behalf. New connection
2814 flag is added to handle backward compatibility.
2816 Severity : enhancement
2817 Bugzilla : 4928, 7341, 9758
2818 Description: allow number of OST service threads to be specified
2819 Details : a module parameter allows the number of OST service threads
2820 to be specified via "options ost ost_num_threads={N}" in the
2821 OSS's /etc/modules.conf or /etc/modprobe.conf.
2825 Bugzilla : 6146, 9635, 9895
2826 Description: servers crash with bad pointer in target_handle_connect()
2827 Details : In rare cases when a client is reconnecting it was possible that
2828 the connection request was the last reference for that export.
2829 We would temporarily drop the export reference and get a new
2830 one, but this may have been the last reference and the export
2831 was just destroyed. Get new reference before dropping old one.
2833 Severity : enhancement
2834 Frequency : if client is started with failover MDS
2836 Description: Allow multiple MDS hostnames in the mount command
2837 Details : Try to read the configuration from all specified MDS
2838 hostnames during a client mount in case the "primary"
2841 Severity : enhancement
2843 Description: Stop sending data to evicted clients as soon as possible.
2844 Details : Check if the client we are about to send or are sending data to
2845 was evicted already. (Check is done every second of waiting,
2846 for which l_wait_event interface was extended to allow checking
2847 of exit condition at specified intervals).
2850 Frequency : rare, normally only when NFS exporting is done from client
2852 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2854 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2857 Frequency : always, if srand() is called before liblustre initialization
2859 Description: Liblustre uses system PRNG disturbing its usage by user application
2860 Details : Introduce internal to lustre fast and high-quality PRNG for
2861 lustre usage and make liblustre and some other places in generic
2862 lustre code to use it.
2864 Severity : enhancement
2865 Bugzilla : 9477, 9557, 9870
2866 Description: Verify that the MDS configuration logs are updated when xml is
2867 Details : Check if the .xml configuration logs are newer than the config
2868 logs stored on the MDS and report an error if this is the case.
2869 Request --write-conf, or allow starting with --old_conf.
2871 Severity : enhancement
2873 Description: Handle symlinks in the path when checking if Lustre is mounted.
2874 Details : Resolve intermediate symlinks when checking if a client has
2875 mounted a filesystem to avoid duplicate client mounts.
2880 Description: lconf can hit an error exception but still return success.
2881 Details : The lconf command catches the Command error exception at the top
2882 level script context and will exit with the associated exit
2883 status, but doesn't ensure that this exit status is non-zero.
2888 Description: failure of ptlrpc thread startup can cause oops
2889 Details : Starting a ptlrpc service thread can fail if there are a large
2890 number of threads or the server memory is very fragmented.
2891 Handle this without oopsing.
2894 Frequency : always, only if liblustre and non-default acceptor port was used
2896 Description: liblustre cannot connect to servers with non-default acceptor port
2897 Details : tcpnal_set_default_params() was not called and was therefore
2898 ignoring the environment varaible TCPNAL_PORT, as well as other
2899 TCPNAL_ environment variables
2904 Description: two objects could be created on the same OST for a single file
2905 Details : If an OST is down, in some cases it was possible to create two
2906 objects on a single OST for a single file. No problems other
2907 than potential performance impact and spurious error messages.
2911 Bugzilla : 5681, 9562
2912 Description: Client may oops in ll_unhash_aliases
2913 Details : Client dcache may become inconsistent in race condition.
2914 In some cases "getcwd" can fail if the current directory is
2920 Description: Inode refcounting problems in NFS export code
2921 Details : link_raw functions used to call d_instantiate without obtaining
2922 extra inode reference first.
2926 Bugzilla : 9942, 9903
2927 Description: Referencing freed requests leading to crash, memleaks with NFS.
2928 Details : We used to require that call to ll_revalidate_it was always
2929 followed by ll_lookup_it. Also with revalidate_special() it is
2930 possible to call ll_revalidate_it() twice for the same dentry
2931 even if first occurence returned success. This fix changes semantic
2932 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2933 reference on a request referred from the intent.
2934 ll_intent_release() then releases such a request.
2937 Frequency : rare, normally benchmark loads only
2939 Description: unlinked inodes were kept in memory on the client
2940 Details : If a client is repeatedly creating and unlinking files it
2941 can accumulate a lot of stale inodes in the inode slab cache.
2942 If there is no other client load running this can cause the
2943 client node to run out of memory. Instead flush old inodes
2944 from client cache that have the same inode number as a new inode.
2947 Frequency : SLES9 2.6.5 kernel and long filenames only
2948 Bugzilla : 9969, 10379
2949 Description: utime reports stale NFS file handle
2950 Details : SLES9 uses out-of-dentry names in some cases, which confused
2951 the lustre dentry revalidation. Change it to always use the
2955 Frequency : rare, unless heavy write-truncate concurrency is continuous
2956 Bugzilla : 4180, 6984, 7171, 9963, 9331
2957 Description: OST becomes very slow and/or deadlocked during object unlink
2958 Details : filter_destroy() was holding onto the parent directory lock
2959 while truncating+unlinking objects. For very large objects this
2960 may block other threads for a long time and slow overall OST
2961 responsiveness. It may also be possible to get a lock ordering
2962 deadlock in this case, or run out of journal credits because of
2963 the combined truncate+unlink. Solution is to do object truncate
2964 first in one transaction without parent lock, and then do the
2965 final unlink in a new transaction with the parent lock. This
2966 reduces the lock hold time dramatically.
2969 Frequency : rare, 2.4 kernels only
2971 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2972 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2973 before unmounting their filesystem in order to avoid deadlock.
2974 The kernel_locked() function in 2.4 kernels only checks whether
2975 the kernel lock is held, not whether it is this process that is
2976 holding it as 2.6 kernels do.
2981 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2982 Details : The client ptlrpc code may be trying to reconnect to a down
2983 server before a previous connection attempt has timed out.
2984 Increase the reconnect interval to be longer than the connection
2985 timeout interval to avoid sending duplicate connections to
2989 Frequency : echo_client brw_test command
2991 Description: fix echo_client to work with OST preallocated code
2992 Details : OST preallocation code (5137) didn't take echo_client IO path
2993 into account: echo_client calls filter methods outside of any
2994 OST thread and, hence, there is no per-thread preallocated
2995 pages and buffers to use. Solution: hijack pga pages for IO. As
2996 a byproduct, this avoids unnecessary data copying.
3000 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3001 Description: Client can oops in mdc_commit_close() after open replay
3002 Details : It was possible for the MDS to return an open request with no
3003 transaction number in mds_finish_transno() if the client was
3004 evicted, but without actually returning an error. Clients
3005 would later try to replay that open and may trip an assertion
3006 Simplify the client close codepath, and always return an error
3007 from the MDS in case the open is not successful.
3010 Frequency : rare, 2.6 OSTs only
3012 Description: OST may deadlock under high load on fragmented files
3013 Details : If there was a heavy load and highly-fragmented OST filesystems
3014 it was possible to have all the OST threads deadlock waiting on
3015 allocation of biovecs, because the biovecs were not released
3016 until the entire RPC IO was completed. Instead, release biovecs
3017 as soon as they are complete to ensure forward IO progress.
3019 Severity : enhancement
3021 Description: Support for specifying external journal device at mount
3022 Details : If an OST or MDS device is formatted with an external journal
3023 device, this device major/minor is stored in the ext3 superblock
3024 and may not be valid for failover. Allow detecting and
3025 specifying the external journal at mount time.
3030 Description: Mounting an MDS with pending unlinked files may cause oops
3031 Details : target_finish_recovery() calls mds_postrecov() which returned
3032 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3033 considers this an error and immediately begins cleaning up the
3034 lov, just after starting the mds_lov process
3036 Severity : enhancement
3038 Description: Implement 'lfs df' to report actual free space on per-OST basis
3039 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3040 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3041 usage of objects; '-h' to report in human readable format.
3043 ------------------------------------------------------------------------------
3045 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
3052 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3053 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3054 at startup during orphan unlinking. Instead, skip buddy update
3055 before setup, as it will be regenerated after recovery is complete.
3060 Description: If an OST is inactive, its locks might reference stale inodes.
3061 Details : lov_change_cbdata() must iterate over all namespaces, even if
3062 they are inactive to clear inode references from the lock.
3064 Severity : enhancement
3065 Frequency : occasional, if non-standard max_dirty_mb used
3067 Description: Client will block write RPCs if not enough grant
3068 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3069 then the client will block writes while waiting for another RPC
3070 to complete instead of consuming its dirty limit. With change
3071 we get improved performance when max_dirty_mb is small.
3073 Severity : enhancement
3074 Bugzilla : 3389, 6253
3075 Description: Add support for supplementary groups on the MDS.
3076 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3077 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3078 lookups for user supplementary groups into a cache.
3082 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3083 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3084 for liblustre O_EXCL is enforced by MDS.
3088 Description: readdir never returns NULL in liblustre.
3089 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3090 dentry in struct dirent.
3094 Frequency : liblustre only, depends on application IO pattern
3095 Description: liblustre clients evicted if not contacting servers
3096 Details : Don't put liblustre clients into the ping_evictor list, so
3097 they will not be evicted by the pinger ever.
3099 Severity : enhancement
3101 Description: Add ability to evict clients by NID from MDS.
3102 Details : By echoing "nid:$NID" string into
3103 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3104 $NID would be instantly evicted from this MDS and from all active
3105 OSTs connected to it.
3109 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3110 Details : lookup_it_finish() used to query file size from OSTs that was not
3115 Description: service threads change working directory to that of init
3116 Details : Starting lustre service threads may pin the working directory
3117 of the parent thread, making that filesystem busy. Threads
3118 now change to the working directory of init to avoid this.
3122 Frequency : during shutdown only
3123 Description: shutdown with a failed MDS or OST can cause unmount to hang
3124 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3129 Frequency : 2.6 only
3130 Description: chmod/chown may include an extra supplementary group
3131 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3132 supplementary group and if none is specified this is used.
3135 Bugzilla : 5479 (6816)
3137 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3138 Details : If lookup is in progress on a file that is unlinked we might try
3139 to revalidate the inode and fail in revalidate after lookup is
3140 complete and ll_file_open() enqueues the open again but
3141 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3144 Frequency : always, if lconf --abort_recovery used
3146 Description: lconf --abort_recovery fails with 'Operation not supported'
3147 Details : lconf was attempting to abort recovery on the MDT device and not
3150 ------------------------------------------------------------------------------
3152 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3157 Frequency : rare (only unsupported configurations with a node running as an
3159 Bugzilla : 6514, 5137
3160 Description: Mounting a Lustre file system on a node running as an OST could
3162 Details : OSTs now preallocates memory needed to write out data at
3163 startup, instead of when needed, to avoid having to
3164 allocate memory in possibly low memory situations.
3165 Specifically, if the file system is mounted on on OST,
3166 memory pressure could force it to try to write out data,
3167 which it needed to allocate memory to do. Due to the low
3168 memory, it would be unable to do so and the node would
3169 become unresponsive.
3171 Severity : enhancement
3173 Description: Addition of lconf --service command line option
3174 Details : lconf now accepts a '--service <arg>' option, which is
3175 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3177 Severity : enhancement
3179 Description: Failover mode is now the default for OSTs.
3180 Details : By default, OSTs will now run in failover mode. To return to
3181 the old behaviour, add '--failout' to the lmc line for OSTs.
3183 Severity : enhancement
3185 Description: Health checks are now provided for MDS and OSTs
3186 Details : Additional detailed health check information on MSD and OSTs
3187 is now provided through the procfs health_check value.
3190 Frequency : occasional, depends on IO load
3192 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3193 after numerous create/delete cycles
3194 Details : The ext3 inode allocation policy would not allocate new inodes
3195 very well on the OSTs because there are no new directories
3196 being created. Instead we look for groups with free space if
3197 the parent directories are nearly full.
3202 Description: Network or server problems during mount may cause partially
3203 mounted clients instead of returning an error.
3204 Details : The config llog parsing code may overwrite the error return
3205 code during mount error handling, returning success instead
3211 Description: MDS can fail to allocate large reply buffers
3212 Details : After long uptimes the MDS can fail to allocate large reply
3213 buffers (e.g. zconf client mount config records) due to memory
3214 fragmentation or consumption by the buffer cache. Preallocate
3215 some large reply buffers so that these replies can be sent even
3216 under memory pressure.
3220 Frequency : rare (liblustre)
3221 Description: fsx running with liblustre complained that using truncate() to
3222 extend the file doesn't work. This patch corrects that issue.
3223 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3224 ATTR_SIZE and lsm use in llu_setattr_raw.
3228 Frequency : rare, only 2.6 kernels
3229 Description: Unusual file access patterns on the MDS may result in inode
3230 data being lost in very rare circumstances.
3231 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3232 code in the 2.6 kernel caused the inode and/or EA data not to
3233 be read from disk, causing single-file corruption.
3237 Frequency : rare, only 2.6 filesystems using extents
3238 Description: Heavy concurrent write and delete load may cause data corruption.
3239 Details : It was possible under high-load situations to have an extent
3240 metadata block in the block device cache from a just-unlinked
3241 file overwrite a newly-allocated data block. We now unmap any
3242 metadata buffers that alias just-allocated data blocks.
3246 Frequency : filesystems with default stripe_count larger than 77
3247 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3248 Details : lconf specifies an inode size of 4096 bytes when the default
3249 stripe_count is larger than 77. This conflicts with the default
3250 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3251 this case to avoid pinning too much memory for large EAs.
3253 ------------------------------------------------------------------------------
3255 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3260 Frequency : rare (extremely heavy IO load with hundreds of clients)
3262 Description: Client is evicted, gets IO error writing to file
3263 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3264 caused clients to be evicted for AST timeouts. The fixes in
3265 bug 5192 mean we no longer need to have such short AST timeouts
3266 so ldlm_timeout has been increased.
3269 Frequency : occasional during --force or --failover shutdown under load
3270 Bugzilla : 5949, 4834
3271 Description: Server oops/LBUG if stopped with --force or --failover under load
3272 Details : a collection of import/export refcount and cleanup ordering
3273 issues fixed for safer force cleanup
3276 Frequency : only filesystems larger than 120 OSTs
3277 Bugzilla : 5990, 6223
3278 Description: lfs getstripe would oops on a very large filesystem
3279 Details : lov_getconfig used kfree on vmalloc'd memory
3282 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3283 Bugzilla : 6242, 6243
3284 Description: reading from files that had been truncated to a non-zero size
3285 but never opened returned no data
3286 Details : ll_file_read() reads zeros from no-object files to EOF
3291 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3292 Details : There's little harm in aborting MDS/OSS recovery and letting it
3293 try again, so I removed the LASSERT and return an error instead.
3295 Severity : enhancement
3297 Description: New debugging infrastructure for tracking down data corruption
3298 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3299 (b) cover more of the client-side code path, and (c) try to narrow
3300 down where problems occurred
3304 Bugzilla : 3819, 4364, 4397, 6313
3305 Description: Racing close and eviction MDS could cause assertion in mds_close
3306 Details : It was possible to get multiple mfd references during close and
3307 client eviction, leading to one thread referencing a freed mfd.
3309 Severity: : enhancement
3310 Bugzilla : 3262, 6359
3311 Description: Attempts to reconnect to servers are now more aggressive.
3312 Details : This builds on the enhanced upcall-less recovery that was added
3313 in 1.4.2. When trying to reconnect to servers, clients will
3314 now try each server in the failover group every 10 seconds. By
3315 default, clients would previously try one server every 25 seconds.
3320 Description: After recovery, certain operations trigger a failed
3321 assertion on a client.
3322 Details : Failing over an mds, using lconf -d --failover, while a
3323 client was doing a readdir() call would cause the client to
3324 LBUG after recovery completed and the readdir() was resent.
3326 Severity : enhancement
3328 Description: Default groups are now added by lconf
3329 Details : You can now run lconf --group <servicename> without having to
3330 manually add groups with lmc.
3333 Frequency : occasional
3335 Description: Nodes with an elan id of 0 trigger a failed assertion
3338 Frequency : always when accessing e.g. tty/console device nodes
3340 Description: tty and some other devices nodes cannot be used on lustre
3341 Details : file's private_data field is used by device data and lustre
3342 values in there got lost. New field was added to struct file to
3343 store fs-specific private data.
3346 Frequency : when exporting Lustre via NFS
3348 Description: NFSD failed occasionally when looking up a path component
3349 Details : NFSD is looking up ".." which was broken in ext3 directories
3350 that had grown large enough to become hashed.
3353 Frequency : Clusters with multiple interfaces not on the same subnet
3355 Description: Nodes will repeatedly try to reconnect to an interface which it
3356 cannot reach and report an error to the log.
3357 Details : Extra peer list entries will be created by lconf with some peers
3358 unreachable. lconf now validates the peer before adding it.
3361 Frequency : Only if a default stripe is set on the filesystem root.
3363 Description: Setting a default stripe on the filesystem root prevented the
3364 filesystem from being remounted.
3365 Details : The client was sending extra request flags in the root getattr
3366 request and did not allocate a reply buffer for the dir EA.
3369 Frequency : occasional, higher if lots of files are accessed by one client
3370 Bugzilla : 6159, 6097
3371 Description: Client trips assertion regarding lsm mismatch/magic
3372 Details : While revalidating inodes the VFS looks up inodes with ifind()
3373 and in rare cases can find an inode that is being freed.
3374 The ll_test_inode() code will free the lsm during ifind()
3375 when it finds an existing inode and then the VFS later attaches
3376 this free lsm to a new inode.
3380 Bugzilla : 6422, 7030
3381 Description: MDS deadlock between mkdir and client eviction
3382 Details : Creating a new file via mkdir or mknod (starting a transaction
3383 and getting the ns lock) can deadlock with client eviction
3384 (gets ns lock and trying to finish a synchronous transaction).
3387 Frequency : occasional
3388 Description: While starting a server, the fsfilt_ext3 module could not be
3390 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3391 kernels. Previously, lconf would still use the ext3 name
3392 when trying to load modules. Now, it will correctly use
3393 ext3 on 2.4 and ldiskfs on 2.6.
3395 Severity : enhancement
3396 Description: The default stripe count has been changed to 1
3397 Details : The interpretation of the default stripe count (0, to lfs
3398 or lmc) has been changed to mean striping across a single
3399 OST, rather than all available. For general usage we have
3400 found a stripe count of 1 or 2 works best.
3402 Severity : enhancement
3403 Description: Add support for compiling against Cray portals.
3404 Details : Conditional compiling for some areas that are different
3408 Frequency : occasional
3409 Bugzilla : 6409, 6834
3410 Description: Creating files with an explicit stripe count may lead to
3411 a failed assertion on the MDS
3412 Details : If some OSTs are full or unavailable, creating files may
3413 trigger a failed assertion on the MDS. Now, Lustre will
3414 try to use other servers or return an error to the
3418 Frequency : occasional
3420 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3421 caused lock timeout during readahead (since 1.4.2).
3422 Details : Processes doing readahead might match a lock that hasn't been
3423 granted yet if there are overlapping and conflicting lock
3424 requests. The readahead process waits on ungranted lock
3425 (original lock is CBPENDING), while OST waits for that process
3426 to cancel CBPENDING read lock and eventually evicts client.
3428 Severity : enhancement
3430 Description: Initial enabling of flock support for clients
3431 Details : Implements fcntl advisory locking and file status functions.
3432 This feature is provided as an optional mount flag (default
3433 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3434 locking are implemented yet, and those that are are not guaranteed
3435 to be completely correct in production environments.
3436 mount -t lustre -o [flock|noflock] ...
3439 Frequency : occasional
3441 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3442 assertion in the ext3 JBD (journaling) layer.
3443 Details : The b_committed_data struct is protected by the big kernel lock
3444 in 2.4 kernels, serializing journal_commit_transaction() and
3445 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3446 access to this struct. In 2.6 kernels there is finer grained
3447 locking to improve SMP performance of the JBD layer.
3451 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3456 Description: merge and cleanup kernel patches.
3459 -----------------------------------------------------------------------------
3461 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3463 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3464 versions of Lustre. Please update all servers and clients to
3465 version 1.4.2 or later at the same time. You must also run
3466 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3467 to update the configuration logs.
3469 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3470 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3471 - swab llog records of type '0' so we get proper header size/idx (5861)
3472 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3473 - fix rename of one directory over another leaking an inode (5953)
3474 - avoid SetPageDirty on 2.6 (5981)
3475 - don't re-add just-being-destroyed locks to the waiting list (5653)
3476 - when creating new directories, inherit the parent's custom
3477 striping settings if present parent (3048)
3478 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3479 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3480 - don't hold client locks on temporary worklist from l_lru (5666)
3481 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3482 - automatically evict dead clients (5921)
3483 - Update file size properly in create+truncate+fstat case (6196)
3484 - Do not unhash mountpoint dentries, do not allow removal of
3486 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3487 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3488 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3489 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3490 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3491 - Correctly handle reads of files with no objects (6243)
3492 - lctl recover will also mark a device active if deactivate used (5933)
3494 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3495 - allow --write-conf on an MDS with different nettype than client (5619)
3496 - don't write config llogs to MDS for mounts not from that MDS (5617)
3497 - lconf should create multiple TCP connections from a client (5201)
3498 - init scripts are now turned off by default; run chkconfig --on
3499 lustre and chkconfig --on lustrefs to use them
3500 - upcalls are no longer needed for clients to recover to failover
3502 - add --abort-recovery option to lconf to abort recovery on device
3504 - add support for an arbitrary number of OSTs (3026)
3505 - Quota support protocol changes.
3506 - forward compatibility changes to wire structs (6007)
3507 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3508 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3509 - improved "lustre status" script
3510 - initialize blocksize for non-regular files (6062)
3511 - added --disable-server and --disable-client configure options (5782)
3512 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3513 - Vanilla 2.4.29 support
3514 - increase maximum number of obd devices to 520 (6242)
3515 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3516 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3517 SLES 9 2.6.5 kernels
3518 - limit stripes per file to 160 (the maximum EA size) (6093)
3520 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3523 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3524 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3525 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3526 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3527 - small ext3 extents cleanups and fixes (5733)
3528 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3529 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3530 - avoid needless client->OST connect, fix handle mismatch (5317)
3531 - fix DLM error path that led to out-of-sync client, long delays (5779)
3532 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3533 - fix several locking issues related to i_size (5492,5624,5654,5672)
3534 - don't move pending lock onto export if it is already evicted (5683)
3535 - fix kernel oops when creating .foo in unlinked directory (5548)
3536 - fix deadlock in obdfilter statistics vs. object create (5811)
3537 - use time_{before,after} to avoid timer jiffies wrap (5882)
3538 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3539 - Do not leak request if server was not able to process it (5154)
3540 - If mds_open unable to find parent dir, make that negative lookup(5154)
3541 - don't create new directories with extent-mapping (5909, 5936)
3543 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3544 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3545 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3546 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3547 - support for RHEL4 kernel on i686 (5773)
3548 - provide error messages when incompatible logs are encountered (5898)
3550 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3551 * version 1.4.0.10 (1.4.1 release candidate 1)
3553 - don't keep a lock reference when lock is not granted (4238)
3554 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3555 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3556 - drop import inflight refcount on signal_completed_replay error (5255)
3557 - unlock page after async write error during send (3677)
3558 - handle missing objects in filter_preprw_read properly (5265)
3559 - no transno return for symlink open, don't save no-trasno open (3440)
3560 - don't try to complete elan receive that already failed (4012)
3561 - free RPC server reply state on error (5406)
3562 - clean up thread from ptlrpc_start_thread() on error (5160)
3563 - readahead could read extra page into cache that wasn't ejected (5388)
3564 - prevent races in class_attach/setup/cleanup/detach (5260)
3565 - don't dereference de->d_inode after l_dput of de (5458)
3566 - use "int" for stripe value returned from lock_to_stripe (5544)
3567 - mballoc allocation and error-checking fixes in 2.6 (5504)
3568 - block device patches to fix I/O request sizes in 2.6 (5482)
3569 - look up hostnames for IB nals (5602)
3570 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3571 - don't start multiple acceptors for the same port (5277)
3572 - fix incorrect LASSERT in mds_getattr_name (5635)
3573 - export a proc file for general "ping" checking (5628)
3574 - fix "lfs check" to not block when the MDS is down (5628)
3576 - service request history (4965)
3577 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3578 - create an "evict_client" /proc entry on OSTs, like the MDS has
3579 - fix mount usage message, return errors per mount(8) (5168)
3580 - change grep [] to grep "[]" in tests so they work in more UMLs
3581 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3582 - remove ancient LOV_MAGIC_V0 EA support (5047)
3583 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3584 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3585 - allow readahead allocations to fail when low on memory (5383)
3586 - mmap locking landed again, after considerable improvement (2828)
3587 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3589 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3592 - send OST transaction number in read/write reply to free req (4966)
3593 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3594 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3595 - the watchdog thread now runs as interruptible (5246)
3596 - flock/lockf fixes (but it's still disabled, pending 5135)
3597 - don't use EXT3 constants in llite code (5094)
3598 - memory shortage at startup could cause assertion (5176)
3600 - reorganization of lov code
3601 - single portals codebase
3603 - add extents/mballoc support (5025)
3604 - direct I/O reads in the obdfilter (4048)
3605 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3607 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3610 - send OST transaction number in read/write reply to free req (4966)
3611 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3612 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3613 - the watchdog thread now runs as interruptible (5246)
3614 - handle missing objects in filter_preprw_read properly (5265)
3615 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3616 - drop import inflight refcount on signal_completed_replay error (5255)
3617 - unlock page after async write error during send (3677)
3618 - return original error code on reconstructed replies (3761)
3619 - no transno return for symlink open, don't save no-trasno open (3440)
3621 - add pid to ldlm debugging output (4922)
3622 - bump the watchdog timeouts -- we can't handle 30sec yet
3623 - extra debugging for orphan dentry/inode bug (5259)
3625 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3628 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3629 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3630 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3631 - fix NULL dereference in /proc/sys/portals/routes (4827)
3632 - allow failed mdc_close() operations to be interrupted (4561)
3633 - stop precreate on OST before MDS would time out on it (4778)
3634 - don't send partial-page writes before EOF from client (4410)
3635 - discard client grant for sub-page writes on large-page clients (4520)
3636 - don't free dentries not owned by NFS code, check generation (4806)
3637 - fix lsm leak if mds_create_objects() fails (4801)
3638 - limit debug_daemon file size, always print CERROR messages (4789)
3639 - use transno after validating reply (3892)
3640 - process timed out requests if import state changes (3754)
3641 - update mtime on OST during writes, return in glimpse (4829)
3642 - add mkfsoptions to LDAP (4679)
3643 - use ->max_readahead method instead of zapping global ra (5039)
3644 - don't interrupt __l_wait_event() during strace
3646 - add software watchdogs to catch hung threads quickly (4941)
3647 - make lustrefs init script start after nfs is mounted
3648 - fix CWARN/ERROR duplication (4930)
3649 - return async write errors to application if possible (2248)
3650 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3651 - print NAL number in %x format (4645)
3652 - update barely-supported suse-2.4.21-171 series (4842)
3653 - support for sles 9 %post scripts
3654 - support for building 2.6 kernel-source packages
3655 - support for sles km_* packages
3657 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3660 - ignore -ENOENT errors in osc_destroy (3639)
3661 - notify osc create thread that OSC is being cleaned up (4600)
3662 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3663 - reconstruct ost_handle() like mds_handle() (4657)
3664 - create a new thread to do import eviction to avoid deadlock (3969)
3665 - let lconf resolve symlinked-to devices (4629)
3666 - don't unlink "objects" from directory with default EA (4554)
3667 - hold socknal file ref over connect in case target is down (4394)
3668 - allow more than 32000 subdirectories in a single directory (3244)
3669 - fix blocks count for O_DIRECT writes (3751)
3670 - OST returns ENOSPC from object create when no space left (4539)
3671 - don't send truncate RPC if file size isn't changing (4410)
3672 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3673 - bind to privileged port in socknal and tcpnal (3689)
3675 - rate limit CERROR/CWARN console message to avoid overload (4519)
3676 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3677 - basic mmap support (3918)
3678 - kernel patch series update from b1_4 (4711)
3680 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3683 - avoid crash during MDS cleanup with OST shut down (2775)
3684 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3685 - don't use bad inodes on the MDS (3744)
3686 - dynamic object preallocation to improve recovery speed (4236)
3687 - don't hold spinlock over lock dumping or change debug flags (4401)
3688 - don't zero obd_dev when it is force cleaned (3651)
3689 - print grants to console if they go negative (4431)
3690 - "lctl deactivate" will stop automatic recovery attempts (3406)
3691 - look for existing locks in ldlm_handle_enqueue() (3764)
3692 - don't resolve lock handle twice in recovery avoiding race (4401)
3693 - revalidate should check working dir is a directory (4134)
3695 - don't always mark "slow" obdfilter messages as errors (4418)
3697 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3700 - don't close LustreDB during write_conf until it is done (3860)
3701 - fix typo in lconf for_each_profile (3821)
3702 - allow dumping logs from multiple threads at one time (3820)
3703 - don't allow multiple threads in OSC recovery (3812)
3704 - fix debug_size parameters (3864)
3705 - fix mds_postrecov to initialize import for llog ctxt (3121)
3706 - replace config semaphore with spinlock (3306)
3707 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3708 - don't allow enqueue to complete on a destroyed export (3822)
3709 - down write_lock before checking llog header bitmap (3825)
3710 - recover from lock replay timeout (3764)
3711 - up llog sem before sending rpc (3652)
3712 - reduce ns lock hold times when setting kms (3267)
3713 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3714 - fix NULL deref and obd_dev leak on setup error (3312)
3715 - replace some LBUG about llog ops with error handling (3841)
3716 - don't match INVALID dentries from d_lookup and spin (3784)
3717 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3718 - fix invalid assertion in ptlrpc_set_wait (3880)
3720 - add libwrap support for the TCP acceptor (3996)
3721 - add /proc/sys/portals/routes for non-root route listing (3994)
3722 - allow setting MDS UUID in .xml (2580)
3723 - print the stack of a process that LBUGs (4228)
3725 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3728 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3729 - make sure to unset replay flag from failed open requests (3440)
3730 - if default stripe count is 0, use OST count for inode size (3636)
3731 - update parent mtime/ctime on client for create/unlink (2611)
3732 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3733 - free recovery state on server during a forced cleanup (3571)
3734 - unregister_reply for resent reqs (3063)
3735 - loop back devices mounting and status check on 2.6 (3563)
3736 - fix resource-creation race that can provoke i_size == 0 (3513)
3737 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3738 - more debugging for page-accounting assertion (3746)
3739 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3740 - avoid deadlock after precreation failure (3758)
3741 - fix race and lock order deadlock in orphan handling (3450, 3750)
3742 - add validity checks when grabbing inodes from l_ast_data (3599)
3744 - add /proc/.../recovery_status to obdfilter (3428)
3745 - lightweight CDEBUG infrastructure, debug daemon (3668)
3746 - change default OSC RPC parameters to be better on small clusters
3747 - turn off OST read cache for files smaller than 32MB
3748 - install man pages and include them in rpms (3100)
3749 - add new init script for (un)mounting lustre filesystems (2593)
3750 - run chkconfig in %post for init scripts (3701)
3751 - drop scimac NAL (unmaintained)
3753 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3756 - clean kiobufs before and after use (3485)
3757 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3758 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3759 - change init script to fail more clearly if not run as root (1528)
3760 - allow clients to reconnect during replay (1742)
3761 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3762 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3763 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3764 - fix cond_resched (3554)
3765 - only evict unfinished clients after recovery (3515)
3766 - allow bulk resend, prevent data loss (3570)
3767 - dynamic ptlrpc request buffer allocation (2102)
3768 - don't allow unlinking open directory if it isn't empty (2904)
3769 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3770 - remove extraneous obd dereference causing LASSERT failure (3334)
3771 - don't use get_cycles() when creating temp. files on the mds (3156)
3772 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3773 - handle EEXIST for set-stripe, set proper directory name (3336)
3775 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3776 - fix ksocknal_fmb_callback() error messages (2918)
3778 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3781 - don't copy lvb into (possibly NULL) reply on error (2983)
3782 - don't deref dentry after dput, don't free lvb on error (2922)
3783 - use the kms to determine writeback rpc length (2947)
3784 - increment oti_logcookies when osc is inactive (2948)
3785 - update client's i_blocks count via lvb messages (2543)
3786 - handle intent open/close of special files properly (1557)
3787 - mount MDS with errors=remount-ro, like obdfilter (2009)
3788 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3789 - don't use cancelling-locks' kms values (2947)
3790 - use highest lock extent for kms, not last one (2925)
3791 - don't dereference ERR_PTR() dentry in error handling path (3107)
3792 - fix thread race in portals_debug_dumplog() (3122)
3793 - create lprocfs device entries at setup instead of at attach (1519)
3794 - common AST error handler, don't evict client on completion race (3145)
3795 - zero nameidata in detach_mnt in 2.6 (3118)
3796 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3797 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3798 - initialize RPC timeout timer earlier for 2.6 (3219)
3799 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3800 - print nal/nid for unknown nid (3258)
3801 - additional checks for oscc recovery before doing precreate (3284)
3802 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3803 - don't crash in mdc_close for bad permissions on open (3285)
3804 - zero i_rdev for non-device files (3147)
3805 - clear page->private before handing to FS, better assertion (3119)
3806 - tune the read pipeline (3236)
3807 - fix incorrect decref of invalidated dentry (2350)
3808 - provide read-ahead stats and refine rpc in flight stats (3328)
3809 - don't hold journal transaction open across create RPC (3313)
3810 - update atime on MDS at close time (3265)
3811 - close LDAP connection when recovering to avoid server load (3315)
3812 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3813 - don't leak open file on MDS after open resend (3325)
3814 - serialize filter_precreate and filter_destroy_precreated (3329)
3815 - loop device shouldn't call sync_dev() for nul device (3092)
3816 - clear page cache after eviction (2766)
3817 - resynchronize MDS->OST in background (2824)
3818 - refuse to mount the same filesystem twice on same mountpoint (3394)
3819 - allow llmount to create routes for mounting behind routers (3320)
3820 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3821 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3822 - fix rare problem with rename on htree directories (3417)
3824 - allow default OST striping configuration per directory (1414)
3825 - fix compilation for qswnal for 2.6 kernels (3125)
3826 - increase maximum number of MDS request buffers for large systems
3827 - change liblustreapi to be useful for external progs like lfsck (3098)
3828 - increase local configuration timeout for slow disks (3353)
3829 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3831 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3834 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3835 - don't overwrite extent policy data in reply if lock was blocked (2901)
3836 - drop filter export grants atomically with removal from device (2663)
3837 - del obd_self_export from work_list in class_disconnect_exports (2908)
3838 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3839 - swab reply message in mdc_close, other PPC fixes (2464)
3840 - fix destroying of named logs (2325)
3841 - overwrite old logs when running lconf --write_conf (2264)
3842 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3843 - fix race in target_handle_connect (2898)
3844 - mds_reint_create() should take same inode create lock (2926)
3845 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3846 - don't close files for self_export to avoid uninitialized obd (2936)
3847 - allow MDS with the same name as client node (2939)
3848 - hold dentry reference for closed log files for unlink (2325)
3849 - reserve space for all logs during transactions (2059)
3850 - don't evict page beyond end of stripe extent (2925)
3851 - don't oops on a deleted current working directory (2399)
3852 - handle hard links to targets without a parent properly (2517)
3853 - don't dereference NULL lock when racing during eviction (2867)
3854 - don't grow lock extents when lots of conflicting locks (2919)
3856 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3859 - account for cache space usage on clients to avoid data loss (974)
3860 - lfsck support in lustre kernel code (2349)
3861 - reduce journal credits needed for BRW writes (2370)
3862 - orphan handling to avoid losing space on client/server crashes
3863 - ptlrpcd can be blocked, stopping ALL progress (2477)
3864 - use lock value blocks to assist in proper KMS, faster stat (1021)
3865 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3866 - recovery for initial connections (2355)
3867 - fixes for mds_cleanup_orphans (1934)
3868 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3869 - block all file creations until orphan recovery completes (1901)
3870 - client remove rq_connection from request struct (2423)
3871 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3872 - recovery timer race (2670)
3873 - mdc_close recovey bug (2532)
3874 - ptlrpc cleanup bug (2710)
3875 - mds timeout on local locks (2588)
3876 - namespace lock held during RPCs (2431)
3877 - handle interrupted sync write properly (2503)
3878 - don't try to handle a message that hasn't been replied to (2699)
3879 - client assert failure during cleanup after abort recovery (2701)
3880 - leak mdc device after failed mount (2712)
3881 - ptlrpc_check_set allows timedout requests to complete (2714)
3882 - wait for inflight reqs when ptlrpcd finishes (2710)
3883 - make sure unregistered services are removed from the srv_list
3884 - reset bulk XID's when resending them (caught by 1138 test)
3885 - unregister_bulk after timeout
3886 - fix lconf error (2694)
3887 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3888 - readahead locks pages, leaves pending causing memory pressure (2673)
3889 - increase OST request buffers to 4096 on large machines (2729)
3890 - fix up permission of existing directories in simple_mkdir (2661)
3891 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3892 - don't assign transno to errored transactions (2742)
3893 - don't delete objects on OST if given a bogus objid from MDS (2751)
3894 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3895 - if rq_no_resend, then timeout request after recovery (2432)
3896 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3897 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3898 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3899 - MDS recovery shouldn't delete objects that it has given out (2730)
3900 - if enqueue arrives after completion, don't clobber LVB (2819)
3901 - don't unlock pages twice when trigger_group_io returns error (2814)
3902 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3903 - don't write pages to disk if there was an error (1450)
3904 - don't ping imports that have recovery disabled (2676)
3905 - take buffered bytes into account when balancing socknal conn (2817)
3906 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3907 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3908 - remove little-endian swabbing of llog records (1987)
3909 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3910 - retry reposting request buffers if they fail (1191)
3911 - grow extent at grant time to avoid granting a revoked lock (2809)
3912 - lock revoke doesn't evict page if covered by a second lock (2765)
3913 - disable VM readahead to avoid reading outside lock extents (2805)
3915 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3916 - updated kernel patches for hp-2.4.20 kernel (2681)
3918 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3921 - fix truncated write corruption (2366)
3922 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3924 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3925 - ensure len(uuid) < 37 in lmc (1171)
3926 - fix ia64 OOPS in llog_test (2255)
3927 - zero end of page at obdfilter for partial page writes (2648)
3928 - don't leave stale dentries around after renames (bug 2428)
3929 - fix timeouts when evicting a client with a single lock held (2642)
3930 - set deadline for the initial HELLO message to drain (2634)
3931 - print out dotted-quad IP addresses in the socknal (2302)
3933 - additional debugging for MDS client eviction problem (2443)
3934 - fix mkfsoptions support for osts (2603, 2604)
3936 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3939 - add series for the vanilla 2.6.0 kernel
3940 - add series for the vanilla 2.4.24 kernel
3941 - add series for a cray x86/64 UL kernel drop
3942 - fix xattr patches for the vanilla 2.4.19 series
3944 - generate true UUIDs in lmc (1171)
3945 - have portals stack dumping break in UML (2466)
3946 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3947 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3948 - wake callback waiting threads on client eviction (2460)
3949 - Add --ptldebug and --subsystem to lmc (1719)
3950 - update assertion to allow safe interrupt allocation
3951 - set rq_no_resend for cancel requests (2432)
3952 - recalculate ptlrpcd timeout after resend (2494)
3953 - call vfs_rmdir when removing pending directories (2368)
3954 - fix renaming a file to itself (2429)
3955 - lmc creates a default one-stripe lov (2454)
3956 - expand procfs space to handle large clusters (2326)
3957 - increase UML stack to avoid overflow
3958 - update lconf's list of debug and subsystem masks
3959 - fix lfs find --obd (2510)
3960 - /proc tunable for disabling filter read caching (2591)
3961 - stop rpm packages from altering slapd.conf (2301)
3962 - disable nagle in the socknal under 0conf (2578)
3963 - choose mds inode size based on stripe count (2572)
3964 - fix kernel-source rpm problems (2516)
3966 - add --disable-doc to avoid pdf generation (2421)
3967 - update documentation, tests, type-os, comments
3968 - avoid format warnings on ia64
3969 - remove the TOE NAL
3970 - tiny code cleanups by removing unused fields
3972 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3975 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3976 - remove the most blatant lies from BUILDING file (2371)
3977 - change default debug level to reasonable production setting
3978 - reduce client side cache size to reduce cache flush time
3979 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3980 - make TCP zerocopy and pinger support enabled by default (2476)
3981 - sync writes completed after process exits caused crashes (2319)
3982 - maintain correct mount count on the MDS (2356)
3983 - backout 1557, because 2316 wasn't really fixed
3984 - better file I/O statistics gathering in /proc
3985 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3986 - another kernel patch to fix zero-copy TCP function export
3987 - don't take duplicate lock when processing re-sent getattr (2420)
3988 - lctl uses obd_self_export instead of creating new conn (2353)
3989 - MDS/OST recovery case which requires object creation asserted (2425)
3990 - move lfs from /usr/sbin to /usr/bin in packages
3991 - fix race between mds_client_add and mds_client_free (2417)
3992 - use kmalloc instead of slabs in portals (2430)
3993 - don't create duplicate records when a failover MDS is present (2442)
3994 - remove unnecessary mount age check (2332)
3995 - don't remove directory inodes from locks prematurely (2451)
3996 - don't break if MDS service name is the same as hostname (2103)
3997 - fix races in client write RPC generation when cache full (2482)
3999 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4002 - remove now-unused request->rq_obd (278)
4003 - if an allocation fails, print out how much memory we've used (1933)
4004 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4005 - add the "configurable stack size" patch to most series files (1256)
4006 - ability to write large log records, for 100+ OST configs (2306)
4007 - fix NULL deref when filter_prep fails (2314)
4008 - fix operator precedence error in filter_sync
4009 - dynamic allocation of socknal TX descriptors (2315)
4010 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4011 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4012 - ensure that CWARN messages in Portals always get to the syslog
4013 - __init/__exit are not for prototype decls (ldlm_init/exit)
4014 - x86-64 compile warning fixes
4015 - fix gateway LMC keyword conflict (2318)
4016 - fix MDS lock inversions in getattr/reint paths (1844)
4017 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4018 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4019 - DLM race condition prevented some lock evictions (2328)
4020 - ENOMEM detection and retry on socknal sends (2230)
4021 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4022 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4023 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4024 - better error messages when a client is rejected during recovery (1505)
4025 - avoid cancelling locks which were never granted, after failure (2330)
4026 - fix i_sem/journal inversion in mds_client_add (2333)
4027 - fix truncate/getattr lock cycle deadlock (2334)
4028 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4029 - fix two rare exit paths which could leak an l_lock() ref (2321)
4030 - fencepost error in MDS/OST orphan recovery (2226)
4031 - make log record alignment 8 bytes (1988)
4032 - lstripe now fails when requested offset > ost_count (2237)
4033 - ensure that all kernel series have a complete list.h (1607)
4034 - fix crashes in special-file operations (2316)
4035 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4037 - allow configurable automake binary, for testing new versions
4038 - small update to the lfs documentation
4040 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4042 * fix negative export reference count in fsfilt_sync (2312)
4044 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4045 * release candidate 0.9.1
4047 - orphans are moved into the PENDING directory for possible recovery
4048 - replayed opens now open by fid for orphan/rename safety (1042)
4049 - last close of an orphan inode generates a transno (683)
4050 - chdir() and mount() now pin the directory entry (1020)
4051 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4052 - discard very old requests without processing them (1502)
4053 - remove some common, well-understood CERRORs (1505)
4054 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4055 - clear "grant" flags in OST replies until OST grant code lands (1644)
4056 - fix read performance by not clobbering i_blksize on client (1598)
4057 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4058 - make LRU size a /proc tunable, clears locks when reduced (707)
4059 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4060 - server locks take references on exports now (1558)
4061 - build fixes for 2.4.20-rh trees (1663)
4062 - return an error from lov_create if all OSCs are inactive (1751)
4063 - fix import levels when a reconnect happens without a timeout (1597)
4064 - exit early from mds_open if we get a lookup error (1749)
4065 - partial page read at EOF wouldn't wait for disk before sending (1642)
4066 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4067 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4068 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4069 - write/truncate lock inversion (1639)
4070 - Don't auto-load obdclass, portals modules during cleanup (1495)
4071 - fix timestamps from jumping to "now" (1763)
4072 - extra journal assertions (1648)
4073 - add an extra multiunlink test (1771)
4074 - fix read_record/write_record API (1776)
4075 - fix leak of offset_extent, possible incorrect i_size later (1772)
4076 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4077 - Debugging for the kqswnal_get_idle_tx problems (1820)
4078 - Allow recovery to be attempted multiple times (1536)
4079 - Write out MDS last_rcvd file after it is first created (1600)
4080 - Fix tx_descriptor leak in failed transmit situations (1827)
4081 - ext3 journaling fixes for assertion failure after IO error (1871)
4082 - class_export_put() on freed export after completion AST error (1896)
4083 - Fix revalidate looping in VFS (1322)
4084 - Don't access a freed export during MDS_REINT timeout (1521)
4085 - Add open-unlink recovery support on the MDS (1673,1764)
4086 - Return an error if no MDS data was read from last_rcvd (1946)
4087 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4088 - Don't setup a disk device that doesn't match exported UUID (317)
4089 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4090 - avoid committing NULL handle in force close
4091 - local.sh is now a one-stripe LOV configuration
4092 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4093 - fix ext3 htree duplicate directory entry corruption (1516)
4094 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4095 - update to new LOV EA format (2097)
4096 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4098 2003-06-15 Phil Schwan <phil@clusterfs.com>
4101 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4102 - per-import/export recovery handling (958, 931, 959)
4103 - multiple last-rcvd slots, for serving multiple FSes (949)
4104 - connections are again shared between multiple imp/exports (963, 964)
4105 - "umount -f" would hang if any requests needed to be sent (393, 978)
4106 - avoid pinning large req buffer by copying for queued messages (989)
4107 - add "uuid" to "lctl device" command to help upcalls (991)
4108 - "open" RPCs with transnos would confuse recovery counters (1037)
4109 - do proper endian conversion of all wire messages (288, 340, 891)
4110 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4111 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4112 - fix ptlrpc_connection leak in target_handle_connect (1174)
4113 - fix import refcounting bug in OST and MDS cleanup (1134)
4114 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4115 - fix very unlikely obd_types race condition (501)
4116 - remove osc_open hack for echo_client (1187)
4117 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4118 - a failure in read_inode2 leads to deadlock (1139)
4119 - cancel ack-locks as soon as transaction is committed (1072)
4120 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4121 - make sure to commitrw after any preprw to avoid deadlock (1162)
4122 - failing to execute a file in a lustre FS would lock inode (1203)
4123 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4124 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4125 - obd_uuid2tgt would walk off the end of the list (1255)
4126 - on IA64 the osc would give portals incorrect bulk size (1258)
4127 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4128 - fix lock inversion caused by new llite matching code (1282)
4129 - limit the number of dirty pages on a client to 10MB (1286)
4130 - timed out locks were not being corrected cancelled (1289)
4131 - fix O_DIRECT above 4GB on IA-32 (1292)
4132 * major user-visible changes
4133 - fail out/fail over policy now controlled by the upcall (993)
4135 - add OBD_PING to check server availability and failure (954)
4136 - lustre messages are now sent in sending host order (288, 340, 891)
4137 - add eadatalen to MDS getattr reply (340)
4138 - OST read replies may contain second buffer, with per-page status (593)
4140 2003-03-11 Phil Schwan <phil@clusterfs.com>
4143 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4144 - failed open()s could cause deadlock; fixed (867, 869)
4145 - stop cancelling OST locks when files are closed (481)
4146 - overlapping XID spaces caused network corruption (851, 853)
4147 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4148 - setattr_raw added, to do single-RPC, server-side setattrs
4149 - lmc/lconf syntax change for OST UUIDs
4150 - fix crashy race condition between ptlrpc_free_req and osc_close
4151 - don't use request in mdc_enqueue if we hit a timeout (889)
4152 - don't set the inode i_size for regular files from the MDS (896)
4153 - handle out of order completion AST (842)
4154 - don't LBUG if a lock request times out after receiving AST (913)
4155 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4156 - if a bad lock AST arrives, send an error instead of dropping entirely
4157 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4158 - fix leak in bulk IO when only partially completed (899, 900, 926)
4159 - fix O_DIRECT for ia64 (55)
4160 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4161 - C-z after timeout could hang a process forever; fixed (977)
4163 - client-side I/O cache (678, 924, 929, 941, 970)
4165 - READPAGE and SETATTRs which don't take server-side locks get
4168 2003-02-11 Phil Schwan <phil@clusterfs.com>
4171 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4172 - Fix fsfilt_extN_readpage() to read a full page of directory
4173 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4174 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4175 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4176 - Most intent operations were removed, in favour of a new RPC mode
4177 that does a single RPC to the server and bypasses most of the VFS
4178 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4179 - Aggressively cancel local locks on DLM servers
4180 - mds_reint_unlink sends EA to the client if it's the last nlink.
4181 client uses that EA to unlink OST objects.
4182 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4183 - recursive symlinks were fixed (439)
4184 - fixed NULL deref in DEBUG_REQ
4185 - filter_update_lastobjid no longer calls sync, which annoyed extN
4186 - fixed multi-client small-writes to a single file problem (445)
4187 - fixed mtime updates during file writes (607)
4188 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4189 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4190 - fixed Linux OST size reporting problem (444, 656)
4191 - OST now updates object mtime with writes or setattr (607, 619)
4192 - client verifies file size before zeroing page past EOF (445)
4193 - OST now writes last allocated objid to disk with allocation (108)
4194 - LOV on echo now works (409)
4196 - mds_reint_unlink sends a new buffer, with the EA included. this
4197 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4198 set if a regular file was being unlinked, and it was the last link
4199 - use PtlGet from the target for bulk writes (315)
4200 - OST now updates object mtime with writes or setattr (607, 619)
4201 - LDLM now has a grant-time callback to revalidate locked items, if
4203 - Many MDS operations were reorganized to combat race conditions
4205 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4206 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4208 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4211 - Fully reactivate OST imports after reconnection (512, others)
4212 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4213 - More graceful error handling for truncating on dead OST (515)
4214 - Don't error out unless we're actually accessing dead stripes (474)
4215 - Fix garbage sizes when stripes are missing (410)
4216 - LRU counters were broken, causing constant lock purge (433, 432)
4217 - garbage on read from stripes with failed OSTs (441)
4218 - mark OSCs as active before reconnecting during recovery (438)
4219 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4220 - lfind did not preserve OST order in output (443)
4221 - symlinks cause hung clients, incorrect data (439)
4222 - stop dereferencing request after dropping refcount (457)
4223 - don't LASSERT(spin_is_locked) on non-SMP (455)
4224 - fixes for many rename() bugs
4225 - fstat didn't correctly synchronize attributes (399)
4226 - server must handle lock cancellation during blocking AST prep (487)
4227 - bulk descriptors were free()d too soon (511)
4228 - fix paths in lconf, which would load incorrect modules (451, 507)
4229 - fix confusing lconf 'host not found' error message (386)
4230 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4231 - fix race condition in mdc_blocking_ast() for inode access (526)
4232 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4233 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4234 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4235 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4236 - fix lstripe in several ways (kernel side) (527)
4237 - fix request leak in ldlm_cli_enqueue (262)
4238 - incorrect OSC was marked inactive after OST failure
4239 - call mds_fs_cleanup before unmounting filesystem (524)
4240 - fix races between taking ns_lock and ldlm_lock_change_resource
4241 - fix races updating LOV export open file list
4242 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4243 - fix recovery NULL deref in ldlm_cli_cancel_unused
4244 - fix some DLM races by using new hash table for lock handles (419)
4245 - permit the client to specify desired inodes, at replay
4246 - duplicate requests when we queue them for replay reintegration
4247 - fix last_rcvd offset calculation
4248 - sync after each recovered transaction, so we always make progress
4249 - never, not always, ERESTART requests without transnos
4250 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4252 - skip replay if the MDS says that the client is already connected
4253 - don't check for a recovery-enabled export to match lctl's UUID
4254 - don't INC_USE_COUNT for phantom exports
4255 - don't crash when cleaning up phantom exports (567)
4256 - don't double-finish or set replay data for errored mdc_open requests
4257 - abort requests when they time out, so we don't get old replies
4258 - send/receive replies for AST messages again
4259 - if the client says that it doesn't have the lock, cancel it on the
4261 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4262 mark it as destroyed, it will all work out when decref is called
4263 - fix module use counts (22, 581)
4265 - ASTs now expect a reply (server cancels lock on error reply)
4267 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4270 - fix many simultaneous client startup (392)
4271 - fix dentry->d_it clobbering
4272 - credentials weren't being shipped for readdir/getattr operations
4273 - remove invalid assertions triggered during some concurrent MD
4275 - proper Lustre versions added (336, 389)
4276 - fix memory leak for create error case (398)
4277 - fix LOV locking bug that would get cli/srv out of sync
4278 - fix echo client over LOV (409)
4279 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4280 - fix double-O_EXCL intent crash (424)
4281 - avoid sending multiple lock CANCELs (352)
4283 - MDS can do multi-client recovery (modulo bugs in new code)
4285 - many updates, edits, cleanups
4287 2002-11-18 Phil Schwan <phil@clusterfs.com>
4290 - fix null d_it dereference (346)
4291 - fix full OST/dbench hang (333)
4292 - fix permission problem with file removal (286)
4293 - fix removal of OSCs from LOV when they fail
4294 - fix NULL deref during bulk timeout (214)
4295 - fix problems related to multiple filesystems on one MDS (241)
4296 - fixed serious subtle metadata locking bugs
4297 - free locks on clients when inodes are removed due to memory
4299 - fix inode pointer in lock data (285)
4300 - partial support for multiple MDS on a single host (241)
4301 - data locks weren't cancelled at clear_inode time (290, 311)
4302 - intent locks could lead to unbounded lock growth (205)
4303 - added a maximum lock count, an LRU list, and a flusher
4304 - fix multiple rename (365)
4305 - properly abstracted the echo client
4306 - OSC locked 1 byte too many; fixed
4307 - rewrote brw callback code:
4308 - fixed recovery bugs related to LOVs (306)
4309 - fixed too-many-pages-in-one-write crash (191)
4310 - fixed (again) crash in sync_io_timeout (214)
4311 - probably fixed callback-related race (385)
4313 - Add capability to MDS protocol
4314 - LDLM cancellations and callbacks on different portals
4316 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4319 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4320 - defer open object destruction to close time (601981, 138)
4321 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4322 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4323 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4324 - don't call obd_setattr in ll_file_release for destroyed objects
4325 * protocol change to lustre_msg: move |version| and add |flags|
4326 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4327 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4328 * added replay of create, unlink, link and rename operations during
4329 MDS failover; recovery should be much more robust now
4330 * remove failed OSCs from LOVs (only lov_create uses this so far)
4331 * the lustre-HOWTO was brought (more) up to date (582544)
4333 2002-10-23 Phil Schwan <phil@clusterfs.com>
4336 - in-use dentries weren't being reused properly (617851)
4337 - prevent multiple LDLM setup (599178)
4338 - fix LOV size calculations for truncate (617853)
4339 - fix client handling of MDS intent errors (POSIX)
4340 - fix permission bug in lovstripe.c test (624321)
4341 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4342 - truncate past end of file could corrupt data
4343 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4344 - a race in recovery could return ETIMEDOUT to apps (623947)
4345 - building outside the source directory was fixed
4346 * the lustre-HOWTO was brought (more) up to date (582544)
4347 * major progress was made on recovery functionality
4349 2002-10-10 Phil Schwan <phil@clusterfs.com>
4352 - recovery deadlock fix
4353 - rm -rf causes LBUG fix (617817)
4354 - file open by multiple tasks fix (618962)
4355 - directory permissions bugs (602707 and 620007)
4356 - journal_stop fixed with locking (611313)
4357 - O_APPEND failures resolved (618273, perhaps 614459)
4358 - lconf PATH fix (619770)
4359 - IA64 build fix (621450)
4360 - RPC buffer sizes scale with amount of memory
4362 2002-10-01 Phil Schwan <phil@clusterfs.com>
4365 - locks would be cancelled without throwing away data pages,
4366 resulting in inconsistent data (605627)
4367 - inode attributes were not always being refreshed (605627, 612449)
4368 - lconf now continues to cleanup after lctl reports an error
4369 - MDS now enforces user permissions (602707)
4370 - lprocfs cleanup fixed, but not yet enabled (614157)
4371 - fixed infinite server hang, should a client not respond to an AST
4372 - avoid going into recovery if user calls readlink() with a buffer
4373 that's too small (613941)
4374 - AST RPCs no longer require replies (614867) -- this may be changed
4375 - don't crash server if client sends an IOV that's too big (611336)
4376 - fixed lock conversion deadlock (611892)
4377 - fixed the following of symlinks (614622)
4378 * recovery: the server can remove locks from a client that dies, other
4379 clients can make progress
4380 * more extN patch fixes
4381 * compile-time configurable ptlrpc buffer allocations
4383 - collaborative read cache document
4384 - Lustre Lite Performance CDR document-in-progress
4386 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4389 - fix typo in patch-2.4.18
4391 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4394 - clear ptlrpc request each time in handle_incoming_request()
4395 - unlink of files now destroys the object on the OST
4397 2002-09-19 Peter Braam <braam@clusterfs.com>
4399 * add hard link support
4400 * change obdfile creation method
4401 * kernel patch changed
4403 2002-09-19 Peter Braam <braam@clusterfs.com>
4406 - stack overflow bug in extN fixed
4408 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4410 * documentation updates
4411 - add man pages for config tools
4412 - update tests/README to describe testing with new config tools
4413 - finish metadata API descriptions
4414 * bug fixes and cleanups
4415 - statfs workaround for 16TB limit
4416 - LOV stripe allocation improved, can stripe on subset of OSTs
4417 - LOV file size/IO offset was wrong for files > 4GB in size
4418 - object EA data was being dropped, caused files to be unreadable
4419 - memory overflow with non-LOV OST caused memory corruption
4420 - fixed regression tests to work with new config tools, obdfilter
4421 - fixed bug when directory size became larger than 1 block
4422 - fixed bug (for single client case) when PWD was deleted
4423 - invalidate local directory pages when doing intent-based ops
4424 - avoid LDLM oops when lock callback contained bad data
4426 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4428 * documentation updates
4429 * bug fixes and cleanups
4430 - configuration tools
4433 - 64-bit compile warnings
4434 - 64-bit internal statfs data
4436 * test_brw on persistent OST devices
4438 * lprocfs (disabled)
4440 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4442 * documentation updates
4443 * bug fixes and cleanups
4444 * configuration tools
4446 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4449 * many small fixes to 0_5_4
4450 * io/network handling
4451 * thinkos in MDS operations
4453 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4456 * crucial basic fixes to 0.5.3
4457 * IOR, Iozone work over Elan
4460 2002-08-07 Phil Schwan <phil@clusterfs.com>
4461 * version 0_5_3, our first alpha
4462 * we use the new Portals iovs
4463 * documentation updates
4464 * bug fixes and cleanups
4465 * small changes in the DLM wire protocol
4467 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4468 * version 0_5_1 with some initial stability,
4469 * locking on MD and file I/O.
4470 * documentation updates
4471 * several bug fixes since 0.5.0
4472 * small changes in wire protocol
4474 2002-07-18 Phil Schwan <phil@clusterfs.com>
4476 * delivered as Lustre Light Alpha
4477 * fixed a crash after handling invalid MDS requests
4478 * fixed directory pages for architectures with non-4k pages sizes
4480 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4481 * release version v0_4_4
4482 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4484 * Fixes a number of interruption problems with OST operations.
4485 * Update documentation for portals header changes
4486 * Move all wire protocol structs/defines to lustre_idl.h
4487 * Fixes symlink length bug.
4488 * Add tcpdump to repository.
4490 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4491 * release version v0_4_3
4492 * Fixes statfs for inodes on extN.
4493 * Fixes bug in runtests which would delete /etc/hosts.
4494 * Use 64-bit object IDs wherever possible (not into VFS though)
4495 Remove ost_get_info, which is unused by lustre, and out of date.
4497 2002-07-03 Peter Braam <braam@clusterfs.com>
4498 * release version v0_4_2 Fixes a lookup error (type not passed)
4499 * move forward to head of Portals
4500 * move forward to latest Lustre kernel
4502 2002-06-25 Peter Braam <braam@clusterfs.com>
4503 * release version v0_4_1. Hopefully stable on single node use.