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 2.6.18-53.1.14.el5 (RHEL 5),
7 2.6.22.14 vanilla (kernel.org).
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.16 - 2.6.21 vanilla (kernel.org)
11 * Recommended e2fsprogs version: 1.40.7-sun1
12 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
13 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
14 removed cwd "./" (refer to Bugzilla 14399).
18 Frequency : on ppc only
19 Description: not convert ost objects for directory because it's not exist.
20 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
24 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
26 Severity : enhancement
28 Description: Update to support 2.6.22.14 vanilla kernel.
32 Frequency : rare, on recovery
33 Description: read procfs can produce deadlock in some situation
34 Details : Holding lprocfs lock which send rpc can produce block for destroy
35 obd objects and this also block reconnect with -EALREADY. This isn't
36 fix all lprocfs bugs - but make it rare.
38 Severity : enhancement
40 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
43 Frequency : frequent on X2 node
45 Description: mdc_set_open_replay_data LBUG
46 Details : Set replay data for requests that are eligible for replay.
50 Description: lustre_mgs: operation 101 on unconnected MGS
51 Details : When MGC is disconnected from MGS long enough, MGS will evict the
52 MGC, and late on MGC cannot successfully connect to MGS and a lot
53 of the error messages complaining that MGS is not connected.
56 Frequency : on start mds
58 Description: Implement get_info(last_id) in obdfilter.
61 Frequency : occasional
63 Description: Correctly check stale fid, not start epoch if ost not support SOM
64 Details : open with flag O_CREATE need set old fid in op_fid3 because op_fid2
65 overwrited with new generated fid, but mds can anwer with one of these
66 two fids and both is not stale. setattr incorectly start epoch and
67 assume will be called done_writeting, but without SOM done_writing
71 Frequency : rare, depends on device drivers and load
73 Description: MDS or OSS nodes crash due to stack overflow
74 Details : Code changes in 1.8.0 increased the stack usage of some functions.
75 In some cases, in conjunction with device drivers that use a lot
76 of stack the MDS (or possibly OSS) service threads could overflow
77 the stack. One change which was identified to consume additional
78 stack has been reworked to avoid the extra stack usage.
81 Frequency : occasional
83 Description: Do not fail import if osc_interpret_create gets -EAGAIN
84 Details : If osc_interpret_create got -EAGAIN it immediately exits and
85 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
86 oscc_has_objects and see OSC has no objests and call
87 oscc_internal_create to resend create request.
89 Severity : enhancement
91 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
93 Severity : enhancement
95 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
100 Description: Fix replay if there is an un-replied request and open
101 Details : In some cases, older replay request will revert the
102 mcd->mcd_last_xid on MDS which is used to record the client's
105 Severity : enhancement
107 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
109 Severity : enhancement
111 Description: Add rhel5 support to HEAD.
113 Serverity : enhancement
115 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
120 Description: Don't allow skipping OSTs if index has been specified.
121 Details : Don't allow skipping OSTs if index has been specified, make locking
122 in internal create lots better.
126 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
127 Details : access to bitfield in structure is always rounded to long
128 and this produce problem with not atomic change any bit.
132 Description: Lustre make rpms failed.
133 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
134 builds Lustre rpms from tarball.
137 Frequency : rare on shutdown ost
139 Description: If llog cancel was not send before clean_exports phase, this can
140 produce deadlock in llog code.
141 Details : If llog thread has last reference to obd and call class_import_put
142 this produce deadlock because llog_cleanup_commit_master wait when
143 last llog_commit_thread exited, but this never success because was
144 called from llog_commit_thread.
148 Description: allow userland application know is lost one of stripes.
149 Details : fill lvb_blocks with error code on ost and return it to
150 application if error flag found.
154 Description: NULL lov_tgts causing MDS oops
155 Details : more safe checks for NULL lov_tgts for avoid oops.
157 Severity : enhancement
159 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
163 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
164 Details : If an OST is down the MDS will hang indefinitely in
165 obd_statfs_rqset() waiting for the statfs data. While for
166 MDS QOS usage of statfs, it should not stuck in waiting.
168 Severity : enhancement
170 Description: remote_acl support
171 Details : Support ACL-based permission check for remote user.
172 Support setfacl/getfacl for remote user with the utils
173 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
174 the system "{s,g}etfacl" utils.
176 Severity : enhancement
178 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
180 Severity : enhancement
182 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
186 Description: make mgs_setparam() handle fsname containing dash
187 Details : fsname containing a dash does not work with lctl conf_param
189 Severity : enhancement
191 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
193 Severity : enhancement
195 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
199 Frequency : rare, at shutdown
200 Description: access already free / zero obd_namespace.
201 Details : if client_disconnect_export was called without force flag set,
202 and exist connect request in flight, this can produce access to
203 NULL pointer (or already free pointer) when connect_interpret
204 store ocd flags in obd_namespace.
208 Frequency : only at startup
209 Description: not alloc memory with spinlock held.
210 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
211 if any spinlock held.
213 Severity : enhancement
215 Description: make lustre randomly fail allocating memory
216 Details : Make lustre randomly failed allocating memory for testing purpose.
218 Severity : enhancement
220 Description: lost problems with lov objid file
221 Details : Fixes some scability and access to not inited memory problems
222 in work with lov objdid file.
227 Description: lfs find does not continue on file error
228 Details : Continue other files processing when a file/dir is absent.
232 Description: Inconsistent usage of lustre_pack_reply()
233 Details : Standardize the usage of lustre_pack_reply() such that it
234 always generate a CERROR on failure.
237 Frequency : occasional
239 Description: MDS hang or stay in waiting lock
240 Details : If client receive lock with CBPENDING flag ldlm need send lock
241 cancel as separate rpc, to avoid situation when cancel request
242 can't processed due all i/o threads stay in wait lock.
246 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
250 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
254 Description: Update to SLES9 kernel-2.6.5-7.287.3.
258 Description: Update to RHEL4 latest kernel.
260 Severity : enhancement
262 Description: Build SLES10 patchless client fails
263 Details : The configure was broken by run ./configure with
264 --with-linux-obj=.... argument for patchless client. When the
265 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
266 can't search header adequately. Use absolute path such as
267 -I($LINUX)/include instead.
271 Description: interrupt oig_wait produce painc on resend.
272 Details : brw_redo_request can be used for resend requests from ptlrpcd and
273 private set, and this produce situation when rq_ptlrpcd_data not
274 copyed to new allocated request and triggered LBUG on assert
275 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
276 ptlrpcd set if request is changed and can be safety changed to use
281 Description: LASSERT_{REQ,REP}SWAB macros are buggy
282 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
283 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
284 these macros and replace them with their difinitions instead.
288 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
289 Details : Update which_patch & target file for SLES10 latest kernel.
292 Description: Remove client patches from SLES 10 kernel.
293 Details : This causes SLES 10 clients to behave as patchless clients
294 even on a Lustre-patched (server) kernel.
296 Severity : enhancement
298 Description: self-adjustable client's lru lists
299 Details : use adaptive algorithm for managing client cached locks lru
300 lists according to current server load, other client's work
301 pattern, memory activities, etc. Both, server and client
302 side namespaces provide number of proc tunables for controlling
305 Severity : enhancement
307 Description: light-weight GSS support
308 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
309 and reduce performance overhead.
311 Severity : enhancement
313 Description: Linux keyring support
314 Details : Support using service of Linux keyring for Lustre GSS internal
315 context refresh/cache mechanism.
319 Description: Fix errors in lfs documentation
320 Details : Fixes man pages
324 Description: don't use GFP_* in generic Lustre code.
325 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
329 Description: obdclass is limited by single OBD_ALLOC(idarray)
330 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
334 Description: warnings with build patchless client with vanila 2.6.19 and up
335 Details : change old ctl_table style and replace ctl_table/ctl_table_header
336 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
340 Frequency : Only for SLES
341 Description: sanity_quota fail test_1
342 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
343 file, which makes TSTID[2] inunique.
347 Frequency : Only for SLES9
348 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
349 Details : Update target/ChangeLog/which_patch .
353 Frequency : Only for test_10 in sanity-quota.sh
354 Description: a bug in quota test code
355 Details : checking a test flag in a wrong place
359 Frequency : Only for test_13 in sanity-quota.sh
360 Description: a bug in quota test script
361 Details : shouldn't check overfull allocated quota
365 Description: jbd statistics
366 Details : Port older jbd statistics patch for sles10
368 Severity : enhancement
370 Description: Add printing inode info into message about error in writepage.
375 Description: reply_lock_interpret crash due to race with it and lock cancel.
376 Details : Do not replay locks that are being cancelled. Do not reference
377 locks by their address during replay, just by their handle.
381 Frequency : When flocks are used.
382 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
383 Details : Flock locks might destroy just granted lock if it could be merged
384 with another existing flock, this is done in completion handler,
385 so teach ldlm_cli_enquque_fini that this is a valid case for
391 Description: Oops in read and write path when failing to allocate lock.
392 Details : Check if lock allocation failed and return error back.
396 Description: lstripe command fails for valid OST index
397 Details : The stripe offset is compared to lov->desc.ld_tgt_count
398 instead of lov->desc.ld_active_tgt_count.
402 Description: sanity.sh failed test 103
403 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
408 Description: df doesn't work properly if diskfs blocksize != 4K
409 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
413 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
414 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
418 Description: obd_health_check_timeout too short
419 Details : set obd_health_check_timeout as 1.5x of obd_timeout
423 Description: llapi_file_create() does not allow some changes
424 Details : add llapi_file_open() that allows specifying the mode and
425 open flags, and also returns an open file handle.
429 Description: merge and cleanup kernel patches.
430 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
434 Description: Add journal checksum support.(Kernel part)
435 Details : The journal checksum feature adds two new flags i.e
436 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
437 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
438 indicates that the commit block contains the checksum for
439 the blocks described by the descriptor blocks. Now commit
440 record can be sent to disk without waiting for descriptor
441 blocks to be written to disk. This behavior is controlled
442 using JBD2_FEATURE_ASYNC_COMMIT flag.
446 Description: OSS needs mutliple precreate threads
447 Details : Add ability to start more than one create thread per OSS.
451 Description: Sanity.sh test_65k failed: setstripe should have succeeded
452 Details : execute lfs setstripe on client
456 Description: mds_obd_create error creating tmp object
457 Details : When the user sets quota on root, llog will be affected and can't
458 create files and write files.
461 Frequency : Always on ia64 patchless client, and possibly others.
463 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
464 Details : This allows the patchless client to be loaded on architectures
469 Description: RedHat Update kernel for RHEL5
470 Details : Add the kernel config file for RHEL5.
474 Description: RedHat Update kernel for RHEL5
475 Details : Modify the kernel config file more closer RHEL5.
479 Description: Build failure against Centos5 (RHEL5)
480 Details : Define PAGE_SIZE when it isn't present.
484 Description: test_7 on sanity-quota.sh can't be used on separate machines
485 Details : add facet to handle it
489 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
490 Details : with NFS, the anon dentry's parent was set to itself in
491 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
492 corresponding dentry other than use rec->ur_name.
494 Severity : enhancement
496 Description: lfs setstripe enhancement
497 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
501 Description: enable data checksumming by default
502 Details : enable checksum by default, allow --disable-checksum
503 configure option and "-o nochecksum" mount option. Checksums
504 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
508 Description: lustre support for RHEL5
509 Details : Add support for RHEL5.
513 Description: Client eviction due to failover config
514 Details : after a connection loss, the lustre client should attempt to
515 reconnect to the last active server first before trying the
516 other potential connections.
520 Description: when mds and osts use different quota unit(32bit and 64bit),
521 quota will be released repeatly.
522 Details : void sending multiple quota reqs to mds, which will keep the
523 status between the reqs.
527 Description: osts not allocated evenly to files
528 Details : change the condition to increase offset_idx
531 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
533 Description: Data corruption for OSTs that are formatted larger than 2TB
535 Details : When generating the bio request for lustre file writes the
536 sector number would overflow a temporary variable before being
537 used for the IO. The data reads correctly from Lustre (which
538 will overflow in a similar manner) but other file data or
539 filesystem metadata may be corrupted in some cases.
543 Description: Tune the kernel for good SCSI performance.
544 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
545 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
550 Description: rewrite ext2-derived code in obdclass/uuid.c
551 Details : rewrite inherited code (uuid parsing code from ext2 utils)
552 from scratch preserving functionality.
556 Description: "lfs find -obd UUID" prints directories
557 Details : "lfs find -obd UUID" will return all directory names instead
558 of just file names. It is incorrect because the directories
559 do not reside on the OSTs.
563 Description: use i_size_read and i_size_write in 2.6 port
564 Details : replace inode->i_size access with i_size_read/write()
567 Frequency : when removing large files
569 Description: scheduling issue during removal of large Lustre files
570 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
571 It causes scheduling issues when removing large files (17TB in the
575 Frequency : only with liblustre clients on XT3
577 Description: evictions taking too long
578 Details : allow llrd to evict clients directly on OSTs
581 Frequency : only on ppc
583 Description: /proc/fs/lustre/devices broken on ppc
584 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
585 We should make sure the type of 'index' is loff_t and then cast
586 explicitly as needed below. Do not assign an explicitly cast
590 Frequency : only for rhel5
592 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
593 Details : Modify the target file & which_kernel.
595 Severity : enhancement
597 Description: omit set fsid for export NFS
598 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
599 as FS_REQUIRES_DEV for avoid problems with generate fsid.
603 Frequency : Always, for kernels after 2.6.16
604 Description: Fix warning idr_remove called for id=.. which is not allocated.
605 Details : Last kernels save old s_dev before kill super and not allow
606 to restore from callback - restore it before call kill_anon_super.
610 Description: buffer overruns could theoretically occur
611 Details : llapi_semantic_traverse() modifies the "path" argument by
612 appending values to the end of the origin string, and a
613 overrun may occur. Adding buffer overrun check in liblustreapi.
617 Description: Fix error on 'ls .' at the top of the Lustre mount.
618 Details : Don't revalidate dentry if it is a root dentry.
622 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
623 Details : Modify vm-tunables-rhel4.patch.
627 Description: Kernel config for 2.6.18-vanilla.
628 Details : Modify targets/2.6-vanilla.target.in.
629 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
630 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
631 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
632 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
636 Description: improve handling recoverable errors
637 Details : if request processig with error which can be recoverable on server
638 request should be resend, otherwise page released from cache and
643 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
644 Details : Modify target file & which_patch.
645 A flaw was found in the IA32 system call emulation provided
646 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
647 value could be stored in the %RAX register, which could trigger an
648 out-of-bounds system call table access. An untrusted local user
649 could exploit this flaw to run code in the kernel
650 (ie a root privilege escalation). (CVE-2007-4573).
654 Description: O_DIRECT bypasses client statistics.
655 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
656 read_ahead_stats were updated. copied the stats section from
657 osc_send_oap_rpc() into async_internal().
661 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
665 Description: change order of libsysio includes
666 Details : '#include sysio.h' should always come before '#include xtio.h'
670 Description: handle "serious error: objid * already exists" more gracefully
671 Details : If LAST_ID value on disk is smaller than the objects existing in
672 the O/0/d* directories, it indicates disk corruption and causes an
673 LBUG(). If the object is 0-length, then we should use the existing
674 object. This will help to avoid a full fsck in most cases.
676 Severity : enhancement
678 Description: adapt the lustre_config script to support the upgrade case
679 Details : Add "-u" option for lustre_config script to support upgrading 1.4
680 server targets to 1.6 in parallel.
684 Description: To avoid grant space > avaible space when the disk is almost
685 full. Without this patch you might see the error "grant XXXX >
686 available" or some LBUG about grant, when the disk is almost
688 Details : In filter_check_grant, for non_grant cache write, we should
689 check the left space by if (*left > ungranted + bytes), instead
690 of (*left > ungranted), because only we are sure the left space
691 is enough for another "bytes", then the ungrant space should be
692 increase. In client, we should update cl_avail_grant only there
693 is OBD_MD_FLGRANT in the reply.
697 Description: Update RHEL 4 kernel to fix local root privilege escalation.
698 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
699 described in CVE-2007-4573. This problem could allow untrusted
700 local users to gain root access.
703 Frequency : when using O_DIRECT and quotas
705 Description: Incorrect file ownership on O_DIRECT output files
706 Details : block usage reported by 'lfs quota' does not take into account
707 files that have been written with O_DIRECT.
712 Description: touch file failed when fs is not full
713 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
714 otherwise we can get ENOSP while fs is not full.
718 Description: parallel lock callbacks
719 Details : Instead of sending blocking and completion callbacks as separated
720 requests, adding them to a set and sending in parallel.
723 Frequency : only for Cray XT3
724 Bugzilla : 12829/13455
725 Description: Changing primary group doesn't change the group lustre assigns to
727 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
728 overridden with the primary group provided by the group upcall,
729 whereas the supplied fsgid can be trusted if it is in the list of
730 supplementary groups returned by the group upcall.
732 Severity : enhancement
734 Description: Allow masking D_WARNING, D_ERROR messages from console
735 Details : Console messages can now be disabled via lnet.printk.
739 Description: User code with malformed file open parameter crashes client node
740 Details : Before packing join_file req, all the related reference should be
741 checked carefully in case some malformed flags cause fake
742 join_file req on client.
746 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
747 Details : ldlm_completion_ast() assumes that a lock is granted when the req
748 mode is equal to the granted mode. However, it should also check
749 that LDLM_FL_CP_REQD is not set.
753 Description: Heavy nfs access might result in deadlocks
754 Details : After ELC code landed, it is now improper to enqueue any mds
755 locks under och_sem, because enqueue might want to decide to
756 cancel open locks for same inode we are holding och_sem for.
760 Description: Client eviction while running blogbench
761 Details : A lot of unlink operations with concurrent I/O can lead to a
762 deadlock causing evictions. To address the problem, the number of
763 oustanding OST_DESTROY requests is now throttled to
764 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
765 callbacks are processed in priority.
769 Description: enable ACLs on MDS by default
770 Details : ACLs must be enabled on MDS by default.
773 Frequency : PPC/PPC64 only
775 Description: conflicts between asm-ppc64/types.h and lustre_types.h
776 Details : fix duplicated definitions between asm-ppc64/types.h and
777 lustre_types.h on PPC.
780 Frequency : PPC/PPC64 only
782 Description: asm-ppc/segment.h does not exist
783 Details : fix compile issue on PPC.
787 Description: better handle error messages in extents code
790 Frequency : RHEL4 only
792 Description: mkfs is very slow on IA64/RHEL4
793 Details : A performance regression has been discovered in the MPT Fusion
794 driver between versions 3.02.73rh and 3.02.99.00rh. As a
795 consequence, we have downgraded the MPT Fusion driver in the RHEL4
796 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
798 Severity : enhancement
800 Description: SNMP support enhancement
801 Details : Adding total number of sampled request for an MDS node in snmp
804 Severity : enhancement
806 Description: Optimize ldlm waiting list processing for PR extent locks
807 Details : When processing waiting list for read extent lock and meeting read
808 lock that is same or wider to it that is not contended, skip
809 processing rest of the list and immediatelly return current
810 status of conflictness, since we are guaranteed there are no
811 conflicting locks in the rest of the list.
815 Description: Time out and refuse to reconnect
816 Details : When the failover node is the primary node, it is possible
817 to have two identical connections in imp_conn_list. We must
818 compare not conn's pointers but NIDs, otherwise we can defeat
819 connection throttling.
823 Description: port llog fixes from b1_6 into HEAD
824 Details : Port llog reference couting and some llog cleanups from b1_6
825 (bug 10800) into HEAD, for protect from panic and access to already
826 free llog structures.
830 Description: Detect stride IO mode in read-ahead
831 Details : When a client does stride read, read-ahead should detect that and
832 read-ahead pages according to the detected stride pattern.
836 Description: data checksumming impacts single node performance
837 Details : add support for several checksum algorithm. Currently, only CRC32
838 and Adler-32 are supported. The checksum type can be changed on
839 the fly via /proc/fs/lustre/osc/*/checksum_type.
843 Description: use adler32 for page checksums
844 Details : when available, use the Adler-32 algorithm instead of CRC32 for
849 Description: build for x2 fails
850 Details : fix compile issue on Cray systems.
854 Description: Properly match for duplicate locks
855 Details : Due to different lock order from skiplists code, we need to
856 traverse entire list for now
859 Frequency : only on PPC/SLES10
861 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
862 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
863 be defined. Add a hack in mkfs_lustre.c to work around this compile
868 Description: LASSERT on MDS when client holding flock lock dies
869 Details : ldlm pool logic depends on number of granted locks equal to
870 number of released locks which is not true for flock locks, so
871 just exclude such locks from consideration.
875 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
876 Details : Use fsfilt_sync() for both the whole filesystem sync and
877 individual file sync to eliminate dangerous inode locking
878 with I_LOCK that can lead to a deadlock.
880 --------------------------------------------------------------------------------
882 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
884 * Support for kernels:
885 2.6.5-7.283 (SLES 9),
886 2.6.9-55.EL (RHEL 4),
887 2.6.16.46-0.14 (SLES 10),
888 2.6.18.8 vanilla (kernel.org)
889 * Client support for unpatched kernels:
890 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
891 2.6.16 - 2.6.22 vanilla (kernel.org)
892 * Due to recently discovered recovery problems, we do not recommend
893 using patchless RHEL 4 clients with this or any earlier release.
894 * Recommended e2fsprogs version: 1.39.cfs8
895 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
896 * Starting with this release, the ldiskfs backing filesystem required
897 by Lustre is now in its own package, lustre-ldiskfs. This package
898 should be installed. It is versioned separately from Lustre and
899 may be released separately in future.
903 Description: block reactivating mgc import until all deactivates complete
904 Details : Fix race when failing back MDT/MGS to itself (testing)
907 Frequency : at statup only
909 Description: mds_lov_synchronize race leads to various problems
910 Details : simultaneous MDT->OST connections at startup can cause the
911 sync to abort, leaving the OSC in a bad state.
913 Severity : enhancement
915 Description: add optional extra BUILD_VERSION info
916 Details : add a new environment variable (namely LUSTRE_VERS) which allows
917 to override the lustre version.
920 Frequency : 2.6.18 servers only
922 Description: ll_kern_mount() doesn't release the module reference
923 Details : The ldiskfs module reference count never drops down to 0
924 because ll_kern_mount() doesn't release the module reference.
929 Description: server LBUG when using old ost_num_threads parameter
930 Details : Accept the old ost_num_threads parameter but warn that it
931 is deprecated, and fix an off-by-one error that caused an LBUG.
936 Description: Transient SCSI error results in persistent IO issue
937 Details : iobuf->dr_error is not reinitialized to 0 between two
941 Frequency : sometimes when underlying device returns I/O errors
943 Description: OSTs not going read-only during write failures
944 Details : OSTs are not remounted read-only when the journal commit threads
945 get I/O errors because fsfilt_ext3 calls journal_start/stop()
946 instead of the ext3 wrappers.
950 Description: poor connect scaling with increasing client count
951 Details : Don't run filter_grant_sanity_check for more than 100 exports
952 to improve scaling for large numbers of clients.
955 Frequency : SLES10 only
957 Description: sanity-quota.sh quotacheck failed: rc = -22
958 Details : Quotas cannot be enabled on SLES10.
961 Frequency : liblustre clients only
963 Description: getdirentries does not give error when run on compute nodes
964 Details : getdirentries does not fail when the size specified as an argument
965 is too small to contain at least one entry
967 Severity : enhancement
969 Description: Add LNET router traceability for debug purposes
970 Details : If a checksum failure occurs with a router as part of the
971 IO path, the NID of the last router that forwarded the bulk data
972 is printed so it can be identified.
977 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
978 Details : Due to a race condition, liblustre clients were occasionally
981 Severity : enhancement
983 Description: lfs setstripe use optional parameters instead of postional
986 Severity : enhancement
988 Description: Nanosecond timestamp support for ldiskfs
989 Details : The on-disk ldiskfs filesystem has added support for nanosecond
990 resolution timestamps. There is not yet support for this at
991 the Lustre filesystem level.
994 Frequency : during server recovery
996 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
997 Details : request with rq_no_resend flag not awake l_wait_event if they get a
1001 Frequency : nfs export on patchless client
1003 Description: connectathon hang when test nfs export over patchless client
1004 Details : Disconnected dentry cannot be found with lookup, so we do not need
1005 to unhash it or make it invalid
1008 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1009 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1010 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1015 Description: open req refcounting wrong on reconnect
1016 Details : If reconnect happened between getting open reply from server and
1017 call to mdc_set_replay_data in ll_file_open, we will schedule
1018 replay for unreferenced request that we are about to free.
1019 Subsequent close will crash in variety of ways.
1020 Check that request is still eligible for replay in
1021 mdc_set_replay_data().
1026 Description: disable writes to filesystem when reading health_check file
1027 Details : the default for reading the health_check proc file has changed
1028 to NOT do a journal transaction and write to disk, because this
1029 can cause reads of the /proc file to hang and block HA state
1030 checking on a healthy but otherwise heavily loaded system. It
1031 is possible to return to the previous behaviour during configure
1032 with --enable-health-write.
1034 Severity : enhancement
1036 Description: 64-bit inode version
1037 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1038 to the inode. This will be required for version-based recovery.
1043 Description: MDS fails to start if a duplicate client export is detected
1044 Details : in some rare cases it was possible for a client to connect to
1045 an MDS multiple times. Upon recovery the MDS would detect this
1046 and fail during startup. Handle this more gracefully.
1048 Severity : enhancement
1050 Description: Add -o localflock option to simulate old noflock
1052 Details : This will achieve local-only flock/fcntl locks
1058 Description: log_commit_thread vs filter_destroy race leads to crash
1059 Details : Take import reference before releasing llog record semaphore
1064 Description: Wrong request locking in request set processing
1065 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1066 imp_delayed_list, in this place should be used imp_lock.
1069 Frequency : when reconnection
1071 Description: Grant Leak when osc reconnect to OST
1072 Details : When osc reconnect ost, OST(filter) should check whether it
1073 should grant more space to client by comparing fed_grant and
1074 cl_avail_grant, and return the granted space to client instead
1075 of "new granted" space, because client will call osc_init_grant
1076 to update the client grant space info.
1079 Frequency : when client reconnect to OST
1081 Description: Grant Leak when osc do resend and replay bulk write
1082 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1083 bulk write request, because the grant info while be sync between
1084 OSC and OST when reconnect, and we should ignore the grant info
1085 these of resend/replay write req.
1090 Description: Grant space more than avaiable left space sometimes.
1091 Details : When then OST is about to be full, if two bulk writing from
1092 different clients came to OST. Accord the avaliable space of the
1093 OST, the first req should be permitted, and the second one
1094 should be denied by ENOSPC. But if the seconde arrived before
1095 the first one is commited. The OST might wrongly permit second
1096 writing, which will cause grant space > avaiable space.
1099 Frequency : when client is evicted
1101 Description: Grant might be wrongly erased when osc is evicted by OST
1102 Details : when the import is evicted by server, it will fork another
1103 thread ptlrpc_invalidate_import_thread to invalidate the
1104 import, where the grant will be set to 0. While the original
1105 thread will update the grant it got when connecting. So if
1106 the former happened latter, the grant will be wrongly errased
1107 because of this race.
1112 Description: Checking Stale with correct fid
1113 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1114 to check whether it is stale, because sometimes, we want the
1115 enqueue happened anyway, and op_data.fid2 will not be initialized.
1117 Severity : enhancement
1119 Description: update patchless client
1120 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1123 Frequency : only with 2.4 kernel
1125 Description: random memory corruption
1126 Details : size of struct ll_inode_info is to big for union inode.u and this
1127 can be cause of random memory corruption.
1132 Description: Memory leak in recovery
1133 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1134 It should also check obd_fail before fsfilt_start, otherwise if
1135 fsfilt_start return -EROFS,(failover mds during mds recovery).
1136 then the req will return with repmsg->transno = 0 and rc = EROFS.
1137 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1138 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1139 be freed no matter whether fsfilt_commit success or not.
1142 Frequency : only with huge count clients
1144 Description: Prevents from taking the superblock lock in llap_from_page for
1146 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1147 for llap_from_page call in ll_removepage prevents from taking the
1148 superblock lock for a soon died page.
1153 Description: Not check open intent error before release open handle
1154 Details : in some rare cases, the open intent error is not checked before
1155 release open handle, which may cause
1156 ASSERTION(open_req->rq_transno != 0), because it tries to release
1157 the failed open handle.
1162 Description: Set cat log bitmap only after create log success.
1163 Details : in some rare cases, the cat log bitmap is set too early. and it
1164 should be set only after create log success.
1168 Description: Accessing a block bevice can re-enable I/O when Lustre is
1169 tearing down a device.
1170 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1175 Description: service threads may hog cpus when there are a lot of requests
1176 Details : Insert cond_resched to give other threads a chance to use some CPU
1181 Description: the cat log was not initialized in recovery
1182 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1183 unlink log on mds will not be initialized until mds post
1184 recovery. And also in mds post recovery, the unlink log will
1185 initialization will be done asynchronausly, so there will be race
1186 between add unlink log and unlink log initialization.
1190 Description: brw_stats were being printed incorrectly
1191 Details : brw_stats were being printed as log2 but all of them were not
1192 recorded as log2. Also remove some code duplication arising from
1193 filter_tally_{read,write}.
1197 Frequency : rare, only in recovery.
1198 Description: ASSERTION(req->rq_type != LI_POISON) failed
1199 Details : imp_lock should be held while iterating over imp_sending_list for
1200 prevent destroy request after get timeout in ptlrpc_queue_wait.
1204 Description: replay-single.sh test 52 fails
1205 Details : A lock's skiplist need to be cleanup when it being unlinked
1206 from its resource list.
1210 Description: Short directio read returns full requested size rather than
1212 Details : Direct I/O operations should return actual amount of bytes
1213 transferred rather than requested size.
1215 Severity : enhancement
1217 Description: metadata RPC reduction (e.g. for rm performance)
1218 Details : decrease the amount of synchronous RPC between clients and servers
1219 by canceling conflicing lock before the operation on the client
1220 and packing thier handles into the main operation RPC to server.
1222 Severity : enhancement
1224 Description: Async OSC create to avoid the blocking unnecessarily.
1225 Details : If a OST has no remain object, system will block on the creating
1226 when need to create a new object on this OST. Now, ways use
1227 pre-created objects when available, instead of blocking on an
1228 empty osc while others are not empty. If we must block, we block
1229 for the shortest possible period of time.
1231 Severity : enhancement
1233 Description: refine locking for avoid write wrong info into lov_objid file
1234 Details : fix possible races with add new target and write/update data in
1237 --------------------------------------------------------------------------------
1239 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1244 Frequency : on some architectures
1246 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1247 Details : Uninitialized flags sometimes cause configuration commands to
1251 Frequency : patchless clients only
1253 Description: missing __iget() symbol export
1254 Details : The __iget() symbol export is missing. To avoid the need for
1255 this on patchless clients the deathrow inode reaper is turned
1256 off, and we depend on the VM to clean up old inodes. This
1257 dependency was during via the fix for bug 12181.
1259 --------------------------------------------------------------------------------
1261 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1263 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1264 INTEROPERATE with older versions automatically. In many cases a
1265 special upgrade step is needed. Please read the
1266 user documentation before upgrading any part of a live system.
1267 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
1268 version will not interoperate with 1.6 betas before beta5 (1.5.95).
1269 * WARNING: Lustre configuration and startup changes are required with
1270 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1276 Severity : enhancement
1278 Description: MountConf
1279 Details : Lustre configuration is now managed via mkfs and mount
1280 commands instead of lmc and lconf. New obd types (MGS, MGC)
1281 are added for dynamic configuration management. See
1282 https://mail.clusterfs.com/wikis/lustre/MountConf for
1285 Severity : enhancement
1287 Description: dynamic OST addition
1288 Details : OSTs can now be added to a live filesystem
1290 Severity : enhancement
1292 Description: startup order invariance
1293 Details : MDTs and OSTs can be started in any order. Clients only
1294 require the MDT to complete startup.
1296 Severity : enhancement
1298 Description: parallel, asynchronous orphan cleanup
1299 Details : orphan cleanup is now performed in separate threads for each
1300 OST, allowing parallel non-blocking operation.
1302 Severity : enhancement
1304 Description: optimized stripe assignment
1305 Details : stripe assignments are now made based on ost space available,
1306 ost previous usage, and OSS previous usage, in order to try
1307 to optimize storage space and networking resources.
1309 Severity : enhancement
1311 Description: Permanently set tunables
1312 Details : All writable /proc/fs/lustre tunables can now be permanently
1313 set on a per-server basis, at mkfs time or on a live system.
1315 Severity : enhancement
1317 Description: Lustre message v2
1318 Details : Add lustre message format v2.
1320 Severity : enhancement
1322 Description: client OST exclusion list
1323 Details : Clients can be started with a list of OSTs that should be
1324 declared "inactive" for known non-responsive OSTs.
1328 Description: SPEC SFS validation failure on NFS v2 over lustre.
1329 Details : Changes the blocksize for regular files to be 2x RPC size,
1330 and not depend on stripe size.
1332 Severity : enhancement
1334 Description: Multiple MD RPCs in flight.
1335 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1336 To avoid overly-overloading MDS, introduce a limit on number of
1337 MDS RPCs in flight for a single client and add /proc controls
1338 to adjust this limit.
1340 Severity : enhancement
1342 Description: client read/write statistics
1343 Details : Add client read/write call usage stats for performance
1344 analysis of user processes.
1345 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1346 file access. extents_stats shows chunk size distribution.
1347 extents_stats_per_process show chunk size distribution per
1350 Severity : enhancement
1352 Description: mds statistics
1353 Details : Add detailed mds operations statistics in
1354 /proc/fs/lustre/mds/*/stats.
1358 Description: Failure of copying files with lustre special EAs.
1359 Details : Client side always return success for setxattr call for lustre
1360 special xattr (currently only "trusted.lov").
1365 Description: Refcount LNET uuids
1366 Details : The global LNET uuid list grew linearly with every startup;
1367 refcount repeated list entries instead of always adding to
1370 Severity : enhancement
1372 Description: Dynamic service threads
1373 Details : Within a small range, start extra service threads
1374 automatically when the request queue builds up.
1377 Frequency : mixed-endian client/server environments
1379 Description: mixed-endian crashes
1380 Details : The new msg_v2 system had some failures in mixed-endian
1383 Severity : enhancement
1385 Description: Easy OST removal
1386 Details : OSTs can be permanently deactivated with e.g. 'lctl
1387 conf_param lustre-OST0001.osc.active=0'
1389 Severity : enhancement
1391 Description: MGS proc entries
1392 Details : Added basic proc entries for the MGS showing what filesystems
1395 Severity : enhancement
1397 Description: provide MGS failover
1398 Details : Added config lock reacquisition after MGS server failover.
1400 Severity : enhancement
1402 Description: add Linux 2.4 support
1403 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1407 Description: a large application tries to do I/O to the same resource and dies
1408 in the middle of it.
1409 Details : Check the req->rq_arrival time after the call to
1410 ost_brw_lock_get(), but before we do anything about
1411 processing it & sending the BULK transfer request. This
1412 should help move old stale pending locks off the queue as
1413 quickly as obd_timeout.
1416 Frequency : when an incorrect nid is specified during startup
1418 Description: ptlrpc connect to non-existant node causes kernel crash
1419 Details : LNET can't be re-entered from an event callback, which
1420 happened when we expire a message after the export has been
1421 cleaned up. Instead, hand the zombie cleanup off to another
1424 Severity : enhancement
1426 Description: plain/inodebits lock performance improvement
1427 Details : Grouping plain/inodebits in granted list by their request modes
1428 and bits policy, thus improving the performance of search through
1432 Frequency : only if OST filesystem is corrupted
1434 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1435 Details : for a short time RPCs with bulk IO are in the replay list,
1436 but replay of bulk IOs is unimplemented. If the OST filesystem
1437 is corrupted due to disk cache incoherency and then replay is
1438 started it is possible to trip an assertion. Avoid putting
1439 committed RPCs into the replay list at all to avoid this issue.
1442 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1444 Description: System hang on startup
1445 Details : This bug allowed the liblustre (e.g. catamount) client to
1446 return to the app before handling all startup RPCs. This
1447 could leave the node unresponsive to lustre network traffic
1448 and manifested as a server ptllnd timeout.
1450 Severity : enhancement
1452 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1453 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1454 boolean to control whether to print peer debug info when a
1455 client's RPC times out.
1458 Frequency : only for kernels with patches from Lustre below 1.4.3
1460 Description: Remove old rdonly API
1461 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1464 Frequency : only for devices with external journals
1466 Description: Set external device read-only also
1467 Details : During a commanded failover stop, we set the disk device
1468 read-only while the server shuts down. We now also set any
1469 external journal device read-only at the same time.
1472 Frequency : when upgrading from 1.4 while trying to change parameters
1474 Description: The wrong (new) MDC name was used when setting parameters for
1475 upgraded MDT's. Also allows changing of OSC (and MDC)
1476 parameters if --writeconf is specified at tunefs upgrade time.
1479 Frequency : when setting specific ost indicies
1481 Description: QOS code breaks on skipped indicies
1482 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1483 created with --index need not be sequential.
1485 Severity : enhancement
1487 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1488 Details : The uninit_groups feature works in conjunction with the kernel
1489 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1490 up the pass1 processing of e2fsck. This is a read-only feature
1491 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1492 mount filesystems that have had this feature enabled.
1494 Severity : enhancement
1496 Description: Improve multi-block allocation algorithm to avoid fragmentation
1497 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1498 allocation locality and avoid filesystem fragmentation.
1500 ------------------------------------------------------------------------------
1502 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1504 * Support for kernels:
1505 2.4.21-47.0.1.EL (RHEL 3)
1506 2.6.5-7.283 (SLES 9)
1507 2.6.9-42.0.10.EL (RHEL 4)
1508 2.6.12.6 vanilla (kernel.org)
1509 2.6.16.27-0.9 (SLES 10)
1510 * Recommended e2fsprogs version: 1.39.cfs5
1512 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1516 Frequency : occasional, depends on client load and configuration
1517 Bugzilla : 12181, 12203
1518 Description: data loss for recently-modified files
1520 Details : In some cases it is possible that recently written or created
1521 files may not be written to disk in a timely manner (this should
1522 normally be within 30s unless client IO load is very high).
1523 The problem appears as zero-length files or files that are a
1524 multiple of 1MB in size after a client crash or client eviction
1525 that are missing data at the end of the file.
1527 This problem is more likely to be hit on clients where files are
1528 repeatedly created and unlinked in the same directory, clients
1529 have a large amount of RAM, have many CPUs, the filesystem has
1530 many OSTs, the clients are rebooted frequently, and/or the files
1531 are not accessed by other nodes after being written.
1533 The presence of the problem can be detected by looking at
1534 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1535 smaller than the second (nr_unused) then dirty files will not
1536 be flushed automatically to disk. "sync; sleep 10" should be
1537 run several times on the node before unmounting it to update
1538 Lustre (this is also safe to run on nodes without this problem).
1540 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1541 that can cause this same problem, so customers using that kernel
1542 also need to update the kernel in addition to Lustre. In order
1543 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1545 It is normal that files written just before a client crash (less
1546 than 30s) may not yet have been flushed to disk, even for local
1550 Frequency : frequent on thin XT3 nodes
1552 Description: UUID collision on thin XT3 Linux nodes
1553 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1554 randomly, since we relied on an insufficiently-seeded PRNG.
1559 Description: OSS hangs after "All ost request buffers busy"
1560 Details : A deadlock between quota and journal operations caused OSS
1561 hangs after printing "All ost request buffers busy."
1564 Frequency : always on liblustre builds
1566 Description: Cleanup compiler warnings on liblustre
1569 Frequency : always on liblustre builds on XT3
1571 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1575 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1576 > 32000 subdirectories
1577 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1578 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1579 subdirectory count crosses 32000. This will aid e2fsck to
1580 correctly handle more than 32000 subdirectories.
1583 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1585 Description: System hang on startup
1586 Details : This bug allowed the liblustre (e.g. catamount) client to
1587 return to the app before handling all startup RPCs. This
1588 could leave the node unresponsive to lustre network traffic
1589 and manifested as a server ptllnd timeout.
1591 Severity : enhancement
1593 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1594 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1595 boolean to control whether to print peer debug info when a
1596 client's RPC times out.
1601 Description: make O_SYNC working on 2.6 kernels
1602 Details : 2.6 kernels use different method for mark pages for write,
1603 so need add a code to lustre for O_SYNC work.
1608 Description: Failure to close file and release space on NFS
1609 Details : Put inode details into lock acquired in ll_intent_file_open.
1610 Use mdc_intent_lock in ll_intent_open to properly
1611 detect all kind of errors unhandled by mdc_enqueue.
1616 Description: proc file read during shutdown sometimes raced obd removal,
1618 Details : Add lock to prevent obd access after proc file removal.
1621 Frequency : Only for files larger than 4GB on 32-bit clients.
1623 Description: improperly doing page alignment of locks
1624 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1625 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1630 Description: RPCs being resent when they shouldn't be
1631 Details : Some RPCs that should not be resent are being resent. This
1632 can cause inconsistencies in the RPC state machine. Do not
1633 resend such requests.
1636 Frequency : rare, only with NFS export
1638 Description: Crash on NFS re-export node
1639 Details : under very unusual load conditions an assertion is hit in
1640 ll_intent_file_open()
1643 Frequency : only if OST filesystem is corrupted
1645 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1646 Details : for a short time RPCs with bulk IO are in the replay list,
1647 but replay of bulk IOs is unimplemented. If the OST filesystem
1648 is corrupted due to disk cache incoherency and then replay is
1649 started it is possible to trip an assertion. Avoid putting
1650 committed RPCs into the replay list at all to avoid this issue.
1655 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1656 Details : Large single O_DIRECT read and write calls can fail to allocate
1657 a sufficiently large buffer to process the request. In case of
1658 allocation failure the allocation is retried with a smaller
1659 buffer and broken into smaller requests.
1661 Severity : enhancement
1663 Description: Add -o localflock option to simulate old noflock behaviour.
1664 Details : This will achieve local-only flock/fcntl locks coherentness.
1669 Description: versioning check is incomplete
1670 Details : Checking the version difference of client vs. server, report
1671 error if the gap is too big.
1676 Description: add support PG_writeback bit
1677 Details : add support for PG_writeback bit for Lustre, for more carefull
1678 work with page cache in 2.6 kernel. This also fix some deadlocks
1679 and remove hack for work O_SYNC with 2.6 kernel.
1681 Severity : enhancement
1683 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1684 Details : The uninit_groups feature works in conjunction with the kernel
1685 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1686 up the pass1 processing of e2fsck. This is a read-only feature
1687 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1688 mount filesystems that have had this feature enabled.
1690 Severity : enhancement
1692 Description: Improve multi-block allocation algorithm to avoid fragmentation
1693 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1694 allocation locality and avoid filesystem fragmentation.
1696 ------------------------------------------------------------------------------
1698 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1700 * Support for kernels:
1701 2.6.9-42.0.3.EL (RHEL 4)
1702 2.6.5-7.276 (SLES 9)
1703 2.4.21-47.0.1.EL (RHEL 3)
1704 2.6.12.6 vanilla (kernel.org)
1705 2.6.16.21-0.8 (SLES10)
1706 * Recommended e2fsprogs version: 1.39.cfs2-0
1708 * The backwards-compatible /proc/sys/portals symlink has been removed
1709 in this release. Before upgrading, please ensure that you change
1710 any configuration scripts or /etc/sysctl.conf files that access
1711 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1712 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1713 in advance of the upgrade on any system running Lustre 1.4.6 or
1714 newer, since /proc/sys/lnet was added in that version.
1715 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1719 Frequency : only when quota is used
1721 Description: avoid scanning export list for quota master
1722 Details : Change the algorithms to avoid scanning export list in order
1723 to improve the efficiency.
1726 Frequency : MDS failover only, very rarely
1728 Description: "went back in time" messages on mds failover
1729 Details : The greatest transno may be lost when the current operation
1730 finishes with an error (transno==0) and the client's last_rcvd
1731 record is over-written. Save the greatest transno in the
1732 mds_last_transno for this case.
1735 Frequency : always for specific kernels and striping counts
1737 Description: client may get "Matching packet too big" without ACL support
1738 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1739 when trying to access files in certain configurations. The
1740 clients should in fact be denied when mounting because they do
1741 not understand ACLs.
1744 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1746 Description: many clients connecting with IO in progress causes connect timeouts
1747 Details : Avoid synchronous journal commits to avoid delays caused by many
1748 clients connecting/disconnecting when bulk IO is in progress.
1749 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1750 OST_IO_PORTAL to avoid delays behind potentially many pending
1754 Frequency : occasionally with multiple writers to a single file
1756 Description: shared writes to file may result in wrong size reported by stat()
1757 Details : Allow growing of kms when extent lock is cancelled
1760 Frequency : always with random mmap IO to multi-striped file
1762 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1763 Details : Only if the hole is at the end of OST object so that kms is too
1764 small. Fix is to increase kms accordingly in ll_nopage.
1767 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1769 Description: writes to a missing object would leak memory on the OST
1770 Details : If there is an inconsistency between the MDS and OST filesystems,
1771 such that the MDS references an object that doesn't exist, writes
1772 to that object will leak memory due to incorrect cleanup in the
1773 error handling path, eventually running out of memory on the OST.
1778 Description: Creating too long symlink causes lustre errors
1779 Details : Check symlink and name lengths before sending requests to MDS.
1782 Frequency : only if flock is enabled (not on by default)
1784 Description: posix locks not released on fd closure on 2.6.9+
1785 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1786 kernels, this caused such locks not to be released on fd close and
1787 then assertions on fs unmount about still used locks.
1790 Frequency : MDS failover only, very rarely
1792 Description: clients may get ASSERTION(granted_lock != NULL)
1793 Details : When request was taking a long time, and a client was resending
1794 a getattr by name lock request. The were multiple lock requests
1795 with the same client lock handle and
1796 mds_getattr_name->fixup_handle_for_resent_request found one of the
1797 lock handles but later failed with ASSERTION(granted_lock != NULL).
1802 Description: handle->h_buffer_credits > 0, assertion failure
1803 Details : h_buffer_credits is zero after truncate, causing assertion
1804 failure. This patch extends the transaction or creates a new
1808 Frequency : NFS re-export or patchless client
1809 Bugzilla : 11179, 10796
1810 Description: Crash on NFS re-export node (__d_move)
1811 Details : We do not want to hash the dentry if we don't have a lock.
1812 But if this dentry is later used in d_move, we'd hit uninitialised
1813 list head d_hash, so we just do this to init d_hash field but
1814 leave dentry unhashed.
1817 Frequency : NFS re-export or patchless client
1819 Description: NFS exports has problem with symbolic link
1820 Details : lustre client didn't properly install dentry when re-exported
1821 to NFS or running patchless client.
1824 Frequency : NFS re-export or patchless client
1826 Description: Various nfs/patchless fixes.
1827 Details : fixes reuse disconected alias for lookup process - this fixes
1828 warning "find_exported_dentry: npd != pd",
1829 fix permission error with open files at nfs.
1830 fix apply umask when do revalidate.
1833 Frequency : occasional
1835 Description: Crash on NFS re-export node
1836 Details : calling clear_page() on the wrong pointer triggered oops in
1837 generic_mapping_read().
1840 Frequency : rarely, using O_DIRECT IO
1842 Description: unaligned directio crashes client with LASSERT
1843 Details : check for unaligned buffers before trying any requests.
1846 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1848 Description: stale data returned from RAID cache
1849 Details : If only a small amount of IO is done to the RAID device before
1850 reading it again it is possible to get stale data from the RAID
1851 cache instead of reading it from disk.
1854 Frequency : always for sles10 kernel
1856 Description: sles10 support
1857 Details : ll_follow_link: compile fixes and using of nd_set_link
1858 under newer kernels.
1861 Frequency : depends on arch, kernel and compiler version, always on sles10
1864 Description: recursive or deep enough symlinks cause stack overflow
1865 Details : getting rid of large stack-allocated variable in
1869 Frequency : depends on hardware
1871 Description: lustre write performance loss in the SLES10 kernel
1872 Details : the performance loss is caused by using of write barriers in the
1873 ext3 code. The SLES10 kernel turns barrier support on by
1874 default. The fix is to undo that change for ldiskfs.
1876 ------------------------------------------------------------------------------
1878 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1880 * Support for kernels:
1881 2.6.9-42.0.3EL (RHEL 4)
1882 2.6.5-7.276 (SLES 9)
1883 2.4.21-47.0.1.EL (RHEL 3)
1884 2.6.12.6 vanilla (kernel.org)
1888 Frequency : quota enabled and large files being deleted
1890 Description: releasing more than 4GB of quota at once hangs OST
1891 Details : If a user deletes more than 4GB of files on a single OST it
1892 will cause the OST to spin in an infinite loop. Release
1893 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1898 Description: statfs data retrieved from /proc may be stale or zero
1899 Details : When reading per-device statfs data from /proc, in the
1900 {kbytes,files}_{total,free,avail} files, it may appear
1901 as zero or be out of date.
1904 Frequency : systems with MD RAID1 external journal devices
1906 Description: lconf's call to blkid is confused by RAID1 journal devices
1907 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1908 of returning all block devices that match the journal UUID.
1911 Frequency : always, for aggregate stripe size over 4GB
1913 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1914 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1915 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1918 Frequency : always if "lfs find" used on a local file/directory
1920 Description: "lfs find" segfaults if used on a local file/directory
1921 Details : The case where a directory component was not specified wasn't
1922 handled correctly. Handle this properly.
1925 Frequency : always on ppc64
1927 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1928 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1929 perform endianness conversion.
1932 Frequency : rarely (truncate to non-zero file size after write under load)
1933 Bugzilla : 10730, 10687
1934 Description: Files padded with zeros to next 4K multiple
1935 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1936 it is possible that files that are truncated to a non-zero size
1937 immediately after being written are filled with zero bytes beyond
1938 the truncated size. No file data is lost.
1940 Severity : enhancement
1942 Description: Allow recovery/failover for liblustre clients.
1943 Details : liblustre clients were unaware of failover configurations until
1946 Severity : enhancement
1948 Description: user file locks should fail when not mounting with flock option
1949 Details : Set up an error-returning stub in ll_file_operations.lock field
1950 to prevent incorrect behaviour when client is mounted without
1951 flock option. Also, set up properly f_op->flock field for
1955 Frequency : always on ia64
1957 Description: "lfs df" loops on printing out MDS statfs information
1958 Details : The obd_ioctl_data was not initialized and in some systems
1959 this caused a failure during the ioctl that did not return
1960 an error. Initialize the struct and return an error on failure.
1963 Frequency : SLES 9 only
1965 Description: Error of copying files with lustre special EAs as root
1966 Details : Client side always return success for setxattr call for lustre
1967 special xattr (currently only "trusted.lov").
1970 Frequency : rarely on clusters with both ia64+i386 clients
1972 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1973 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1974 clients are doing concurrent writes to the same file it is
1975 possible that the ia64 clients may LASSERT because the OST
1976 extent locks are not PAGE_SIZE aligned. Ensure that grown
1977 locks are always aligned on the request boundary.
1980 Frequency : specific use, occasional
1982 Description: Overwriting in use executable truncates on-disk binary image
1983 Details : If one node attempts to overwrite an executable in use by
1984 another node, we now correctly return ETXTBSY instead of
1985 truncating the file.
1987 Severity : enhancement
1989 Description: Async OSC create to avoid the blocking unnecessarily.
1990 Details : If a OST has no remain object, system will block on the creating
1991 when need to create a new object on this OST. Now, ways use
1992 pre-created objects when available, instead of blocking on an
1993 empty osc while others are not empty. If we must block, we block
1994 for the shortest possible period of time.
1999 Description: chmod on Lustre root is propagated to other clients
2000 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2001 invalid by the follow_mount time.
2004 Frequency : liblustre clients only
2006 Description: Race in 'instant cancel' lock handling could lead to such locks
2007 never to be granted in case of SMP MDS
2008 Details : Do not destroy not yet granted but cbpending locks in
2012 Frequency : replay/resend of open
2014 Description: non null lock assetion failure in mds_intent_policy
2015 Details : Trying to replay/resend lockless open requests resulted in
2016 mds_open() returning 0 with no lock. Now it sets a flag if
2017 a lock is going to be returned.
2019 Severity : enhancement
2021 Description: Checksum enhancements
2022 Details : New checksum enhancements allow for resending RPCs that failed
2025 Severity : enhancement
2027 Description: Tunables on number of dirty pages in cacche
2028 Details : Allow to set limit on number of dirty pages cached.
2033 Description: client crash on unmount - lock still has references
2034 Details : In some error handling cases it was possible to leak a lock
2035 reference on a client while accessing a file. This was not
2036 harmful to the client during operation, but would cause the
2037 client to crash when the filesystem is unmounted.
2040 Frequency : specific case, rare
2042 Description: ETXTBSY on mds though file not in use by client
2043 Details : ETXTBSY is no longer incorrectly returned when attempting to
2044 chmod or chown a directory that the user previously tried to
2045 execute or a currently-executing binary.
2048 Frequency : extremely rare except on liblustre-based clients
2050 Description: Lustre space not freed when files are deleted
2051 Details : Clean up open-unlinked files after client eviction. Previously
2052 the unlink was skipped and the files remained as orphans.
2057 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2058 Details : In some cases it was possible to send a blocking callback to
2059 a client doing a glimpse, even though that client didn't get
2060 a lock granted. When the glimpse lock is cancelled on the OST
2061 the freed lock is left on the waiting list and corrupted the list.
2064 Frequency : all core dumps
2066 Description: Broke core dumps to lustre
2067 Details : Negative dentry may be unhashed if parent does not have UPDATE
2068 lock, but some callers, e.g. do_coredump, expect dentry to be
2069 hashed after successful create, hash it in ll_create_it.
2071 ------------------------------------------------------------------------------
2073 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2075 * Support for kernels:
2076 2.6.9-42.0.2.EL (RHEL 4)
2077 2.6.5-7.276 (SLES 9)
2078 2.4.21-40.EL (RHEL 3)
2079 2.6.12.6 vanilla (kernel.org)
2083 Frequency : always on RHEL 3
2085 Description: Number of open files grows over time
2086 Details : The number of open files grows over time, whether or not
2087 Lustre is started. This was due to a filp leak introduced
2088 by one of our kernel patches.
2090 ------------------------------------------------------------------------------
2092 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2094 * Support for kernels:
2095 2.6.9-42.EL (RHEL 4)
2096 2.6.5-7.276 (SLES 9)
2097 2.4.21-40.EL (RHEL 3)
2098 2.6.12.6 vanilla (kernel.org)
2103 Bugzilla : 5719, 9635, 9792, 9684
2104 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2105 Details : If a server is under heavy load and cannot reply to new
2106 connection requests before the client resends the (re)connect,
2107 the connection handling code can behave badly if two service
2108 threads are concurrently handing separate (re)connections from
2109 the same client. Add better locking to the connection handling
2110 code, and ensure that only a single connection will be processed
2111 for a given client UUID, even if the lock is dropped.
2113 Severity : enhancement
2115 Description: add TCP zero-copy support to kernel
2116 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2117 sends if the hardware supports scatter-gather and checksumming.
2118 This allows socklnd to do client-write and server-read more
2119 efficiently and reduce CPU utilization from skbuf copying.
2122 Frequency : only if NFS exporting from client
2124 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2125 Details : If setting attributes on a file created under NFS that had
2126 never been opened it would be possible to oops the client
2127 if the file had no objects.
2130 Frequency : always for liblustre
2132 Description: liblustre client does MDS+OSTs setattr RPC for each write
2133 Details : When doing a write from a liblustre client, the client
2134 incorrectly issued an RPC to the MDS and each OST the file was
2135 striped over in order to update the timestamps. When writing
2136 with small chunks and many clients this could overwhelm the MDS
2137 with RPCs. In all cases it would slow down the write because
2138 these RPCs are unnecessary.
2140 Severity : enhancement
2142 Description: allow number of MDS service threads to be changed at module load
2143 Details : It is now possible to change the number of MDS service threads
2144 running. Adding "options mds mds_num_threads={N}" to the MDS's
2145 /etc/modprobe.conf will set the number of threads for the next
2146 time Lustre is restarted (assuming the "mds" module is also
2147 reloaded at that time). The default number of threads will
2148 stay the same, 32 for most systems.
2153 Description: OST crash if filesystem is unformatted or corrupt
2154 Details : If an OST is started on a device that has never been formatted
2155 or if the filesystem is corrupt and cannot even mount then the
2156 error handling cleanup routines would dereference a NULL pointer.
2161 Description: NULL pointer deref in llap_from_page.
2162 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2163 incorrect) mapping if the page was truncated/reclaimed while it was
2164 searched for. Check for this condition and skip such pages when
2165 doing readahead. Introduce extra check to llap_from_page() to
2166 verify page->mapping->host is non-NULL (so page is not anonymous).
2169 Frequency : Sometimes when using sys_sendfile
2171 Description: "page not covered by a lock" warnings from ll_readpage
2172 Details : sendfile called ll_readpage without right page locks present.
2173 Now we introduced ll_file_sendfile that does necessary locking
2174 around call to generic_file_sendfile() much like we do in
2178 Frequency : with certain MDS communication failures at client mount time
2180 Description: NULL pointer deref after failed client mount
2181 Details : a client connection request may delayed by the network layer
2182 and not be sent until after the PTLRPC layer has timed out the
2183 request. If the client fails the mount immediately it will try
2184 to clean up before the network times out the request. Add a
2185 reference from the request import to the obd device and delay
2186 the cleanup until the network drops the request.
2189 Frequency : occasionally during client (re)connect
2191 Description: assertion failure during client (re)connect
2192 Details : processing a client connection request may be delayed by the
2193 client or server longer than the client connect timeout. This
2194 causes the client to resend the connection request. If the
2195 original connection request is replied in this interval, the
2196 client may trip an assertion failure in ptlrpc_connect_interpret()
2197 which thought it would be the only running connect process.
2200 Frequency : only with obd_echo servers and clients that are rebooted
2202 Description: kernel BUG accessing uninitialized data structure
2203 Details : When running an obd_echo server it did not start the ping_evictor
2204 thread, and when a client was evicted an uninitialized data
2205 structure was accessed. Start the ping_evictor in the RPC
2206 service startup instead of the OBD startup.
2208 Severity : enhancement
2209 Bugzilla : 10193 (patchless)
2210 Description: Remove dependency on various unexported kernel interfaces.
2211 Details : No longer need reparent_to_init, exit_mm, exit_files,
2212 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2215 Frequency : rare (only users of deprecated and unsupported LDAP config)
2217 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2218 Details : LDAP apparently contains 'lustreName' attributes instead of
2219 'name'. A simple remapping of the name is sufficient.
2222 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2224 Description: waiting_locks_callback trips kernel BUG if client is evicted
2225 Details : Running with the dump_on_timeout debug flag turned on makes
2226 it possible that the waiting_locks_callback() can try to dump
2227 the Lustre kernel debug logs from an interrupt handler. Defer
2228 this log dumping to the expired_lock_main() thread.
2230 Severity : enhancement
2232 Description: Support NFS exporting on 2.6 kernels.
2233 Details : Implement non-rawops metadata methods for NFS server to use without
2234 changing NFS server code.
2237 Frequency : very rare (synthetic metadata workload only)
2239 Description: two racing renames might cause an MDS thread to deadlock
2240 Details : Running the "racer" program may cause one MDS thread to rename
2241 a file from being the source of a rename to being the target of
2242 a rename at exactly the same time that another thread is doing
2243 so, and the second thread has already enqueued these locks after
2244 doing a lookup of the target and is trying to relock them in
2245 order. Ensure that we don't try to re-lock the same resource.
2248 Frequency : only very large systems with liblustre clients
2250 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2251 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2252 all OSTs in parallel. This allows the eviction of stale liblustre
2253 clients to proceed much faster than if they were done in series,
2254 and also offers similar improvements for other set_info RPCs.
2259 Description: excessive CPU usage during initial read phase on client
2260 Details : During the initial read phase on a client, it would agressively
2261 retry readahead on the file, consuming too much CPU and impacting
2262 performance (since 1.4.5.8). Improve the readahead algorithm
2263 to avoid this, and also improve some other common cases (read
2264 of small files in particular, where "small" is files smaller than
2265 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2270 Description: MDS crash when receiving packet with unknown intent.
2271 Details : Do not LBUG in unknown intent case, just return -EFAULT
2273 Severity : enhancement
2274 Bugzilla : 9293, 9385
2275 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2276 Details : Do not serialize getattr (non-intent version) and statfs.
2279 Frequency : occasional, when OST network is overloaded/intermittent
2281 Description: client evicted by OST after bulk IO timeout
2282 Details : If a client sends a bulk IO request (read or write) the OST
2283 may evict the client if it is unresposive to its data GET/PUT
2284 request. This is incorrect if the network is overloaded (takes
2285 too long to transfer the RPC data) or dropped the OST GET/PUT
2286 request. There is no need to evict the client at all, since
2287 the pinger and/or lock callbacks will handle this, and the
2288 client can restart the bulk request.
2291 Frequency : Always when mmapping file with no objects
2293 Description: client crashes when mmapping file with no objects
2294 Details : Check that we actually have objects in a file before doing any
2295 operations on objects in ll_vm_open, ll_vm_close and
2301 Description: Request leak when working with deleted CWD
2302 Details : Introduce advanced request refcount tracking for requests
2303 referenced from lustre intent.
2305 Severity : Enhancement
2307 Description: Cache open file handles on client.
2308 Details : MDS now will return special lock along with openhandle, if
2309 requested and client is allowed to hold openhandle, even if unused,
2310 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2311 closing files for every read/write openration.
2313 Severity : Enhancement
2315 Description: Cache open negative dentries on client when possible.
2316 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2317 negative dentries on lock revocation.
2322 Description: Remounting a client read-only wasn't possible with a zconf mount
2323 Details : It wasn't possible to remount a client read-only with llmount.
2325 Severity : enhancement
2326 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2327 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2328 MPICH is included to add Lustre-specific ADIO interfaces.
2329 This is based closely on the UFS ADIO layer and only differs
2330 in file creation, in order to allow the OST striping to be set.
2331 This is user-contributed code and not supported by CFS.
2336 Description: extended inode attributes (immutable, append-only) work improperly
2337 when 2.4 and 2.6 kernels are used on client/server or vice versa
2338 Details : Introduce kernel-independent values for these flags.
2340 Severity : enhancement
2343 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2344 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2345 granularity. Now, Fractional MB granularity is be supported,
2346 this is very useful for low memory system.
2348 Severity : enhancement
2350 Description: Getattr by fid
2351 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2352 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2353 have locks and avoids getattr rpc requests for it on every stat.
2356 Frequency : Always, for filesystems larger than 2TB
2358 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2359 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2360 filesystems larger than 2TB to be mounted with mballoc enabled.
2363 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2365 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2366 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2367 filesystems larger than 2TB is mounted then the kernel will
2368 silently corrupt the start of the filesystem. CONFIG_LBD is
2369 enabled for all CFS-supported kernels, but the possibility of
2370 this happening with a modified kernel config exists.
2372 Severity : enhancement
2374 Description: add client O_DIRECT support for 2.6 kernels
2375 Details : It is now possible to do O_DIRECT reads and writes to files
2376 in the Lustre client mountpoint on 2.6 kernel clients.
2378 Severity : enhancement
2380 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2381 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2382 parallel, not waiting for response from every OST before sending
2383 a rpc to the next OST.
2388 Description: setattr vs write race when updating file timestamps
2389 Details : Client processes that update a file timestamp into the past
2390 right after writing to the file (e.g. tar) it is possible that
2391 the updated file modification time can be reset to the current
2392 time due to a race between processing the setattr and write RPC.
2394 Severity : enhancement
2396 Description: Bring 'lfs find' closer in line with regular Linux find.
2397 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2398 -print0 options and obtains all the needed info through the lustre
2401 Severity : enhancement
2403 Description: support up to 1024 configured devices on one node
2404 Details : change obd_dev array from statically allocated to dynamically
2405 allocated structs as they are first used to reduce memory usage
2410 Description: Flush dirty partially truncated pages during truncate
2411 Details : Immediatelly flush partially truncated pages in filter_setattr,
2412 this way we completely avoid having any pages in page cache on OST
2413 and can retire ugly workarounds during writes to flush such pages.
2418 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2419 Details : protect inode from truncation within vfs_unlink() context
2420 just take a reference before calling vfs_unlink() and release it
2421 when parent's i_sem is free.
2426 Description: last_id value checked outside lock on OST caused LASSERT failure
2427 Details : If there were multiple MDS->OST object precreate requests in
2428 flight, it was possible that the OST's last object id was checked
2429 outside a lock and incorrectly tripped an assertion. Move checks
2430 inside locks, and discard old precreate requests.
2433 Frequency : always, if extents are used on OSTs
2435 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2436 Details : OSTs using the extents format would not zero the high 16 bits of
2437 the index physical block number. This is not a problem for any
2438 OST filesystems smaller than 16TB, and no kernels support ext3
2439 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2440 new/modified files) and can be fixed for existing filesystems
2441 with e2fsprogs-1.39-cfs1.
2446 Description: import connection selection may be incorrect if timer wraps
2447 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2448 import connections to be ignored if the 32-bit jiffies counter
2449 wraps. Use a 64-bit jiffies counter.
2452 Frequency : very large clusters immediately after boot
2454 Description: LNET request buffers exhausted under heavy short-term load
2455 Details : If a large number of client requests are generated on a service
2456 that has previously never seen so many requests it is possible
2457 that the request buffer growth cannot keep up with the spike in
2458 demand. Instead of dropping incoming requests, they are held in
2459 the LND until the RPC service can accept more requests.
2462 Frequency : Sometimes during replay
2464 Description: Assertion failure in ll_local_open after replay.
2465 Details : If replay happened on an open request reply before we were able
2466 to set replay handler, reply will become not swabbed tripping the
2467 assertion in ll_local_open. Now we set the handler right after
2468 recognising of open request
2471 Frequency : very rare
2473 Description: kernel reports "badness in vsnprintf"
2474 Details : Reading from the "recovery_status" /proc file in small chunks
2475 may cause a negative length in lprocfs_obd_rd_recovery_status()
2476 call to vsnprintf() (which is otherwise harmless). Exit early
2477 if there is no more space in the output buffer.
2479 Severity : enhancement
2481 Description: clear OBD RPC statistics by writing to them
2482 Details : It is now possible to clear the OBD RPC statistics by writing
2483 to the "stats" file.
2488 Description: Client mtime is not the same on different clients after utimes
2489 Details : In some cases, the client was using the utimes() syscall on
2490 a file cached on another node. The clients now validate the
2491 ctime from the MDS + OSTs to determine which one is right.
2496 Description: Inability to activate failout mode
2497 Details : lconf script incorrectly assumed that in pythong string's numeric
2498 value is used in comparisons.
2501 Frequency : always with multiple stripes per file
2503 Description: Inefficient object allocation for mutli-stripe files
2504 Details : When selecting which OSTs to stripe files over, for files with
2505 a stripe count that divides evenly into the number of OSTs,
2506 the MDS is always picking the same starting OST for each file.
2507 Return the OST selection heuristic to the original design.
2512 Description: mount failures may take full timeout to return an error
2513 Details : Under some heavy load conditions it is possible that a
2514 failed mount can wait for the full obd_timeout interval,
2515 possibly several minutes, before reporting an error.
2516 Instead return an error as soon as the status is known.
2518 Frequency : quota enabled and large files being deleted
2520 Description: releasing more than 4GB of quota at once hangs OST
2521 Details : If a user deletes more than 4GB of files on a single OST it
2522 will cause the OST to spin in an infinite loop. Release
2523 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2528 Description: statfs data retrieved from /proc may be stale or zero
2529 Details : When reading per-device statfs data from /proc, in the
2530 {kbytes,files}_{total,free,avail} files, it may appear
2531 as zero or be out of date.
2534 Frequency : systems with MD RAID1 external journal devices
2536 Description: lconf's call to blkid is confused by RAID1 journal devices
2537 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2538 of returning all block devices that match the journal UUID.
2541 Frequency : always, for aggregate stripe size over 4GB
2543 Description: assertion fails when trying to use 4GB stripe size
2544 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2545 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2548 Frequency : always on ppc64
2550 Description: the first write on an ext3 filesystem with mballoc got stuck
2551 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2552 perform endianness conversion.
2554 ------------------------------------------------------------------------------
2556 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2558 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2559 INTEROPERATE with older versions automatically. Please read the
2560 user documentation before upgrading any part of a live system.
2561 * WARNING: Lustre networking configuration changes are required with
2562 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2565 * Support for kernels:
2566 2.6.9-22.0.2.EL (RHEL 4)
2567 2.6.5-7.244 (SLES 9)
2568 2.6.12.6 vanilla (kernel.org)
2571 Severity : enhancement
2572 Bugzilla : 7981/8208
2573 Description: Introduced Lustre Networking (LNET)
2574 Details : LNET is new networking infrastructure for Lustre, it includes
2575 a reorganized network configuration mode (see the user
2576 documentation for full details) as well as support for routing
2577 between different network fabrics. Lustre Networking Devices
2578 (LNDs) for the supported network fabrics have also been
2579 created for this new infrastructure.
2581 Severity : enhancement
2582 Description: Introduced Access control lists
2583 Details : clients can set ACLs on files and directories in order to have
2584 more fine-grained permissions than the standard Unix UGO+RWX.
2585 The MDS must be started with the "-o acl" mount option.
2587 Severity : enhancement
2588 Description: Introduced filesystem quotas
2589 Details : Administrators may now establish per-user quotas on the
2592 Severity : enhancement
2594 Description: Configuration change for the XT3
2595 The PTLLND is now used to run Lustre over Portals on the XT3
2596 The configure option(s) --with-cray-portals are no longer used.
2597 Rather --with-portals=<path-to-portals-includes> is used to
2598 enable building on the XT3. In addition to enable XT3 specific
2599 features the option --enable-cray-xt3 must be used.
2604 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2605 Details : A race between allocating a new llcd and re-getting the llcd_lock
2606 allowed another thread to grab newly-allocated llcd.
2608 Severity : enhancement
2610 Description: 2.6 OST async journal commit and locking fix to improve performance
2611 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2612 2.6 kernels are now async as they already were in 2.4 kernels so
2613 that they can commit concurrently with the network bulk transfer.
2614 For block-allocated files the filter allocation semaphore is held
2615 to avoid filesystem fragmentation during allocation. BKL lock
2616 removed for 2.6 xattr operations where it is no longer needed.
2621 Description: lconf incorrectly determined whether two IP networks could talk
2622 Details : In some more complicated routing and multiple-network
2623 configurations, lconf will avoid trying to make a network
2624 connection to a disjoint part of the IP space. It was doing the
2625 math incorrectly for one set of cases.
2630 Description: Fix for potential infinite loop processing records in an llog.
2631 Details : If an llog record is corrupted/zeroed, it is possible to loop
2632 forever in llog_process(). Validate the llog record length
2633 and skip the remainder of the block on error.
2636 Frequency : occasional (liblustre only)
2638 Description: liblustre could not open files whose last component is a symlink
2639 Details : sysio_path_walk() would incorrectly pass the open intent to
2640 intermediate path components.
2643 Frequency : rare (liblustre only with non-standard tuning)
2644 Bugzilla : 7201 (7350)
2645 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2646 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2647 after it has been released, only do this for referenced locks
2649 Severity : enhancement
2651 Description: specifying an (invalid) directory default stripe_size of -1
2652 would reset the directory default striping
2653 Details : stripe_size -1 was used internally to signal directory stripe
2654 removal, now use "all default" to signal dir stripe removal
2655 as a directory striping of "all default" is not useful
2658 Frequency : common for large clusters running liblustre clients
2660 Description: doing an ls when liblustre clients are running is slow
2661 Details : sending a glimpse AST to a liblustre client waits for every AST
2662 to time out, as liblustre clients will not respond. Since they
2663 cannot cache data we refresh the OST lock LVB from disk instead.
2665 Severity : enhancement
2667 Description: doing an ls at the same time as file IO can be slow
2668 Details : enqueue and other "small" requests can be blocked behind many
2669 large IO requests. Create a new OST IO portal for non-IO
2670 requests so they can be processed faster.
2673 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2675 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2676 created file with mode 000
2677 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2678 allow an ftruncate() as a non-root user to a file with mode 000.
2679 Lustre now respects this flag to disable mode checks when
2680 truncating a file owned by the user
2683 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2686 Description: Revoking locks from clients that went dead or catatonic might take
2688 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2689 cancellation of such locks instant on servers without waiting for
2690 any reply from clients. Clients drops these locks when cancel
2691 notification from server is received without replying.
2694 Frequency : liblustre-only, when liblustre client dies or becomes busy
2696 Description: Doing ls on Linux clients can take a long time with active
2698 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2699 granting such locks to it in the first place if possible. Locks
2700 are taken by proxy on the OST during the read or write and
2701 dropped immediately afterward. Add connect flags handling, do
2702 not grant locks to liblustre clients for glimpse ASTs.
2704 Severity : enhancement
2706 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2707 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2708 sequential reads and behaves badly with applications doing random
2709 reads. Improve it by reading ahead at least read region, and
2710 avoiding excessive large RPC for small reads.
2712 Severity : enhancement
2714 Description: Creating more than 1000 files for a single job may cause a load
2715 imbalance on the OSTs if there are also a large number of OSTs.
2716 Details : qos_prep_create() uses an OST index reseed value that is an
2717 even multiple of the number of available OSTs so that if the
2718 reseed happens in the middle of the object allocation it will
2719 still utilize the OSTs as uniformly as possible.
2724 Description: OST or MDS may oops in ping_evictor_main()
2725 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2726 but doesn't restart at beginning of obd_exports_timed list
2729 Severity : enhancement
2731 Description: improve by-nid export eviction on the MDS and OST
2732 Details : allow multiple exports with the same NID to be evicted at one
2733 time without re-searching the exports list.
2736 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2738 Description: MDS may oops in groups_free()
2739 Details : in rare race conditions a newly allocated group_info struct is
2740 freed again, and this can be NULL. The 2.4 compatibility code
2741 for groups_free() checked for a NULL pointer, but 2.6 did not.
2744 Frequency : common for liblustre clients doing little filesystem IO
2745 Bugzilla : 9352, 7313
2746 Description: server may evict liblustre clients accessing contended locks
2747 Details : if a client is granted a lock or receives a completion AST
2748 with a blocking AST already set it would not reply to the AST
2749 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2752 Frequency : lfs setstripe, only systems with more than 160 OSTs
2754 Description: unable to set striping with a starting offset beyond OST 160
2755 Details : llapi_create_file() incorrectly limited the starting stripe
2756 index to the maximum single-file stripe count.
2759 Frequency : LDAP users only
2761 Description: lconf did not handle in-kernel recovery with LDAP properly
2762 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2764 Severity : enhancement
2766 Description: bind OST threads to NUMA nodes to improve performance
2767 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2768 node and do their allocations there to localize memory access
2770 Severity : enhancement
2772 Description: llmount can determine client NID directly from Myrinet (GM)
2773 Details : the client NID code from gmnalnid was moved directly into
2774 llmount, removing the need to use this or specifying the
2775 client NID explicitly when mounting GM clients with zeroconf
2778 Frequency : if client is started with down MDS
2780 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2781 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2782 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2783 ptlrpc_import_delayed_req() because we didn't check if the
2784 request was interrupted, and we also didn't break out of the
2785 event loop if there was no timeout
2790 Description: data loss during non-page-aligned writes to a single file from
2791 both multiple nodes and multiple threads on one node at same time
2792 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2793 inconsistency led to false short-reads, that were cached and later
2794 used by ->prepare_write() to fill in partially written page,
2795 leading to data loss.
2798 Frequency : always, if lconf --abort_recovery used
2800 Description: lconf --abort_recovery fails with 'Operation not supported'
2801 Details : lconf was attempting to abort recovery on the MDT device and not
2804 Severity : enhancement
2806 Description: remove cleanup logs
2807 Details : replace lconf-generated cleanup logs with lustre internal
2808 cleanup routines. Eliminates the need for client-cleanup and
2811 Severity : enhancement
2813 Description: add support for EAs (user and system) on lustre filesystems
2814 Details : it is now possible to store extended attributes in the Lustre
2815 client filesystem, and with the user_xattr mount option it
2816 is possible to allow users to store EAs on their files also
2818 Severity : enhancement
2820 Description: Add possibility (config option) to show minimal available OST free
2822 Details : When compiled with --enable-mindf configure option, statfs(2)
2823 (and so, df) will return least minimal free space available from
2824 all OSTs as amount of free space on FS, instead of summary of
2825 free spaces of all OSTs.
2827 Severity : enhancement
2829 Description: do not expand extent locks acquired on OST-side
2830 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2831 by server: they are not cached anyway.
2834 Frequency : when mmap is used/binaries executed from Lustre
2836 Description: Unmmap pages before throwing them away from read cache.
2837 Details : llap_shrink cache now attempts to unmap pages before discarding
2838 them (if unmapping failed - do not discard). SLES9 kernel has
2839 extra checks that trigger if this unmapping is not done first.
2844 Description: lconf didn't resolve symlinks before checking to see whether a
2845 given mountpoint was already in use
2848 Frequency : when migrating failover services
2849 Bugzilla : 6395, 9514
2850 Description: When migrating a subset of services from a node (e.g. failback
2851 from a failover service node) the remaining services would
2852 time out and evict clients.
2853 Details : lconf --force (implied by --failover) sets the global obd_timeout
2854 to 5 seconds in order to quickly disconnect, but this caused
2855 other RPCs to time out too quickly. Do not change the global
2856 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2858 Severity : enhancement
2859 Frequency : if MDS is started with down OST
2860 Bugzilla : 9439,5706
2861 Description: Allow startup/shutdown of an MDS without depending on the
2862 availability of the OSTs.
2863 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2864 Add appropriate locking and lov-osc refcounts for safe
2865 cleaning. Add osc abort_inflight calls in case the
2866 synchronize never started.
2869 Frequency : occasional (Cray XT3 only)
2871 Description: root not authorized to access files in CRAY_PORTALS environment
2872 Details : The client process capabilities were not honoured on the MDS in
2873 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2874 been accessed by an authorized user then root was able to access
2875 the file on the local client also. The root user capabilities
2876 are now allowed on the MDS, as this environment has secure UID.
2879 Frequency : occasional
2881 Description: ldiskfs "too long searching" message happens too often
2882 Details : A debugging message (otherwise harmless) prints too often on
2883 the OST console. This has been reduced to only happen when
2884 there are fragmentation problems on the filesystem.
2889 Description: Division by zero in statfs when all OSCs are inactive
2890 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2891 lov_statfs divides by value returned by lov_get_stripecnt().
2895 Bugzilla : 9489, 3273
2896 Description: First write from each client to each OST was only 4kB in size,
2897 to initialize client writeback cache, which caused sub-optimal
2898 RPCs and poor layout on disk for the first writen file.
2899 Details : Clients now request an initial cache grant at (re)connect time
2900 and so that they can start streaming writes to the cache right
2901 away and always do full-sized RPCs if there is enough data.
2902 If the OST is rebooted the client also re-establishes its grant
2903 so that client cached writes will be honoured under the grant.
2908 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2909 Details : Now I/O RPCs go to different portal number and (presumably) fast
2910 lock requests (and glimses) and other RPCs get their own service
2911 threads pool that should be able to service those RPCs
2914 Severity : enhancement
2916 Description: Ability to exchange lustre version between client and servers and
2917 issue warnings at client side if client is too old. Also for
2918 liblustre clients there is ability to refuse connection of too old
2920 Details : New 'version' field is added to connect data structure that is
2921 filled with version info. That info is later checked by server and
2925 Frequency : rare, liblustre only.
2926 Bugzilla : 9296, 9581
2927 Description: Two simultaneous writes from liblustre at offset within same page
2928 might proceed at the same time overwriting eachother with stale
2930 Details : I/O lock withing llu_file_prwv was released too early, before data
2931 actually was hitting the wire. Extended lock-holding time until
2932 server acknowledges receiving data.
2935 Frequency : extremely rare. Never observed in practice.
2937 Description: avoid generating lustre_handle cookie of 0.
2938 Details : class_handle_hash() generates handle cookies by incrementing
2939 global counter, and can hit 0 occasionaly (this is unlikely, but
2940 not impossible, because initial value of cookie counter is
2941 selected randonly). Value of 0 is used as a sentinel meaning
2942 "unassigned handle" --- avoid it. Also coalesce two critical
2943 sections in this function into one.
2945 Severity : enhancement
2947 Description: allow liblustre clients to delegate truncate locking to OST
2948 Details : To avoid overhead of locking, liblustre client instructs OST to
2949 take extent lock in ost_punch() on client's behalf. New connection
2950 flag is added to handle backward compatibility.
2952 Severity : enhancement
2953 Bugzilla : 4928, 7341, 9758
2954 Description: allow number of OST service threads to be specified
2955 Details : a module parameter allows the number of OST service threads
2956 to be specified via "options ost ost_num_threads={N}" in the
2957 OSS's /etc/modules.conf or /etc/modprobe.conf.
2961 Bugzilla : 6146, 9635, 9895
2962 Description: servers crash with bad pointer in target_handle_connect()
2963 Details : In rare cases when a client is reconnecting it was possible that
2964 the connection request was the last reference for that export.
2965 We would temporarily drop the export reference and get a new
2966 one, but this may have been the last reference and the export
2967 was just destroyed. Get new reference before dropping old one.
2969 Severity : enhancement
2970 Frequency : if client is started with failover MDS
2972 Description: Allow multiple MDS hostnames in the mount command
2973 Details : Try to read the configuration from all specified MDS
2974 hostnames during a client mount in case the "primary"
2977 Severity : enhancement
2979 Description: Stop sending data to evicted clients as soon as possible.
2980 Details : Check if the client we are about to send or are sending data to
2981 was evicted already. (Check is done every second of waiting,
2982 for which l_wait_event interface was extended to allow checking
2983 of exit condition at specified intervals).
2986 Frequency : rare, normally only when NFS exporting is done from client
2988 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2990 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2993 Frequency : always, if srand() is called before liblustre initialization
2995 Description: Liblustre uses system PRNG disturbing its usage by user application
2996 Details : Introduce internal to lustre fast and high-quality PRNG for
2997 lustre usage and make liblustre and some other places in generic
2998 lustre code to use it.
3000 Severity : enhancement
3001 Bugzilla : 9477, 9557, 9870
3002 Description: Verify that the MDS configuration logs are updated when xml is
3003 Details : Check if the .xml configuration logs are newer than the config
3004 logs stored on the MDS and report an error if this is the case.
3005 Request --write-conf, or allow starting with --old_conf.
3007 Severity : enhancement
3009 Description: Handle symlinks in the path when checking if Lustre is mounted.
3010 Details : Resolve intermediate symlinks when checking if a client has
3011 mounted a filesystem to avoid duplicate client mounts.
3016 Description: lconf can hit an error exception but still return success.
3017 Details : The lconf command catches the Command error exception at the top
3018 level script context and will exit with the associated exit
3019 status, but doesn't ensure that this exit status is non-zero.
3024 Description: failure of ptlrpc thread startup can cause oops
3025 Details : Starting a ptlrpc service thread can fail if there are a large
3026 number of threads or the server memory is very fragmented.
3027 Handle this without oopsing.
3030 Frequency : always, only if liblustre and non-default acceptor port was used
3032 Description: liblustre cannot connect to servers with non-default acceptor port
3033 Details : tcpnal_set_default_params() was not called and was therefore
3034 ignoring the environment varaible TCPNAL_PORT, as well as other
3035 TCPNAL_ environment variables
3040 Description: two objects could be created on the same OST for a single file
3041 Details : If an OST is down, in some cases it was possible to create two
3042 objects on a single OST for a single file. No problems other
3043 than potential performance impact and spurious error messages.
3047 Bugzilla : 5681, 9562
3048 Description: Client may oops in ll_unhash_aliases
3049 Details : Client dcache may become inconsistent in race condition.
3050 In some cases "getcwd" can fail if the current directory is
3056 Description: Inode refcounting problems in NFS export code
3057 Details : link_raw functions used to call d_instantiate without obtaining
3058 extra inode reference first.
3062 Bugzilla : 9942, 9903
3063 Description: Referencing freed requests leading to crash, memleaks with NFS.
3064 Details : We used to require that call to ll_revalidate_it was always
3065 followed by ll_lookup_it. Also with revalidate_special() it is
3066 possible to call ll_revalidate_it() twice for the same dentry
3067 even if first occurence returned success. This fix changes semantic
3068 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3069 reference on a request referred from the intent.
3070 ll_intent_release() then releases such a request.
3073 Frequency : rare, normally benchmark loads only
3075 Description: unlinked inodes were kept in memory on the client
3076 Details : If a client is repeatedly creating and unlinking files it
3077 can accumulate a lot of stale inodes in the inode slab cache.
3078 If there is no other client load running this can cause the
3079 client node to run out of memory. Instead flush old inodes
3080 from client cache that have the same inode number as a new inode.
3083 Frequency : SLES9 2.6.5 kernel and long filenames only
3084 Bugzilla : 9969, 10379
3085 Description: utime reports stale NFS file handle
3086 Details : SLES9 uses out-of-dentry names in some cases, which confused
3087 the lustre dentry revalidation. Change it to always use the
3091 Frequency : rare, unless heavy write-truncate concurrency is continuous
3092 Bugzilla : 4180, 6984, 7171, 9963, 9331
3093 Description: OST becomes very slow and/or deadlocked during object unlink
3094 Details : filter_destroy() was holding onto the parent directory lock
3095 while truncating+unlinking objects. For very large objects this
3096 may block other threads for a long time and slow overall OST
3097 responsiveness. It may also be possible to get a lock ordering
3098 deadlock in this case, or run out of journal credits because of
3099 the combined truncate+unlink. Solution is to do object truncate
3100 first in one transaction without parent lock, and then do the
3101 final unlink in a new transaction with the parent lock. This
3102 reduces the lock hold time dramatically.
3105 Frequency : rare, 2.4 kernels only
3107 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3108 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3109 before unmounting their filesystem in order to avoid deadlock.
3110 The kernel_locked() function in 2.4 kernels only checks whether
3111 the kernel lock is held, not whether it is this process that is
3112 holding it as 2.6 kernels do.
3117 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3118 Details : The client ptlrpc code may be trying to reconnect to a down
3119 server before a previous connection attempt has timed out.
3120 Increase the reconnect interval to be longer than the connection
3121 timeout interval to avoid sending duplicate connections to
3125 Frequency : echo_client brw_test command
3127 Description: fix echo_client to work with OST preallocated code
3128 Details : OST preallocation code (5137) didn't take echo_client IO path
3129 into account: echo_client calls filter methods outside of any
3130 OST thread and, hence, there is no per-thread preallocated
3131 pages and buffers to use. Solution: hijack pga pages for IO. As
3132 a byproduct, this avoids unnecessary data copying.
3136 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3137 Description: Client can oops in mdc_commit_close() after open replay
3138 Details : It was possible for the MDS to return an open request with no
3139 transaction number in mds_finish_transno() if the client was
3140 evicted, but without actually returning an error. Clients
3141 would later try to replay that open and may trip an assertion
3142 Simplify the client close codepath, and always return an error
3143 from the MDS in case the open is not successful.
3146 Frequency : rare, 2.6 OSTs only
3148 Description: OST may deadlock under high load on fragmented files
3149 Details : If there was a heavy load and highly-fragmented OST filesystems
3150 it was possible to have all the OST threads deadlock waiting on
3151 allocation of biovecs, because the biovecs were not released
3152 until the entire RPC IO was completed. Instead, release biovecs
3153 as soon as they are complete to ensure forward IO progress.
3155 Severity : enhancement
3157 Description: Support for specifying external journal device at mount
3158 Details : If an OST or MDS device is formatted with an external journal
3159 device, this device major/minor is stored in the ext3 superblock
3160 and may not be valid for failover. Allow detecting and
3161 specifying the external journal at mount time.
3166 Description: Mounting an MDS with pending unlinked files may cause oops
3167 Details : target_finish_recovery() calls mds_postrecov() which returned
3168 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3169 considers this an error and immediately begins cleaning up the
3170 lov, just after starting the mds_lov process
3172 Severity : enhancement
3174 Description: Implement 'lfs df' to report actual free space on per-OST basis
3175 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3176 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3177 usage of objects; '-h' to report in human readable format.
3179 ------------------------------------------------------------------------------
3181 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
3188 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3189 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3190 at startup during orphan unlinking. Instead, skip buddy update
3191 before setup, as it will be regenerated after recovery is complete.
3196 Description: If an OST is inactive, its locks might reference stale inodes.
3197 Details : lov_change_cbdata() must iterate over all namespaces, even if
3198 they are inactive to clear inode references from the lock.
3200 Severity : enhancement
3201 Frequency : occasional, if non-standard max_dirty_mb used
3203 Description: Client will block write RPCs if not enough grant
3204 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3205 then the client will block writes while waiting for another RPC
3206 to complete instead of consuming its dirty limit. With change
3207 we get improved performance when max_dirty_mb is small.
3209 Severity : enhancement
3210 Bugzilla : 3389, 6253
3211 Description: Add support for supplementary groups on the MDS.
3212 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3213 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3214 lookups for user supplementary groups into a cache.
3218 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3219 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3220 for liblustre O_EXCL is enforced by MDS.
3224 Description: readdir never returns NULL in liblustre.
3225 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3226 dentry in struct dirent.
3230 Frequency : liblustre only, depends on application IO pattern
3231 Description: liblustre clients evicted if not contacting servers
3232 Details : Don't put liblustre clients into the ping_evictor list, so
3233 they will not be evicted by the pinger ever.
3235 Severity : enhancement
3237 Description: Add ability to evict clients by NID from MDS.
3238 Details : By echoing "nid:$NID" string into
3239 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3240 $NID would be instantly evicted from this MDS and from all active
3241 OSTs connected to it.
3245 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3246 Details : lookup_it_finish() used to query file size from OSTs that was not
3251 Description: service threads change working directory to that of init
3252 Details : Starting lustre service threads may pin the working directory
3253 of the parent thread, making that filesystem busy. Threads
3254 now change to the working directory of init to avoid this.
3258 Frequency : during shutdown only
3259 Description: shutdown with a failed MDS or OST can cause unmount to hang
3260 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3265 Frequency : 2.6 only
3266 Description: chmod/chown may include an extra supplementary group
3267 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3268 supplementary group and if none is specified this is used.
3271 Bugzilla : 5479 (6816)
3273 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3274 Details : If lookup is in progress on a file that is unlinked we might try
3275 to revalidate the inode and fail in revalidate after lookup is
3276 complete and ll_file_open() enqueues the open again but
3277 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3280 Frequency : always, if lconf --abort_recovery used
3282 Description: lconf --abort_recovery fails with 'Operation not supported'
3283 Details : lconf was attempting to abort recovery on the MDT device and not
3286 ------------------------------------------------------------------------------
3288 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3293 Frequency : rare (only unsupported configurations with a node running as an
3295 Bugzilla : 6514, 5137
3296 Description: Mounting a Lustre file system on a node running as an OST could
3298 Details : OSTs now preallocates memory needed to write out data at
3299 startup, instead of when needed, to avoid having to
3300 allocate memory in possibly low memory situations.
3301 Specifically, if the file system is mounted on on OST,
3302 memory pressure could force it to try to write out data,
3303 which it needed to allocate memory to do. Due to the low
3304 memory, it would be unable to do so and the node would
3305 become unresponsive.
3307 Severity : enhancement
3309 Description: Addition of lconf --service command line option
3310 Details : lconf now accepts a '--service <arg>' option, which is
3311 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3313 Severity : enhancement
3315 Description: Failover mode is now the default for OSTs.
3316 Details : By default, OSTs will now run in failover mode. To return to
3317 the old behaviour, add '--failout' to the lmc line for OSTs.
3319 Severity : enhancement
3321 Description: Health checks are now provided for MDS and OSTs
3322 Details : Additional detailed health check information on MSD and OSTs
3323 is now provided through the procfs health_check value.
3326 Frequency : occasional, depends on IO load
3328 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3329 after numerous create/delete cycles
3330 Details : The ext3 inode allocation policy would not allocate new inodes
3331 very well on the OSTs because there are no new directories
3332 being created. Instead we look for groups with free space if
3333 the parent directories are nearly full.
3338 Description: Network or server problems during mount may cause partially
3339 mounted clients instead of returning an error.
3340 Details : The config llog parsing code may overwrite the error return
3341 code during mount error handling, returning success instead
3347 Description: MDS can fail to allocate large reply buffers
3348 Details : After long uptimes the MDS can fail to allocate large reply
3349 buffers (e.g. zconf client mount config records) due to memory
3350 fragmentation or consumption by the buffer cache. Preallocate
3351 some large reply buffers so that these replies can be sent even
3352 under memory pressure.
3356 Frequency : rare (liblustre)
3357 Description: fsx running with liblustre complained that using truncate() to
3358 extend the file doesn't work. This patch corrects that issue.
3359 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3360 ATTR_SIZE and lsm use in llu_setattr_raw.
3364 Frequency : rare, only 2.6 kernels
3365 Description: Unusual file access patterns on the MDS may result in inode
3366 data being lost in very rare circumstances.
3367 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3368 code in the 2.6 kernel caused the inode and/or EA data not to
3369 be read from disk, causing single-file corruption.
3373 Frequency : rare, only 2.6 filesystems using extents
3374 Description: Heavy concurrent write and delete load may cause data corruption.
3375 Details : It was possible under high-load situations to have an extent
3376 metadata block in the block device cache from a just-unlinked
3377 file overwrite a newly-allocated data block. We now unmap any
3378 metadata buffers that alias just-allocated data blocks.
3382 Frequency : filesystems with default stripe_count larger than 77
3383 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3384 Details : lconf specifies an inode size of 4096 bytes when the default
3385 stripe_count is larger than 77. This conflicts with the default
3386 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3387 this case to avoid pinning too much memory for large EAs.
3389 ------------------------------------------------------------------------------
3391 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3396 Frequency : rare (extremely heavy IO load with hundreds of clients)
3398 Description: Client is evicted, gets IO error writing to file
3399 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3400 caused clients to be evicted for AST timeouts. The fixes in
3401 bug 5192 mean we no longer need to have such short AST timeouts
3402 so ldlm_timeout has been increased.
3405 Frequency : occasional during --force or --failover shutdown under load
3406 Bugzilla : 5949, 4834
3407 Description: Server oops/LBUG if stopped with --force or --failover under load
3408 Details : a collection of import/export refcount and cleanup ordering
3409 issues fixed for safer force cleanup
3412 Frequency : only filesystems larger than 120 OSTs
3413 Bugzilla : 5990, 6223
3414 Description: lfs getstripe would oops on a very large filesystem
3415 Details : lov_getconfig used kfree on vmalloc'd memory
3418 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3419 Bugzilla : 6242, 6243
3420 Description: reading from files that had been truncated to a non-zero size
3421 but never opened returned no data
3422 Details : ll_file_read() reads zeros from no-object files to EOF
3427 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3428 Details : There's little harm in aborting MDS/OSS recovery and letting it
3429 try again, so I removed the LASSERT and return an error instead.
3431 Severity : enhancement
3433 Description: New debugging infrastructure for tracking down data corruption
3434 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3435 (b) cover more of the client-side code path, and (c) try to narrow
3436 down where problems occurred
3440 Bugzilla : 3819, 4364, 4397, 6313
3441 Description: Racing close and eviction MDS could cause assertion in mds_close
3442 Details : It was possible to get multiple mfd references during close and
3443 client eviction, leading to one thread referencing a freed mfd.
3445 Severity: : enhancement
3446 Bugzilla : 3262, 6359
3447 Description: Attempts to reconnect to servers are now more aggressive.
3448 Details : This builds on the enhanced upcall-less recovery that was added
3449 in 1.4.2. When trying to reconnect to servers, clients will
3450 now try each server in the failover group every 10 seconds. By
3451 default, clients would previously try one server every 25 seconds.
3456 Description: After recovery, certain operations trigger a failed
3457 assertion on a client.
3458 Details : Failing over an mds, using lconf -d --failover, while a
3459 client was doing a readdir() call would cause the client to
3460 LBUG after recovery completed and the readdir() was resent.
3462 Severity : enhancement
3464 Description: Default groups are now added by lconf
3465 Details : You can now run lconf --group <servicename> without having to
3466 manually add groups with lmc.
3469 Frequency : occasional
3471 Description: Nodes with an elan id of 0 trigger a failed assertion
3474 Frequency : always when accessing e.g. tty/console device nodes
3476 Description: tty and some other devices nodes cannot be used on lustre
3477 Details : file's private_data field is used by device data and lustre
3478 values in there got lost. New field was added to struct file to
3479 store fs-specific private data.
3482 Frequency : when exporting Lustre via NFS
3484 Description: NFSD failed occasionally when looking up a path component
3485 Details : NFSD is looking up ".." which was broken in ext3 directories
3486 that had grown large enough to become hashed.
3489 Frequency : Clusters with multiple interfaces not on the same subnet
3491 Description: Nodes will repeatedly try to reconnect to an interface which it
3492 cannot reach and report an error to the log.
3493 Details : Extra peer list entries will be created by lconf with some peers
3494 unreachable. lconf now validates the peer before adding it.
3497 Frequency : Only if a default stripe is set on the filesystem root.
3499 Description: Setting a default stripe on the filesystem root prevented the
3500 filesystem from being remounted.
3501 Details : The client was sending extra request flags in the root getattr
3502 request and did not allocate a reply buffer for the dir EA.
3505 Frequency : occasional, higher if lots of files are accessed by one client
3506 Bugzilla : 6159, 6097
3507 Description: Client trips assertion regarding lsm mismatch/magic
3508 Details : While revalidating inodes the VFS looks up inodes with ifind()
3509 and in rare cases can find an inode that is being freed.
3510 The ll_test_inode() code will free the lsm during ifind()
3511 when it finds an existing inode and then the VFS later attaches
3512 this free lsm to a new inode.
3516 Bugzilla : 6422, 7030
3517 Description: MDS deadlock between mkdir and client eviction
3518 Details : Creating a new file via mkdir or mknod (starting a transaction
3519 and getting the ns lock) can deadlock with client eviction
3520 (gets ns lock and trying to finish a synchronous transaction).
3523 Frequency : occasional
3524 Description: While starting a server, the fsfilt_ext3 module could not be
3526 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3527 kernels. Previously, lconf would still use the ext3 name
3528 when trying to load modules. Now, it will correctly use
3529 ext3 on 2.4 and ldiskfs on 2.6.
3531 Severity : enhancement
3532 Description: The default stripe count has been changed to 1
3533 Details : The interpretation of the default stripe count (0, to lfs
3534 or lmc) has been changed to mean striping across a single
3535 OST, rather than all available. For general usage we have
3536 found a stripe count of 1 or 2 works best.
3538 Severity : enhancement
3539 Description: Add support for compiling against Cray portals.
3540 Details : Conditional compiling for some areas that are different
3544 Frequency : occasional
3545 Bugzilla : 6409, 6834
3546 Description: Creating files with an explicit stripe count may lead to
3547 a failed assertion on the MDS
3548 Details : If some OSTs are full or unavailable, creating files may
3549 trigger a failed assertion on the MDS. Now, Lustre will
3550 try to use other servers or return an error to the
3554 Frequency : occasional
3556 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3557 caused lock timeout during readahead (since 1.4.2).
3558 Details : Processes doing readahead might match a lock that hasn't been
3559 granted yet if there are overlapping and conflicting lock
3560 requests. The readahead process waits on ungranted lock
3561 (original lock is CBPENDING), while OST waits for that process
3562 to cancel CBPENDING read lock and eventually evicts client.
3564 Severity : enhancement
3566 Description: Initial enabling of flock support for clients
3567 Details : Implements fcntl advisory locking and file status functions.
3568 This feature is provided as an optional mount flag (default
3569 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3570 locking are implemented yet, and those that are are not guaranteed
3571 to be completely correct in production environments.
3572 mount -t lustre -o [flock|noflock] ...
3575 Frequency : occasional
3577 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3578 assertion in the ext3 JBD (journaling) layer.
3579 Details : The b_committed_data struct is protected by the big kernel lock
3580 in 2.4 kernels, serializing journal_commit_transaction() and
3581 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3582 access to this struct. In 2.6 kernels there is finer grained
3583 locking to improve SMP performance of the JBD layer.
3587 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3592 Description: merge and cleanup kernel patches.
3595 -----------------------------------------------------------------------------
3597 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3599 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3600 versions of Lustre. Please update all servers and clients to
3601 version 1.4.2 or later at the same time. You must also run
3602 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3603 to update the configuration logs.
3605 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3606 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3607 - swab llog records of type '0' so we get proper header size/idx (5861)
3608 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3609 - fix rename of one directory over another leaking an inode (5953)
3610 - avoid SetPageDirty on 2.6 (5981)
3611 - don't re-add just-being-destroyed locks to the waiting list (5653)
3612 - when creating new directories, inherit the parent's custom
3613 striping settings if present parent (3048)
3614 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3615 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3616 - don't hold client locks on temporary worklist from l_lru (5666)
3617 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3618 - automatically evict dead clients (5921)
3619 - Update file size properly in create+truncate+fstat case (6196)
3620 - Do not unhash mountpoint dentries, do not allow removal of
3622 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3623 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3624 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3625 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3626 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3627 - Correctly handle reads of files with no objects (6243)
3628 - lctl recover will also mark a device active if deactivate used (5933)
3630 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3631 - allow --write-conf on an MDS with different nettype than client (5619)
3632 - don't write config llogs to MDS for mounts not from that MDS (5617)
3633 - lconf should create multiple TCP connections from a client (5201)
3634 - init scripts are now turned off by default; run chkconfig --on
3635 lustre and chkconfig --on lustrefs to use them
3636 - upcalls are no longer needed for clients to recover to failover
3638 - add --abort-recovery option to lconf to abort recovery on device
3640 - add support for an arbitrary number of OSTs (3026)
3641 - Quota support protocol changes.
3642 - forward compatibility changes to wire structs (6007)
3643 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3644 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3645 - improved "lustre status" script
3646 - initialize blocksize for non-regular files (6062)
3647 - added --disable-server and --disable-client configure options (5782)
3648 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3649 - Vanilla 2.4.29 support
3650 - increase maximum number of obd devices to 520 (6242)
3651 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3652 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3653 SLES 9 2.6.5 kernels
3654 - limit stripes per file to 160 (the maximum EA size) (6093)
3656 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3659 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3660 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3661 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3662 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3663 - small ext3 extents cleanups and fixes (5733)
3664 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3665 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3666 - avoid needless client->OST connect, fix handle mismatch (5317)
3667 - fix DLM error path that led to out-of-sync client, long delays (5779)
3668 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3669 - fix several locking issues related to i_size (5492,5624,5654,5672)
3670 - don't move pending lock onto export if it is already evicted (5683)
3671 - fix kernel oops when creating .foo in unlinked directory (5548)
3672 - fix deadlock in obdfilter statistics vs. object create (5811)
3673 - use time_{before,after} to avoid timer jiffies wrap (5882)
3674 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3675 - Do not leak request if server was not able to process it (5154)
3676 - If mds_open unable to find parent dir, make that negative lookup(5154)
3677 - don't create new directories with extent-mapping (5909, 5936)
3679 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3680 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3681 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3682 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3683 - support for RHEL4 kernel on i686 (5773)
3684 - provide error messages when incompatible logs are encountered (5898)
3686 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3687 * version 1.4.0.10 (1.4.1 release candidate 1)
3689 - don't keep a lock reference when lock is not granted (4238)
3690 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3691 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3692 - drop import inflight refcount on signal_completed_replay error (5255)
3693 - unlock page after async write error during send (3677)
3694 - handle missing objects in filter_preprw_read properly (5265)
3695 - no transno return for symlink open, don't save no-trasno open (3440)
3696 - don't try to complete elan receive that already failed (4012)
3697 - free RPC server reply state on error (5406)
3698 - clean up thread from ptlrpc_start_thread() on error (5160)
3699 - readahead could read extra page into cache that wasn't ejected (5388)
3700 - prevent races in class_attach/setup/cleanup/detach (5260)
3701 - don't dereference de->d_inode after l_dput of de (5458)
3702 - use "int" for stripe value returned from lock_to_stripe (5544)
3703 - mballoc allocation and error-checking fixes in 2.6 (5504)
3704 - block device patches to fix I/O request sizes in 2.6 (5482)
3705 - look up hostnames for IB nals (5602)
3706 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3707 - don't start multiple acceptors for the same port (5277)
3708 - fix incorrect LASSERT in mds_getattr_name (5635)
3709 - export a proc file for general "ping" checking (5628)
3710 - fix "lfs check" to not block when the MDS is down (5628)
3712 - service request history (4965)
3713 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3714 - create an "evict_client" /proc entry on OSTs, like the MDS has
3715 - fix mount usage message, return errors per mount(8) (5168)
3716 - change grep [] to grep "[]" in tests so they work in more UMLs
3717 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3718 - remove ancient LOV_MAGIC_V0 EA support (5047)
3719 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3720 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3721 - allow readahead allocations to fail when low on memory (5383)
3722 - mmap locking landed again, after considerable improvement (2828)
3723 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3725 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3728 - send OST transaction number in read/write reply to free req (4966)
3729 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3730 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3731 - the watchdog thread now runs as interruptible (5246)
3732 - flock/lockf fixes (but it's still disabled, pending 5135)
3733 - don't use EXT3 constants in llite code (5094)
3734 - memory shortage at startup could cause assertion (5176)
3736 - reorganization of lov code
3737 - single portals codebase
3739 - add extents/mballoc support (5025)
3740 - direct I/O reads in the obdfilter (4048)
3741 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3743 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3746 - send OST transaction number in read/write reply to free req (4966)
3747 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3748 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3749 - the watchdog thread now runs as interruptible (5246)
3750 - handle missing objects in filter_preprw_read properly (5265)
3751 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3752 - drop import inflight refcount on signal_completed_replay error (5255)
3753 - unlock page after async write error during send (3677)
3754 - return original error code on reconstructed replies (3761)
3755 - no transno return for symlink open, don't save no-trasno open (3440)
3757 - add pid to ldlm debugging output (4922)
3758 - bump the watchdog timeouts -- we can't handle 30sec yet
3759 - extra debugging for orphan dentry/inode bug (5259)
3761 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3764 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3765 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3766 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3767 - fix NULL dereference in /proc/sys/portals/routes (4827)
3768 - allow failed mdc_close() operations to be interrupted (4561)
3769 - stop precreate on OST before MDS would time out on it (4778)
3770 - don't send partial-page writes before EOF from client (4410)
3771 - discard client grant for sub-page writes on large-page clients (4520)
3772 - don't free dentries not owned by NFS code, check generation (4806)
3773 - fix lsm leak if mds_create_objects() fails (4801)
3774 - limit debug_daemon file size, always print CERROR messages (4789)
3775 - use transno after validating reply (3892)
3776 - process timed out requests if import state changes (3754)
3777 - update mtime on OST during writes, return in glimpse (4829)
3778 - add mkfsoptions to LDAP (4679)
3779 - use ->max_readahead method instead of zapping global ra (5039)
3780 - don't interrupt __l_wait_event() during strace
3782 - add software watchdogs to catch hung threads quickly (4941)
3783 - make lustrefs init script start after nfs is mounted
3784 - fix CWARN/ERROR duplication (4930)
3785 - return async write errors to application if possible (2248)
3786 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3787 - print NAL number in %x format (4645)
3788 - update barely-supported suse-2.4.21-171 series (4842)
3789 - support for sles 9 %post scripts
3790 - support for building 2.6 kernel-source packages
3791 - support for sles km_* packages
3793 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3796 - ignore -ENOENT errors in osc_destroy (3639)
3797 - notify osc create thread that OSC is being cleaned up (4600)
3798 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3799 - reconstruct ost_handle() like mds_handle() (4657)
3800 - create a new thread to do import eviction to avoid deadlock (3969)
3801 - let lconf resolve symlinked-to devices (4629)
3802 - don't unlink "objects" from directory with default EA (4554)
3803 - hold socknal file ref over connect in case target is down (4394)
3804 - allow more than 32000 subdirectories in a single directory (3244)
3805 - fix blocks count for O_DIRECT writes (3751)
3806 - OST returns ENOSPC from object create when no space left (4539)
3807 - don't send truncate RPC if file size isn't changing (4410)
3808 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3809 - bind to privileged port in socknal and tcpnal (3689)
3811 - rate limit CERROR/CWARN console message to avoid overload (4519)
3812 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3813 - basic mmap support (3918)
3814 - kernel patch series update from b1_4 (4711)
3816 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3819 - avoid crash during MDS cleanup with OST shut down (2775)
3820 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3821 - don't use bad inodes on the MDS (3744)
3822 - dynamic object preallocation to improve recovery speed (4236)
3823 - don't hold spinlock over lock dumping or change debug flags (4401)
3824 - don't zero obd_dev when it is force cleaned (3651)
3825 - print grants to console if they go negative (4431)
3826 - "lctl deactivate" will stop automatic recovery attempts (3406)
3827 - look for existing locks in ldlm_handle_enqueue() (3764)
3828 - don't resolve lock handle twice in recovery avoiding race (4401)
3829 - revalidate should check working dir is a directory (4134)
3831 - don't always mark "slow" obdfilter messages as errors (4418)
3833 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3836 - don't close LustreDB during write_conf until it is done (3860)
3837 - fix typo in lconf for_each_profile (3821)
3838 - allow dumping logs from multiple threads at one time (3820)
3839 - don't allow multiple threads in OSC recovery (3812)
3840 - fix debug_size parameters (3864)
3841 - fix mds_postrecov to initialize import for llog ctxt (3121)
3842 - replace config semaphore with spinlock (3306)
3843 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3844 - don't allow enqueue to complete on a destroyed export (3822)
3845 - down write_lock before checking llog header bitmap (3825)
3846 - recover from lock replay timeout (3764)
3847 - up llog sem before sending rpc (3652)
3848 - reduce ns lock hold times when setting kms (3267)
3849 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3850 - fix NULL deref and obd_dev leak on setup error (3312)
3851 - replace some LBUG about llog ops with error handling (3841)
3852 - don't match INVALID dentries from d_lookup and spin (3784)
3853 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3854 - fix invalid assertion in ptlrpc_set_wait (3880)
3856 - add libwrap support for the TCP acceptor (3996)
3857 - add /proc/sys/portals/routes for non-root route listing (3994)
3858 - allow setting MDS UUID in .xml (2580)
3859 - print the stack of a process that LBUGs (4228)
3861 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3864 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3865 - make sure to unset replay flag from failed open requests (3440)
3866 - if default stripe count is 0, use OST count for inode size (3636)
3867 - update parent mtime/ctime on client for create/unlink (2611)
3868 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3869 - free recovery state on server during a forced cleanup (3571)
3870 - unregister_reply for resent reqs (3063)
3871 - loop back devices mounting and status check on 2.6 (3563)
3872 - fix resource-creation race that can provoke i_size == 0 (3513)
3873 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3874 - more debugging for page-accounting assertion (3746)
3875 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3876 - avoid deadlock after precreation failure (3758)
3877 - fix race and lock order deadlock in orphan handling (3450, 3750)
3878 - add validity checks when grabbing inodes from l_ast_data (3599)
3880 - add /proc/.../recovery_status to obdfilter (3428)
3881 - lightweight CDEBUG infrastructure, debug daemon (3668)
3882 - change default OSC RPC parameters to be better on small clusters
3883 - turn off OST read cache for files smaller than 32MB
3884 - install man pages and include them in rpms (3100)
3885 - add new init script for (un)mounting lustre filesystems (2593)
3886 - run chkconfig in %post for init scripts (3701)
3887 - drop scimac NAL (unmaintained)
3889 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3892 - clean kiobufs before and after use (3485)
3893 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3894 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3895 - change init script to fail more clearly if not run as root (1528)
3896 - allow clients to reconnect during replay (1742)
3897 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3898 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3899 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3900 - fix cond_resched (3554)
3901 - only evict unfinished clients after recovery (3515)
3902 - allow bulk resend, prevent data loss (3570)
3903 - dynamic ptlrpc request buffer allocation (2102)
3904 - don't allow unlinking open directory if it isn't empty (2904)
3905 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3906 - remove extraneous obd dereference causing LASSERT failure (3334)
3907 - don't use get_cycles() when creating temp. files on the mds (3156)
3908 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3909 - handle EEXIST for set-stripe, set proper directory name (3336)
3911 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3912 - fix ksocknal_fmb_callback() error messages (2918)
3914 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3917 - don't copy lvb into (possibly NULL) reply on error (2983)
3918 - don't deref dentry after dput, don't free lvb on error (2922)
3919 - use the kms to determine writeback rpc length (2947)
3920 - increment oti_logcookies when osc is inactive (2948)
3921 - update client's i_blocks count via lvb messages (2543)
3922 - handle intent open/close of special files properly (1557)
3923 - mount MDS with errors=remount-ro, like obdfilter (2009)
3924 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3925 - don't use cancelling-locks' kms values (2947)
3926 - use highest lock extent for kms, not last one (2925)
3927 - don't dereference ERR_PTR() dentry in error handling path (3107)
3928 - fix thread race in portals_debug_dumplog() (3122)
3929 - create lprocfs device entries at setup instead of at attach (1519)
3930 - common AST error handler, don't evict client on completion race (3145)
3931 - zero nameidata in detach_mnt in 2.6 (3118)
3932 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3933 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3934 - initialize RPC timeout timer earlier for 2.6 (3219)
3935 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3936 - print nal/nid for unknown nid (3258)
3937 - additional checks for oscc recovery before doing precreate (3284)
3938 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3939 - don't crash in mdc_close for bad permissions on open (3285)
3940 - zero i_rdev for non-device files (3147)
3941 - clear page->private before handing to FS, better assertion (3119)
3942 - tune the read pipeline (3236)
3943 - fix incorrect decref of invalidated dentry (2350)
3944 - provide read-ahead stats and refine rpc in flight stats (3328)
3945 - don't hold journal transaction open across create RPC (3313)
3946 - update atime on MDS at close time (3265)
3947 - close LDAP connection when recovering to avoid server load (3315)
3948 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3949 - don't leak open file on MDS after open resend (3325)
3950 - serialize filter_precreate and filter_destroy_precreated (3329)
3951 - loop device shouldn't call sync_dev() for nul device (3092)
3952 - clear page cache after eviction (2766)
3953 - resynchronize MDS->OST in background (2824)
3954 - refuse to mount the same filesystem twice on same mountpoint (3394)
3955 - allow llmount to create routes for mounting behind routers (3320)
3956 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3957 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3958 - fix rare problem with rename on htree directories (3417)
3960 - allow default OST striping configuration per directory (1414)
3961 - fix compilation for qswnal for 2.6 kernels (3125)
3962 - increase maximum number of MDS request buffers for large systems
3963 - change liblustreapi to be useful for external progs like lfsck (3098)
3964 - increase local configuration timeout for slow disks (3353)
3965 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3967 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3970 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3971 - don't overwrite extent policy data in reply if lock was blocked (2901)
3972 - drop filter export grants atomically with removal from device (2663)
3973 - del obd_self_export from work_list in class_disconnect_exports (2908)
3974 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3975 - swab reply message in mdc_close, other PPC fixes (2464)
3976 - fix destroying of named logs (2325)
3977 - overwrite old logs when running lconf --write_conf (2264)
3978 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3979 - fix race in target_handle_connect (2898)
3980 - mds_reint_create() should take same inode create lock (2926)
3981 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3982 - don't close files for self_export to avoid uninitialized obd (2936)
3983 - allow MDS with the same name as client node (2939)
3984 - hold dentry reference for closed log files for unlink (2325)
3985 - reserve space for all logs during transactions (2059)
3986 - don't evict page beyond end of stripe extent (2925)
3987 - don't oops on a deleted current working directory (2399)
3988 - handle hard links to targets without a parent properly (2517)
3989 - don't dereference NULL lock when racing during eviction (2867)
3990 - don't grow lock extents when lots of conflicting locks (2919)
3992 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3995 - account for cache space usage on clients to avoid data loss (974)
3996 - lfsck support in lustre kernel code (2349)
3997 - reduce journal credits needed for BRW writes (2370)
3998 - orphan handling to avoid losing space on client/server crashes
3999 - ptlrpcd can be blocked, stopping ALL progress (2477)
4000 - use lock value blocks to assist in proper KMS, faster stat (1021)
4001 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4002 - recovery for initial connections (2355)
4003 - fixes for mds_cleanup_orphans (1934)
4004 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4005 - block all file creations until orphan recovery completes (1901)
4006 - client remove rq_connection from request struct (2423)
4007 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4008 - recovery timer race (2670)
4009 - mdc_close recovey bug (2532)
4010 - ptlrpc cleanup bug (2710)
4011 - mds timeout on local locks (2588)
4012 - namespace lock held during RPCs (2431)
4013 - handle interrupted sync write properly (2503)
4014 - don't try to handle a message that hasn't been replied to (2699)
4015 - client assert failure during cleanup after abort recovery (2701)
4016 - leak mdc device after failed mount (2712)
4017 - ptlrpc_check_set allows timedout requests to complete (2714)
4018 - wait for inflight reqs when ptlrpcd finishes (2710)
4019 - make sure unregistered services are removed from the srv_list
4020 - reset bulk XID's when resending them (caught by 1138 test)
4021 - unregister_bulk after timeout
4022 - fix lconf error (2694)
4023 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4024 - readahead locks pages, leaves pending causing memory pressure (2673)
4025 - increase OST request buffers to 4096 on large machines (2729)
4026 - fix up permission of existing directories in simple_mkdir (2661)
4027 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4028 - don't assign transno to errored transactions (2742)
4029 - don't delete objects on OST if given a bogus objid from MDS (2751)
4030 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4031 - if rq_no_resend, then timeout request after recovery (2432)
4032 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4033 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4034 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4035 - MDS recovery shouldn't delete objects that it has given out (2730)
4036 - if enqueue arrives after completion, don't clobber LVB (2819)
4037 - don't unlock pages twice when trigger_group_io returns error (2814)
4038 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4039 - don't write pages to disk if there was an error (1450)
4040 - don't ping imports that have recovery disabled (2676)
4041 - take buffered bytes into account when balancing socknal conn (2817)
4042 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4043 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4044 - remove little-endian swabbing of llog records (1987)
4045 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4046 - retry reposting request buffers if they fail (1191)
4047 - grow extent at grant time to avoid granting a revoked lock (2809)
4048 - lock revoke doesn't evict page if covered by a second lock (2765)
4049 - disable VM readahead to avoid reading outside lock extents (2805)
4051 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4052 - updated kernel patches for hp-2.4.20 kernel (2681)
4054 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4057 - fix truncated write corruption (2366)
4058 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4060 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4061 - ensure len(uuid) < 37 in lmc (1171)
4062 - fix ia64 OOPS in llog_test (2255)
4063 - zero end of page at obdfilter for partial page writes (2648)
4064 - don't leave stale dentries around after renames (bug 2428)
4065 - fix timeouts when evicting a client with a single lock held (2642)
4066 - set deadline for the initial HELLO message to drain (2634)
4067 - print out dotted-quad IP addresses in the socknal (2302)
4069 - additional debugging for MDS client eviction problem (2443)
4070 - fix mkfsoptions support for osts (2603, 2604)
4072 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4075 - add series for the vanilla 2.6.0 kernel
4076 - add series for the vanilla 2.4.24 kernel
4077 - add series for a cray x86/64 UL kernel drop
4078 - fix xattr patches for the vanilla 2.4.19 series
4080 - generate true UUIDs in lmc (1171)
4081 - have portals stack dumping break in UML (2466)
4082 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4083 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4084 - wake callback waiting threads on client eviction (2460)
4085 - Add --ptldebug and --subsystem to lmc (1719)
4086 - update assertion to allow safe interrupt allocation
4087 - set rq_no_resend for cancel requests (2432)
4088 - recalculate ptlrpcd timeout after resend (2494)
4089 - call vfs_rmdir when removing pending directories (2368)
4090 - fix renaming a file to itself (2429)
4091 - lmc creates a default one-stripe lov (2454)
4092 - expand procfs space to handle large clusters (2326)
4093 - increase UML stack to avoid overflow
4094 - update lconf's list of debug and subsystem masks
4095 - fix lfs find --obd (2510)
4096 - /proc tunable for disabling filter read caching (2591)
4097 - stop rpm packages from altering slapd.conf (2301)
4098 - disable nagle in the socknal under 0conf (2578)
4099 - choose mds inode size based on stripe count (2572)
4100 - fix kernel-source rpm problems (2516)
4102 - add --disable-doc to avoid pdf generation (2421)
4103 - update documentation, tests, type-os, comments
4104 - avoid format warnings on ia64
4105 - remove the TOE NAL
4106 - tiny code cleanups by removing unused fields
4108 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4111 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4112 - remove the most blatant lies from BUILDING file (2371)
4113 - change default debug level to reasonable production setting
4114 - reduce client side cache size to reduce cache flush time
4115 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4116 - make TCP zerocopy and pinger support enabled by default (2476)
4117 - sync writes completed after process exits caused crashes (2319)
4118 - maintain correct mount count on the MDS (2356)
4119 - backout 1557, because 2316 wasn't really fixed
4120 - better file I/O statistics gathering in /proc
4121 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4122 - another kernel patch to fix zero-copy TCP function export
4123 - don't take duplicate lock when processing re-sent getattr (2420)
4124 - lctl uses obd_self_export instead of creating new conn (2353)
4125 - MDS/OST recovery case which requires object creation asserted (2425)
4126 - move lfs from /usr/sbin to /usr/bin in packages
4127 - fix race between mds_client_add and mds_client_free (2417)
4128 - use kmalloc instead of slabs in portals (2430)
4129 - don't create duplicate records when a failover MDS is present (2442)
4130 - remove unnecessary mount age check (2332)
4131 - don't remove directory inodes from locks prematurely (2451)
4132 - don't break if MDS service name is the same as hostname (2103)
4133 - fix races in client write RPC generation when cache full (2482)
4135 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4138 - remove now-unused request->rq_obd (278)
4139 - if an allocation fails, print out how much memory we've used (1933)
4140 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4141 - add the "configurable stack size" patch to most series files (1256)
4142 - ability to write large log records, for 100+ OST configs (2306)
4143 - fix NULL deref when filter_prep fails (2314)
4144 - fix operator precedence error in filter_sync
4145 - dynamic allocation of socknal TX descriptors (2315)
4146 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4147 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4148 - ensure that CWARN messages in Portals always get to the syslog
4149 - __init/__exit are not for prototype decls (ldlm_init/exit)
4150 - x86-64 compile warning fixes
4151 - fix gateway LMC keyword conflict (2318)
4152 - fix MDS lock inversions in getattr/reint paths (1844)
4153 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4154 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4155 - DLM race condition prevented some lock evictions (2328)
4156 - ENOMEM detection and retry on socknal sends (2230)
4157 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4158 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4159 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4160 - better error messages when a client is rejected during recovery (1505)
4161 - avoid cancelling locks which were never granted, after failure (2330)
4162 - fix i_sem/journal inversion in mds_client_add (2333)
4163 - fix truncate/getattr lock cycle deadlock (2334)
4164 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4165 - fix two rare exit paths which could leak an l_lock() ref (2321)
4166 - fencepost error in MDS/OST orphan recovery (2226)
4167 - make log record alignment 8 bytes (1988)
4168 - lstripe now fails when requested offset > ost_count (2237)
4169 - ensure that all kernel series have a complete list.h (1607)
4170 - fix crashes in special-file operations (2316)
4171 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4173 - allow configurable automake binary, for testing new versions
4174 - small update to the lfs documentation
4176 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4178 * fix negative export reference count in fsfilt_sync (2312)
4180 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4181 * release candidate 0.9.1
4183 - orphans are moved into the PENDING directory for possible recovery
4184 - replayed opens now open by fid for orphan/rename safety (1042)
4185 - last close of an orphan inode generates a transno (683)
4186 - chdir() and mount() now pin the directory entry (1020)
4187 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4188 - discard very old requests without processing them (1502)
4189 - remove some common, well-understood CERRORs (1505)
4190 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4191 - clear "grant" flags in OST replies until OST grant code lands (1644)
4192 - fix read performance by not clobbering i_blksize on client (1598)
4193 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4194 - make LRU size a /proc tunable, clears locks when reduced (707)
4195 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4196 - server locks take references on exports now (1558)
4197 - build fixes for 2.4.20-rh trees (1663)
4198 - return an error from lov_create if all OSCs are inactive (1751)
4199 - fix import levels when a reconnect happens without a timeout (1597)
4200 - exit early from mds_open if we get a lookup error (1749)
4201 - partial page read at EOF wouldn't wait for disk before sending (1642)
4202 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4203 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4204 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4205 - write/truncate lock inversion (1639)
4206 - Don't auto-load obdclass, portals modules during cleanup (1495)
4207 - fix timestamps from jumping to "now" (1763)
4208 - extra journal assertions (1648)
4209 - add an extra multiunlink test (1771)
4210 - fix read_record/write_record API (1776)
4211 - fix leak of offset_extent, possible incorrect i_size later (1772)
4212 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4213 - Debugging for the kqswnal_get_idle_tx problems (1820)
4214 - Allow recovery to be attempted multiple times (1536)
4215 - Write out MDS last_rcvd file after it is first created (1600)
4216 - Fix tx_descriptor leak in failed transmit situations (1827)
4217 - ext3 journaling fixes for assertion failure after IO error (1871)
4218 - class_export_put() on freed export after completion AST error (1896)
4219 - Fix revalidate looping in VFS (1322)
4220 - Don't access a freed export during MDS_REINT timeout (1521)
4221 - Add open-unlink recovery support on the MDS (1673,1764)
4222 - Return an error if no MDS data was read from last_rcvd (1946)
4223 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4224 - Don't setup a disk device that doesn't match exported UUID (317)
4225 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4226 - avoid committing NULL handle in force close
4227 - local.sh is now a one-stripe LOV configuration
4228 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4229 - fix ext3 htree duplicate directory entry corruption (1516)
4230 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4231 - update to new LOV EA format (2097)
4232 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4234 2003-06-15 Phil Schwan <phil@clusterfs.com>
4237 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4238 - per-import/export recovery handling (958, 931, 959)
4239 - multiple last-rcvd slots, for serving multiple FSes (949)
4240 - connections are again shared between multiple imp/exports (963, 964)
4241 - "umount -f" would hang if any requests needed to be sent (393, 978)
4242 - avoid pinning large req buffer by copying for queued messages (989)
4243 - add "uuid" to "lctl device" command to help upcalls (991)
4244 - "open" RPCs with transnos would confuse recovery counters (1037)
4245 - do proper endian conversion of all wire messages (288, 340, 891)
4246 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4247 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4248 - fix ptlrpc_connection leak in target_handle_connect (1174)
4249 - fix import refcounting bug in OST and MDS cleanup (1134)
4250 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4251 - fix very unlikely obd_types race condition (501)
4252 - remove osc_open hack for echo_client (1187)
4253 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4254 - a failure in read_inode2 leads to deadlock (1139)
4255 - cancel ack-locks as soon as transaction is committed (1072)
4256 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4257 - make sure to commitrw after any preprw to avoid deadlock (1162)
4258 - failing to execute a file in a lustre FS would lock inode (1203)
4259 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4260 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4261 - obd_uuid2tgt would walk off the end of the list (1255)
4262 - on IA64 the osc would give portals incorrect bulk size (1258)
4263 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4264 - fix lock inversion caused by new llite matching code (1282)
4265 - limit the number of dirty pages on a client to 10MB (1286)
4266 - timed out locks were not being corrected cancelled (1289)
4267 - fix O_DIRECT above 4GB on IA-32 (1292)
4268 * major user-visible changes
4269 - fail out/fail over policy now controlled by the upcall (993)
4271 - add OBD_PING to check server availability and failure (954)
4272 - lustre messages are now sent in sending host order (288, 340, 891)
4273 - add eadatalen to MDS getattr reply (340)
4274 - OST read replies may contain second buffer, with per-page status (593)
4276 2003-03-11 Phil Schwan <phil@clusterfs.com>
4279 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4280 - failed open()s could cause deadlock; fixed (867, 869)
4281 - stop cancelling OST locks when files are closed (481)
4282 - overlapping XID spaces caused network corruption (851, 853)
4283 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4284 - setattr_raw added, to do single-RPC, server-side setattrs
4285 - lmc/lconf syntax change for OST UUIDs
4286 - fix crashy race condition between ptlrpc_free_req and osc_close
4287 - don't use request in mdc_enqueue if we hit a timeout (889)
4288 - don't set the inode i_size for regular files from the MDS (896)
4289 - handle out of order completion AST (842)
4290 - don't LBUG if a lock request times out after receiving AST (913)
4291 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4292 - if a bad lock AST arrives, send an error instead of dropping entirely
4293 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4294 - fix leak in bulk IO when only partially completed (899, 900, 926)
4295 - fix O_DIRECT for ia64 (55)
4296 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4297 - C-z after timeout could hang a process forever; fixed (977)
4299 - client-side I/O cache (678, 924, 929, 941, 970)
4301 - READPAGE and SETATTRs which don't take server-side locks get
4304 2003-02-11 Phil Schwan <phil@clusterfs.com>
4307 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4308 - Fix fsfilt_extN_readpage() to read a full page of directory
4309 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4310 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4311 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4312 - Most intent operations were removed, in favour of a new RPC mode
4313 that does a single RPC to the server and bypasses most of the VFS
4314 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4315 - Aggressively cancel local locks on DLM servers
4316 - mds_reint_unlink sends EA to the client if it's the last nlink.
4317 client uses that EA to unlink OST objects.
4318 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4319 - recursive symlinks were fixed (439)
4320 - fixed NULL deref in DEBUG_REQ
4321 - filter_update_lastobjid no longer calls sync, which annoyed extN
4322 - fixed multi-client small-writes to a single file problem (445)
4323 - fixed mtime updates during file writes (607)
4324 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4325 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4326 - fixed Linux OST size reporting problem (444, 656)
4327 - OST now updates object mtime with writes or setattr (607, 619)
4328 - client verifies file size before zeroing page past EOF (445)
4329 - OST now writes last allocated objid to disk with allocation (108)
4330 - LOV on echo now works (409)
4332 - mds_reint_unlink sends a new buffer, with the EA included. this
4333 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4334 set if a regular file was being unlinked, and it was the last link
4335 - use PtlGet from the target for bulk writes (315)
4336 - OST now updates object mtime with writes or setattr (607, 619)
4337 - LDLM now has a grant-time callback to revalidate locked items, if
4339 - Many MDS operations were reorganized to combat race conditions
4341 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4342 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4344 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4347 - Fully reactivate OST imports after reconnection (512, others)
4348 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4349 - More graceful error handling for truncating on dead OST (515)
4350 - Don't error out unless we're actually accessing dead stripes (474)
4351 - Fix garbage sizes when stripes are missing (410)
4352 - LRU counters were broken, causing constant lock purge (433, 432)
4353 - garbage on read from stripes with failed OSTs (441)
4354 - mark OSCs as active before reconnecting during recovery (438)
4355 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4356 - lfind did not preserve OST order in output (443)
4357 - symlinks cause hung clients, incorrect data (439)
4358 - stop dereferencing request after dropping refcount (457)
4359 - don't LASSERT(spin_is_locked) on non-SMP (455)
4360 - fixes for many rename() bugs
4361 - fstat didn't correctly synchronize attributes (399)
4362 - server must handle lock cancellation during blocking AST prep (487)
4363 - bulk descriptors were free()d too soon (511)
4364 - fix paths in lconf, which would load incorrect modules (451, 507)
4365 - fix confusing lconf 'host not found' error message (386)
4366 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4367 - fix race condition in mdc_blocking_ast() for inode access (526)
4368 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4369 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4370 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4371 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4372 - fix lstripe in several ways (kernel side) (527)
4373 - fix request leak in ldlm_cli_enqueue (262)
4374 - incorrect OSC was marked inactive after OST failure
4375 - call mds_fs_cleanup before unmounting filesystem (524)
4376 - fix races between taking ns_lock and ldlm_lock_change_resource
4377 - fix races updating LOV export open file list
4378 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4379 - fix recovery NULL deref in ldlm_cli_cancel_unused
4380 - fix some DLM races by using new hash table for lock handles (419)
4381 - permit the client to specify desired inodes, at replay
4382 - duplicate requests when we queue them for replay reintegration
4383 - fix last_rcvd offset calculation
4384 - sync after each recovered transaction, so we always make progress
4385 - never, not always, ERESTART requests without transnos
4386 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4388 - skip replay if the MDS says that the client is already connected
4389 - don't check for a recovery-enabled export to match lctl's UUID
4390 - don't INC_USE_COUNT for phantom exports
4391 - don't crash when cleaning up phantom exports (567)
4392 - don't double-finish or set replay data for errored mdc_open requests
4393 - abort requests when they time out, so we don't get old replies
4394 - send/receive replies for AST messages again
4395 - if the client says that it doesn't have the lock, cancel it on the
4397 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4398 mark it as destroyed, it will all work out when decref is called
4399 - fix module use counts (22, 581)
4401 - ASTs now expect a reply (server cancels lock on error reply)
4403 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4406 - fix many simultaneous client startup (392)
4407 - fix dentry->d_it clobbering
4408 - credentials weren't being shipped for readdir/getattr operations
4409 - remove invalid assertions triggered during some concurrent MD
4411 - proper Lustre versions added (336, 389)
4412 - fix memory leak for create error case (398)
4413 - fix LOV locking bug that would get cli/srv out of sync
4414 - fix echo client over LOV (409)
4415 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4416 - fix double-O_EXCL intent crash (424)
4417 - avoid sending multiple lock CANCELs (352)
4419 - MDS can do multi-client recovery (modulo bugs in new code)
4421 - many updates, edits, cleanups
4423 2002-11-18 Phil Schwan <phil@clusterfs.com>
4426 - fix null d_it dereference (346)
4427 - fix full OST/dbench hang (333)
4428 - fix permission problem with file removal (286)
4429 - fix removal of OSCs from LOV when they fail
4430 - fix NULL deref during bulk timeout (214)
4431 - fix problems related to multiple filesystems on one MDS (241)
4432 - fixed serious subtle metadata locking bugs
4433 - free locks on clients when inodes are removed due to memory
4435 - fix inode pointer in lock data (285)
4436 - partial support for multiple MDS on a single host (241)
4437 - data locks weren't cancelled at clear_inode time (290, 311)
4438 - intent locks could lead to unbounded lock growth (205)
4439 - added a maximum lock count, an LRU list, and a flusher
4440 - fix multiple rename (365)
4441 - properly abstracted the echo client
4442 - OSC locked 1 byte too many; fixed
4443 - rewrote brw callback code:
4444 - fixed recovery bugs related to LOVs (306)
4445 - fixed too-many-pages-in-one-write crash (191)
4446 - fixed (again) crash in sync_io_timeout (214)
4447 - probably fixed callback-related race (385)
4449 - Add capability to MDS protocol
4450 - LDLM cancellations and callbacks on different portals
4452 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4455 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4456 - defer open object destruction to close time (601981, 138)
4457 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4458 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4459 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4460 - don't call obd_setattr in ll_file_release for destroyed objects
4461 * protocol change to lustre_msg: move |version| and add |flags|
4462 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4463 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4464 * added replay of create, unlink, link and rename operations during
4465 MDS failover; recovery should be much more robust now
4466 * remove failed OSCs from LOVs (only lov_create uses this so far)
4467 * the lustre-HOWTO was brought (more) up to date (582544)
4469 2002-10-23 Phil Schwan <phil@clusterfs.com>
4472 - in-use dentries weren't being reused properly (617851)
4473 - prevent multiple LDLM setup (599178)
4474 - fix LOV size calculations for truncate (617853)
4475 - fix client handling of MDS intent errors (POSIX)
4476 - fix permission bug in lovstripe.c test (624321)
4477 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4478 - truncate past end of file could corrupt data
4479 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4480 - a race in recovery could return ETIMEDOUT to apps (623947)
4481 - building outside the source directory was fixed
4482 * the lustre-HOWTO was brought (more) up to date (582544)
4483 * major progress was made on recovery functionality
4485 2002-10-10 Phil Schwan <phil@clusterfs.com>
4488 - recovery deadlock fix
4489 - rm -rf causes LBUG fix (617817)
4490 - file open by multiple tasks fix (618962)
4491 - directory permissions bugs (602707 and 620007)
4492 - journal_stop fixed with locking (611313)
4493 - O_APPEND failures resolved (618273, perhaps 614459)
4494 - lconf PATH fix (619770)
4495 - IA64 build fix (621450)
4496 - RPC buffer sizes scale with amount of memory
4498 2002-10-01 Phil Schwan <phil@clusterfs.com>
4501 - locks would be cancelled without throwing away data pages,
4502 resulting in inconsistent data (605627)
4503 - inode attributes were not always being refreshed (605627, 612449)
4504 - lconf now continues to cleanup after lctl reports an error
4505 - MDS now enforces user permissions (602707)
4506 - lprocfs cleanup fixed, but not yet enabled (614157)
4507 - fixed infinite server hang, should a client not respond to an AST
4508 - avoid going into recovery if user calls readlink() with a buffer
4509 that's too small (613941)
4510 - AST RPCs no longer require replies (614867) -- this may be changed
4511 - don't crash server if client sends an IOV that's too big (611336)
4512 - fixed lock conversion deadlock (611892)
4513 - fixed the following of symlinks (614622)
4514 * recovery: the server can remove locks from a client that dies, other
4515 clients can make progress
4516 * more extN patch fixes
4517 * compile-time configurable ptlrpc buffer allocations
4519 - collaborative read cache document
4520 - Lustre Lite Performance CDR document-in-progress
4522 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4525 - fix typo in patch-2.4.18
4527 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4530 - clear ptlrpc request each time in handle_incoming_request()
4531 - unlink of files now destroys the object on the OST
4533 2002-09-19 Peter Braam <braam@clusterfs.com>
4535 * add hard link support
4536 * change obdfile creation method
4537 * kernel patch changed
4539 2002-09-19 Peter Braam <braam@clusterfs.com>
4542 - stack overflow bug in extN fixed
4544 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4546 * documentation updates
4547 - add man pages for config tools
4548 - update tests/README to describe testing with new config tools
4549 - finish metadata API descriptions
4550 * bug fixes and cleanups
4551 - statfs workaround for 16TB limit
4552 - LOV stripe allocation improved, can stripe on subset of OSTs
4553 - LOV file size/IO offset was wrong for files > 4GB in size
4554 - object EA data was being dropped, caused files to be unreadable
4555 - memory overflow with non-LOV OST caused memory corruption
4556 - fixed regression tests to work with new config tools, obdfilter
4557 - fixed bug when directory size became larger than 1 block
4558 - fixed bug (for single client case) when PWD was deleted
4559 - invalidate local directory pages when doing intent-based ops
4560 - avoid LDLM oops when lock callback contained bad data
4562 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4564 * documentation updates
4565 * bug fixes and cleanups
4566 - configuration tools
4569 - 64-bit compile warnings
4570 - 64-bit internal statfs data
4572 * test_brw on persistent OST devices
4574 * lprocfs (disabled)
4576 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4578 * documentation updates
4579 * bug fixes and cleanups
4580 * configuration tools
4582 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4585 * many small fixes to 0_5_4
4586 * io/network handling
4587 * thinkos in MDS operations
4589 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4592 * crucial basic fixes to 0.5.3
4593 * IOR, Iozone work over Elan
4596 2002-08-07 Phil Schwan <phil@clusterfs.com>
4597 * version 0_5_3, our first alpha
4598 * we use the new Portals iovs
4599 * documentation updates
4600 * bug fixes and cleanups
4601 * small changes in the DLM wire protocol
4603 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4604 * version 0_5_1 with some initial stability,
4605 * locking on MD and file I/O.
4606 * documentation updates
4607 * several bug fixes since 0.5.0
4608 * small changes in wire protocol
4610 2002-07-18 Phil Schwan <phil@clusterfs.com>
4612 * delivered as Lustre Light Alpha
4613 * fixed a crash after handling invalid MDS requests
4614 * fixed directory pages for architectures with non-4k pages sizes
4616 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4617 * release version v0_4_4
4618 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4620 * Fixes a number of interruption problems with OST operations.
4621 * Update documentation for portals header changes
4622 * Move all wire protocol structs/defines to lustre_idl.h
4623 * Fixes symlink length bug.
4624 * Add tcpdump to repository.
4626 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4627 * release version v0_4_3
4628 * Fixes statfs for inodes on extN.
4629 * Fixes bug in runtests which would delete /etc/hosts.
4630 * Use 64-bit object IDs wherever possible (not into VFS though)
4631 Remove ost_get_info, which is unused by lustre, and out of date.
4633 2002-07-03 Peter Braam <braam@clusterfs.com>
4634 * release version v0_4_2 Fixes a lookup error (type not passed)
4635 * move forward to head of Portals
4636 * move forward to latest Lustre kernel
4638 2002-06-25 Peter Braam <braam@clusterfs.com>
4639 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob