1 tbd Sun Microsystems, Inc.
5 2.6.9-67.0.22.EL (RHEL 4),
6 2.6.16.60-0.33 (SLES 10),
7 2.6.18-92.1.22.el5 (RHEL 5),
8 2.6.22.14 vanilla (kernel.org)
9 * Client support for unpatched kernels:
10 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
11 2.6.16 - 2.6.22 vanilla (kernel.org)
12 * Client support for unpatched kernels:
13 we do not recommend using patchless RHEL4 clients with kernels
14 prior to 2.6.9-55EL (RHEL4U5).
15 * Recommended e2fsprogs version: 1.40.11-sun1
16 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
17 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
18 removed cwd "./" (refer to Bugzilla 14399).
19 * A new quota file format has been introduced in 1.6.5.
20 The format conversion from prior releases is handled transparently,
21 but releases older than 1.4.12/1.6.5 don't understand this new
22 format. The automatic format conversion can be avoided by running
23 the following command on the MDS:
24 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
25 For more information, please refer to bugzilla 13904.
26 * Output of lfs quota has been made less detailed by default,
27 old (verbose) output can be obtained by using -v option.
28 * File join has been disabled in this release, refer to Bugzilla 16929.
29 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
30 * NFS export disabled when stack size < 8192. Since the NFSv4 export
31 of Lustre filesystem with 4K stack may cause a stack overflow. For
32 more information, please refer to bugzilla 17630.
35 Frequency : rare, connect and disconnect target at same time
37 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
38 Details : don't call obd_disconnect under lov_lock. this long time
39 operation and can block ptlrpcd which answer to connect request.
41 Severity : enhancement
43 Description: Update to SLES10 kernel-2.6.16.60-0.33.
45 Severity : enhancement
47 Description: Update to sles9 kernel-2.6.5-7.315.
50 Frequency : start MDS on uncleanly shutdowned MDS device
52 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
53 Details : stay in waiting mds<>ost recovery finished produce random bugs
54 due race between two ll_sync thread for one lov target. send
55 ACTIVATE event only if connect realy finished and import have
58 Severity : enhancement
60 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
63 Frequency : start MDS on uncleanly shutdowned MDS device
65 Descriptoin: aborting recovery hang on MDS
66 Details : don't throttle destroy RPCs for the MDT.
68 Severity : enhancement
70 Descriptoin: Don't sync journal after every i/o
71 Details : Implement write RPC replay to allow server replies for write RPCs
72 before data is on disk.
76 Description: Slow reads beyond 8Tb offsets.
77 Details : Page index integer overflow in ll_read_ahead_page
81 Description: Soft lockup on OSS after MDS failover
82 Details : MSG_CONNECT_INITIAL is not set on the initial MDS->OST connect.
83 As a conseqence, the patch from bug 18224 is not operational
84 and the MDS export cannot be reused on the OSTs until it gets
88 Frequency : rare, only if using MMP with Linux RAID
90 Description: MMP doesn't work with Linux RAID
91 Details : While using HA for Lustre servers with Linux RAID, it is possible
92 that MMP will not detect multiple mounts. To make this work we
93 need to unplug the device queue in RAID when the MMP block is being
94 written. Also while reading the MMP block, we should read it from
95 disk and not the cached one.
97 -------------------------------------------------------------------------------
98 2008-12-31 Sun Microsystems, Inc.
100 * Support for kernels:
101 2.6.5-7.314 (SLES 9),
102 2.6.9-67.0.22.EL (RHEL 4),
103 2.6.16.60-0.31 (SLES 10),
104 2.6.18-92.1.17.el5 (RHEL 5),
105 2.6.22.14 vanilla (kernel.org)
106 * Client support for unpatched kernels:
107 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
108 2.6.16 - 2.6.22 vanilla (kernel.org)
109 * Client support for unpatched kernels:
110 we do not recommend using patchless RHEL4 clients with kernels
111 prior to 2.6.9-55EL (RHEL4U5).
112 * Recommended e2fsprogs version: 1.40.11-sun1
113 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
114 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
115 removed cwd "./" (refer to Bugzilla 14399).
116 * A new quota file format has been introduced in 1.6.5.
117 The format conversion from prior releases is handled transparently,
118 but releases older than 1.4.12/1.6.5 don't understand this new
119 format. The automatic format conversion can be avoided by running
120 the following command on the MDS:
121 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
122 For more information, please refer to bugzilla 13904.
123 * A new quota file format was introduced in 1.6.6/1.8.0 (kernels 2.6.16+).
124 The format conversion from prior releases is handled transparently,
125 but releases older than 1.6.6/1.8.0 don't understand this new
126 format. The automatic format conversion can be avoided by running
127 the following commands on the MDS and OSS servers (for
128 pre 1.4.12-1.6.5 quota files):
129 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV',
130 'tunefs.lustre --param="ost.quota_type=ug1" $MDTDEV'
131 or (for 1.4.12/1.6.5 quota files)
132 'tunefs.lustre --param="mdt.quota_type=ug2" $MDTDEV',
133 'tunefs.lustre --param="ost.quota_type=ug2" $MDTDEV'
134 For more information, please refer to bugzilla 13904.
135 * Output of lfs quota has been made less detailed by default,
136 old (verbose) output can be obtained by using -v option.
137 * File join has been disabled in this release, refer to Bugzilla 16929.
138 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
139 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
140 Lustre filesystem with 4K stack may cause a stack overflow. For more
141 information, please refer to bugzilla 17630.
146 Descriptoin: don't lose wakeup for imp_recovery_waitq
147 Details : recover_import_no_retry or invalidate_import and import_close can
148 both sleep on imp_recovery_waitq, but we was send only one wakeup
154 Descriptoin: panic in mds_open
155 Details : don't confuse mds_finish_transno() with PTR_ERR(-ENOENT)
160 Descriptoin: stuck in cache_remove_extent() or panic with accessing to already
162 Details : release lock refernce only after add page to pages list.
165 Frequency : always with long access acl
167 Descriptoin: mds can't pack reply with long acl.
168 Details : mds don't control size of acl but they limited by reint/getattr
173 Frequency : on remount
175 Description: external journal device not working after the remount
176 Details : clear dev_rdonly flag for external journal devices in
182 Description: shutdown vs evict race
183 Details : client_disconnect_export vs connect request race.
184 if client will evicted at this time - we start invalidate
185 thread without referece to import and import can be freed
191 Description: shrink LOV EAs before replying
192 Details : correctly adjust LOV EA buffer for reply.
197 Description: don't skip ost target if they assigned to file
198 Details : Drop slow OSCs if we can, but not for requested start idx.
199 This means "if OSC is slow and it is not the requested
200 start OST, then it can be skipped, otherwise skip it only
201 if it is inactive/recovering/out-of-space.
203 Severity : enhancement
205 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
208 Frequency : rare, need acl's on inode.
210 Description: client can't handle ost additional correctly
211 Details : if ost was added after client connected to mds client can have
212 hit lnet_try_match_md ... to big messages to wide striped files.
213 in this case need teach client to handle config events about add
214 lov target and update client max ea size at that event.
216 Severity : enhancement
218 Description: Update to sles9 kernel-2.6.5-7.314.
220 Severity : enhancement
222 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
225 Frequency : Create a symlink file with a very long name
227 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
228 Details : If there is no extra space in the request for early cancels,
229 ldlm_req_handles_avail() returns 0 instead of a negative value.
234 Description: mds is deadlocked
235 Details : in rare cases, inode in catalog can have i_no less than have parent
236 i_no, this produce wrong order for locking during open, and
237 parallel unlink can be lock open. this need teach mds_open to grab
238 locks in resource id order, not at parent -> child order.
240 Severity : enhancement
242 Description: Add /proc entry for import status
243 Details : The mdc, osc, and mgc import directories now have
244 an import directory that contains useful import data for debugging
247 Severity : enhancement
249 Description: Re-disable certain /proc logging
250 Details : Enable and disable client's offset_stats, extents_stats and
251 extents_stats_per_process stats logging on the fly.
254 Frequency : Only on FC kernels 2.6.22+
256 Description: oops in statahead
257 Details : Do not drop reference count for the dentry from VFS when lookup,
258 VFS will do that by itself.
260 Severity : enhancement
262 Description: Generic /proc file permissions
263 Details : Set /Proc file permissions in a more generic way to enable non-
264 root users operate on some /proc files.
268 Description: Hitting mdc_commit_close() ASSERTION
269 Details : Properly handle request reference release in
270 ll_release_openhandle().
274 Frequency : only patchless client
275 Description: add workaround for race between add/remove dentry from hash
277 Severity : enhancement
279 Description: Allow OST glimpses to return PW locks
283 Description: LBUG when llog conf file is full
284 Details : When llog bitmap is full, ENOSPC should be returned for plain
289 Description: Prevent import from entering FULL state when server in recovery
293 Description: service mount cannot take device name with ":"
294 Details : Only when device name contains ":/" will mount treat it as
300 Description: replace ptlrpcd with the statahead thread to interpret the async
301 statahead RPC callback
305 Frequency : on recovery
306 Description: I/O failures after umount during fail back
307 Details : if client reconnected to restarted server we need join to recovery
308 instead of find server handler is changed and process self eviction
309 with cancel all locks.
311 Severity : enhancement
313 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
317 Description: Kernel BUG tries to release flock
318 Details : Lustre does not destroy flock lock before last reference goes
319 away. So always drop flock locks when client is evicted and
320 perform unlock regardless of successfulness of speaking to MDS.
322 Severity : enhancement
324 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
326 Severity : enhancement
328 Description: Upcall on Lustre log has been dumped
329 Details : Allow for a user mode script to be called once a Lustre log has
330 been dumped. It passes the filename of the dumped log to the
331 script, the location of the script can be specified via
332 /proc/sys/lnet/debug_log_upcall.
337 Description: avoid messages about idr_remove called for id that is not allocated
338 Details : Move assigment s_dev for clustered nfs to end of initialization,
339 for avoid problem with error handling.
344 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
345 Details : When connection is reused this not moved from CONN_UNUSED_HASH into
346 CONN_USED_HASH and this prodice warning when put connection again
352 Description: avoid ASSERTION(client_stat->nid_exp_ref_count == 0) failed
353 Details : release reference to stats when client disconnected, not
354 when export destroyed for avoid races when client destroyed
355 after main ost export.
359 Description: more cleanup in mds_lov
360 Details : add workaround for get valid ost count for avoid warnings about
361 drop too big messages, not init llog cat under semphore which
362 can be blocked on reconnect and break normal replay, fix access
365 Severity : enhancement
367 Description: File striping can now be set to use an arbitrary pool of OSTs.
369 Severity : enhancement
371 Description: Export bytes_read/bytes_write count on OSC/OST.
375 Description: Early reply size mismatch, MGC loses connection
376 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
377 the connect flags are properly negotiated.
381 Description: Properly propagate oinfo flags from lov to osc for statfs
382 Details : restore missing copy oi_flags to lov requests.
386 Description: exports in /proc are broken
387 Details : recreate /proc entries for clients when they reconnect.
389 Severity : enhancement
391 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
392 : l_getgroups(8), lst(8), routerstat(8)
393 Details : included man pages for llobdstat(8), llstat(8),
394 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
396 Severity : enhancement
398 Description: Implement lustre ll_show_options method.
400 Severity : enhancement
402 Description: Update to SLES9 kernel-2.6.5-7.312.
404 Severity : enhancement
406 Description: Update to RHEL4 kernel-2.6.9-67.0.22.EL.
410 Description: exports in /proc are broken
411 Details : recreate /proc entries for clients when they reconnect.
415 Description: don't fail open with -ERANGE
416 Details : if client connected until mds will be know about real ost count
417 get LOV EA can be fail because mds not allocate enougth buffer
422 Description: Resolve device initialization race
423 Details : Prevent proc handler from accessing devices added to the
424 obd_devs array but yet be intialized.
426 Severity : enhancement
428 Description: configure's --enable-quota should check the
429 : kernel .config for CONFIG_QUOTA
430 Details : configure is terminated if --enable-quota is passed but
431 : no quota support is in kernel
433 Severity : enhancement
435 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
437 Severity : enhancement
439 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
443 Frequency : rare, on PPC clients
444 Description: don't swab ost objects in response about directory, because
446 Details : bug similar bug 14856, but in different function.
448 Severity : enhancement
450 Description: lfs quota tool enhancement
451 Details : added units specifiers support for setquota, default to
452 current uid/gid for quota report, short quota stats by
453 default, nonpositional parameters for setquota, added
454 llapi_quotactl manual page.
456 Severity : enhancement
458 Description: *optional* service tags registration
459 Details : if the "service tags" package is installed on a Lustre node
460 When the filesystem is mounted, a local-node service tag will
461 be created. See http://inventory.sun.com/ for more information
462 about the Service Tags asset management system.
466 Description: Client runs out of low memory
467 Details : Consider only lowmem when counting initial number of llap pages
470 Description: Kernel BUG tries to release flock
471 Details : Lustre does not destroy flock lock before last reference goes
472 away. So always drop flock locks when client is evicted and
473 perform unlock regardless of successfulness of speaking to MDS.
476 Frequency : occasional
478 Description: add refcount for osc callbacks, so avoid panic on shutdown
480 Severity : enhancement
482 Description: Update to RHEL4 kernel-2.6.9-67.0.20.
485 Frequency : testing only
487 Description: sanity test 65a fails if stripecount of -1 is set
488 Details : handle -1 striping on filesystem in ll_dirstripe_verify
491 Frequency : only in unusual configurations
493 Description: Kernel panic with find ost index.
494 Details : lov_obd have panic if some OST's have sparse indexes.
496 Severity : enhancement
498 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
501 Frequency : rarely, if filesystem is mounted with -o flock
503 Description: do not process already freed flock
504 Details : flock can possibly be freed by another thread before it reaches
505 to ldlm_flock_completion_ast.
508 Frequency : rarely, if filesystem is mounted with -o flock
510 Description: LBUG during stress test
511 Details : Need properly lock accesses the flock deadlock detection list.
514 Frequency : rarely, if binaries are being run from Lustre
516 Description: oops in page fault handler
517 Details : kernel page fault handler can return two special 'pages' in
518 error case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
521 Frequency : rarely, during shutdown
523 Description: timeout with invalidate import.
524 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
525 should be handled by ptlrpcd. This produce long age waiting and
526 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
532 Description: ASSERTION(CheckWriteback(page,cmd)) failed
533 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
537 Frequency : only with broken builds/installations
539 Description: no LBUG if lquota.ko and fsfilt_ldiskfs.ko are different versions
540 Details : just return an error to a user, put a console error message
542 Severity : enhancement
544 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
546 Severity : enhancement
548 Description: Update to RHEL4 kernel-2.6.9-67.0.15.
550 Severity : enhancement
552 Description: enable MGS and MDT services start separately
553 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
554 a co-located MGS without starting the MGS, which is a complement
555 to 'nosvc' mount option.
558 Frequency : always, on big-endian systems
560 Description: cleanup in ptlrpc code, related to PPC platform
561 Details : store magic in native order avoid panic's in recovery on PPC
562 node and forbid from this error in future. Also fix posibily
563 of twice swab data. Fix get lov striping to userpace.
566 Frequency : rarely, if replay get lost on server
568 Description: server incorrectly drop resent replays lead to recovery failure.
569 Details : do not drop replay according to msg flags, instead we check the
570 per-export recovery request queue for duplication of transno.
573 Frequency : after recovery
575 Description: precreate to many object's after del orphan.
576 Details : del orphan st in oscc last_id == next_id and this triger growing
577 count of precreated objects. Set flag LOW to skip increase count
578 of precreated objects.
581 Frequency : rare, on clear nid stats
583 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
584 Details : when clean nid stats sometimes try destroy live entry,
585 and this produce panic in free.
588 Frequency : occasionally since 1.6.4
590 Description: Stack overflow during MDS log replay
591 Details : ease stack pressure by using a thread dealing llog_process.
594 Frequency : very rare
596 Description: MDT cannot be unmounted, reporting "Mount still busy"
597 Details : Mountpoint references were being leaked during open reply
598 reconstruction after an MDS restart. Drop mountpoint reference
599 in reconstruct_open() and free dentry reference also.
604 Description: wait until IO finished before start new when do lock cancel.
605 Details : VM protocol want old IO finished before start new, in this case
606 need wait until PG_writeback is cleared until check dirty flag
607 and call writepages in lock cancel callback.
612 Description: mds_mfd_close() ASSERTION(rc == 0)
613 Details : In mds_mfd_close(), we need protect inode's writecount change
614 within its orphan write semaphore to prevent possible races.
617 Frequency : rare, on shutdown ost
619 Description: don't hit live lock with umount ost.
620 Details : shrink_dcache_parent can be in long loop with destroy dentries,
621 use shrink_dcache_sb instead.
624 Frequency : only when echo_client is used
626 Description: don't panic with use echo_client
627 Details : echo client pass NULL as client nid pointer and this produce NULL
631 Frequency : Always on 32-bit PowerPC systems
633 Description: fix build on PPC32
634 Details : compile code with -m64 flag produce wrong object file for PPC32.
639 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
640 Details : In reconstruct_* functions, LASSERTs on both the data supplied
641 by a client, and the data on disk are dangerous and incorrect.
642 Change them with client eviction.
644 Severity : enhancement
646 Description: skiplist implementation simplification
647 Details : skiplists are used to group compatible locks on granted list
648 that was implemented as tracking first and last lock of each lock
649 group the patch changes that to using doubly linked lists
653 Description: delete compatibility for 32bit qdata
654 Details : as planned, when lustre is beyond b1_8, lquota won't support 32bit
655 qunit. That means servers of b1_4 and servers of b1_8 can't be
656 used together if users want to use quota.
659 Frequency : only with administrator action
661 Description: mount failure if config log has invalid conf_param setting
662 Details : If administrator specified an incorrect configuration parameter
663 with "lctl conf_param" this would cause an error during future
664 client mounts. Instead, ignore the bad configuration parameter.
667 Frequency : blocks per group < blocksize*8 and uninit_groups is enabled
669 Description: ldiskfs error: XXX blocks in bitmap, YYY in gd
670 Details : If blocks per group is less than blocksize*8, set rest of the
674 Frequency : Application do stride read on lustre
676 Description: The read performance will drop a lot if the application does
678 Details : Because the stride_start_offset are missing in stride read-ahead,
679 it will cause clients read a lot of unused pages in read-ahead,
680 then the read-performance drops.
684 Description: more ldlm soft lockups
685 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
686 starve other threads from the resource lock for a long time in
687 case of long waiting queue, so change the debug level from
688 D_OTHER to the less frequently used D_INFO.
690 Severity : enhancement
692 Description: add -gid, -group, -uid, -user options to lfs find
694 Severity : enhancement
696 Description: ll_recover_lost_found_objs - recover objects in lost+found
697 Details : OST corruption and subsequent e2fsck can leave objects in the
698 lost+found directory. Using the "ll_recover_lost_found_objs"
699 tool, these objects can be retrieved and data can be salvaged
700 by using the object ID saved in the fid EA on each object.
705 Description: this bug _only_ happens when inode quota limitation is very low
706 (less than 12), so that inode quota unit is 1 at initialization.
707 Details : if remaining quota equates 1, it is a sign to demonstate that quota
708 is effective now. So least quota qunit should be 2.
712 Description: Hung threads in invalidate_inode_pages2_range
713 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
714 one is completed. As a result, some RPCs are stuck in the queue
719 Description: Procfs and llog threads access destoryed import sometimes.
720 Details : Sync the import destoryed process with procfs and llog threads by
721 the import refcount and semaphore.
725 Description: mds fails to respond, threads stuck in ldlm_completion_ast
726 Details : Sort source/child resource pair after updating child resource.
731 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
732 Details : If insertion of an extent fails, then discard the inode
733 preallocation and free data blocks else it can lead to duplicate
738 Description: don't always update ctime in ext3_xattr_set_handle()
739 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle()
740 In some cases the ctime should not be updated, for example for
741 2.0->1.8 compatibility it is necessary to delete an xattr and it
742 should not update the ctime.
746 Description: add quota statistics
747 Details : 1. sort out quota proc entries and proc code.
748 2. add quota statistics
753 Description: quotas are not honored with O_DIRECT
754 Details : all writes with the flag O_DIRECT will use grants which leads to
755 this problem. Now using OBD_BRW_SYNC to guard this.
759 Bugzilla : 15713/16362
760 Description: Assertion in iopen_connect_dentry in 1.6.3
761 Details : looking up an inode via iopen with the wrong generation number can
762 populate the dcache with a disconneced dentry while the inode
763 number is in the process of being reallocated. This causes an
764 assertion failure in iopen since the inode's dentry list contains
765 both a connected and disconnected dentry.
769 Description: assertion failure in ldlm_handle2lock()
770 Details : fix a race between class_handle_unhash() and class_handle2object()
771 introduced in lustre 1.6.5 by bug 13622.
773 Severity : enhancement
775 Description: superblock lock contention with many SMP cores on one client
776 Details : several client filesystem locks were highly contended on SMP
777 NUMA systems with 8 or more cores. Per-CPU datastructures
778 and more efficient locking implemented to reduce contention.
783 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
784 Details : remove the limit of 256 scsi disks in the sd_iostat patch
789 Description: oops in sd_iostats_seq_show()
790 Details : unloading/reloading the scsi low level driver triggers a kernel
791 bug when trying to access the sd iostat file.
796 Description: Kernel panics during QLogic driver reload
797 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
798 patch, causing memory corruption.
803 Description: journal_dev option does not work in b1_6
804 Details : pass mount option during pre-mount.
806 Severity : enhancement
808 Description: Add a FIEMAP(FIle Extent MAP) ioctl for ldiskfs
809 Details : FIEMAP ioctl will allow an application to efficiently fetch the
810 extent information of a file. It can be used to map logical blocks
811 in a file to physical blocks in the block device.
814 Frequency : only with adaptive timeout enabled
816 Description: DEBUG_REQ() bad paging request
817 Details : ptlrpc_at_recv_early_reply() should not modify req->rq_repmsg
818 because it can be accessed by reply_in_callback() without the
822 Frequency : only on Cray X2
824 Description: X2 build failures
825 Details : fix build failures on Cray X2.
829 Description: xid & resent requests
830 Details : Initialize RPC XID from clock at startup (randomly if clock is
835 Description: quota recovery deadlock during mds failover
836 Details : This patch includes att18982, att18236, att18237 in bz14840.
838 1. fix osts hang when mds does failover with quotaon
839 2. prevent watchdog storm when osts threads wait for the
844 Description: kernel panic on racer
845 Details : Do not access dchild->d_inode when IS_ERR(dchild) is true.
847 Severity : enhancement
849 Description: Add lustre_start utility to start or stop multiple Lustre servers
854 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
855 Details : In case of memory pressure, list_del() can be called twice on
856 req->rq_history_list, causing a kernel oops.
860 Description: (ptllnd_peer.c:557:kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
861 Details : fix stack overflow in the distributed lock manager by defering
862 export eviction after a failed ast to the elt thread instead of
863 handling it in the dlm interpret routine.
865 Severity : enhancement
867 Description: More exported tunables for mballoc
868 Details : Add support for tunable preallocation window and new tunables for
873 Description: Detect corruption of block bitmap and checking for preallocations
874 Details : Checks validity of on-disk block bitmap. Also it does better
875 checking of number of applied preallocations. When corruption is
876 found, it turns filesystem readonly to prevent further corruptions.
880 Frequency : only for big-endian servers
881 Description: Check for big-endian system while mounting fs with extents feature
882 Details : Mounting a filesystem with extents feature will fail on big-endian
883 systems since ext3-based ldiskfs is not supported on big-endian
884 systems. Can be over-ridden with "bigendian_extents" mount option.
888 Description: Excessive recovery window
889 Details : With AT enabled, the recovery window can be excessively long (6000+
890 seconds). To address this problem, we no longer use
891 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
892 timeout no longer depends on the service time, it is set to
893 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
894 time via pb_service_time.
898 Description: Watchdog triggered on MDS failover
899 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
900 the OSTs know that the MDS "UUID" can be reused for the same export
901 from a different NID, so we do not need to wait for the export to
904 -------------------------------------------------------------------------------
906 2008-05-26 Sun Microsystems, Inc.
908 * Support for kernels:
909 2.6.5-7.311 (SLES 9),
910 2.6.9-67.0.7.EL (RHEL 4),
911 2.6.16.54-0.2.5 (SLES 10),
912 2.6.18-53.1.14.el5 (RHEL 5),
913 2.6.22.14 vanilla (kernel.org)
914 * Client support for unpatched kernels:
915 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
916 2.6.16 - 2.6.22 vanilla (kernel.org)
917 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
918 we do not recommend using patchless RHEL4 clients with kernels
919 prior to 2.6.9-55EL (RHEL4U5).
920 * Recommended e2fsprogs version: 1.40.7-sun1
921 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
922 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
923 removed cwd "./" (refer to Bugzilla 14399).
924 * A new quota file format has been introduced in 1.6.5.
925 The format conversion from prior releases is handled transparently,
926 but releases older than 1.4.12/1.6.5 will not understand this new
927 format. The automatic format conversion can be avoided by running
928 the following command on the MDS before upgrading:
929 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
930 For more information, please refer to bugzilla 13904.
934 Description: quota performance fix
935 Details : quota data is written in journalled mode instead of ordered to
940 Description: lfs support for human-readable quota grace time strings
941 Details : lfs setquota -t and lfs quota -t represent quota grace times
942 in "XXwXXdXXhXXmXXs" format instead of large values in seconds
945 Frequency : always with o2ib 1.3 and sles10
947 Description: fix build with SLES10 and o2ib v3.
948 Details : sles10 uses diffrent name for Module.symver file but configure
949 assume this file has same name on RHEL/SLES/vanila kernels.
952 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
954 Description: MDS may lose file striping (and hence file data) in some cases
955 Details : If there are additional extended attributes stored on the MDS,
956 in particular ACLs, SELinux, or user attributes (if user_xattr
957 is specified for the client mount options) then there is a risk
958 of attribute loss. Additionally, the Lustre file striping
959 needs to be larger than default (e.g. striped over all OSTs),
960 and an additional attribute must be stored initially in the
961 inode and then increase in size enough to be moved to the
962 external attribute block (e.g. ACL growing in size) for file
965 Severity : enhancement
967 Description: add message levels for liblustreapi
972 Description: MDT cannot be unmounted, reporting "Mount still busy"
973 Details : Mountpoint references were being leaked during open reply
974 reconstruction after an MDS restart. Drop mountpoint reference
975 in reconstruct_open() and free dentry reference also.
980 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
981 Details : Move the 'good_osts' check before the 'total_bavail' check. This
982 will result in an -EAGAIN and in the exit call path we call
983 alloc_rr() which will with increasing aggressiveness attempt to
984 aquire precreated objects on the minimum number of required OSCs.
988 Description: Use old size assignment to avoid deadlock
989 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
990 the scheduling while holding a spinlock. We do not need locking
991 for size in ll_update_inode() because size is only updated from
992 the MDS for directories or files without objects, so there is no
993 other place to do the update, and concurrent access to such inodes
994 are protected by the inode lock.
998 Description: Use __u64 instead of int for valid bits
1002 Description: resolve "_IOWR redefined" build error on SLES10
1006 Description: dump the memory debugging after all modules are unloaded to
1007 suppress false negative in conf_sanity test 39
1011 Description: the recovery timer never expires
1012 Details : for new client connect request, the recovery timer should not be
1013 reset, otherwise recovery timer will never expired, if the old
1014 client never come. Only old client connect and first connection
1015 req should trigger recovery timer reset.
1019 Description: the min numbers of lproc stats are wrong
1020 Details : adding a new constant LC_MIN_INIT and use it for initialization
1024 Frequency : always with interactive lfs
1026 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1029 Frequency : with multiple concurrent readdir processes in same directory
1030 Bugzilla : 15406, 15169, 15175
1031 Description: misc fixes for directory readahead.
1032 Details : prevent previous statahead async RPC callback from processing the
1033 current "statahead_info", race condition between async RPC callback
1034 add dentry into dentry hash table and "ls" thread revalidate such
1035 dentry, statahead his/miss control for hidden items, and so on.
1037 Severity : enhancement
1039 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1043 Description: incore types cleaning in quota code (with respect to 64-bit limits)
1044 Details : several u32 variables declarations are replaced with u64 declarations
1049 Description: fix SLES kernel versioning
1050 Details : the kernel version for our SLES 10 kernel did not include a "-"
1051 before the "smp" at the end. while this was not a problem in
1052 general, it did mean that software trying to use the kernel
1053 version to try to detect a vendor specific kernel would fail.
1054 this was most evident by the OFED build scripts.
1059 Description: Don't update lov_desc members until making sure they are valid
1060 Details : When updating lov_desc members via proc fs, need fix their
1061 validities before doing the real update.
1064 Frequency : very rare
1066 Description: don't put request into delay list while invalidate in flight.
1067 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1068 import in flight. this produce timeout for invalidate and sometimes
1069 can cause stale data.
1071 Severity : enhancement
1073 Description: Update kernel to SLES9 2.6.5-7.311.
1075 Severity : enhancement
1077 Description: Update kernel to RHEL4 2.6.9-67.0.7.
1082 Frequency : on PPC only
1083 Description: not convert ost objects for directory because it's not exist.
1084 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1086 Severity : enhancement
1088 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
1089 __u64 as usingied long long type.
1092 Frequency : rare, on shutdown
1094 Description: race process ast vs remove callback
1095 Details : removing callback before disconnect import open race with
1096 processing callback.
1098 Severity : enhancement
1100 Description: Update kernel to SLES9 2.6.5-7.311.
1102 Severity : enhancement
1104 Description: Files open for execute are not marked busy on SLES10
1105 Details : Add FMODE_EXEC to SLES10 SP1 server kernel series.
1107 Severity : enhancement
1109 Description: Add server support for vanilla-2.6.22.14.
1112 Frequency : occasional
1114 Description: Avoid lov_create() getting stuck in obd_statfs_rqset()
1115 Details : If an OST is down the MDS will hang indefinitely in
1116 obd_statfs_rqset() waiting for the statfs data. While for
1117 MDS QOS usage of statfs, it should not stuck in waiting.
1119 Severity : enhancement
1121 Description: Disable adaptive timeouts by default
1124 Frequency : on network error
1126 Description: panic with double free request if network error
1127 Details : mdc_finish_enqueue is finish request if any network error occuring,
1128 but it's true only for synchronus enqueue, for async enqueue
1129 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1133 Frequency : rare, on recovery
1135 Description: read procfs can produce deadlock in some situation
1136 Details : Holding lprocfs lock with send rpc can produce block for destroy
1137 obd objects and this also block reconnect with -EALREADY.
1138 This isn't fix all lprocfs bugs - but make it rare.
1140 Severity : enhancement
1142 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1145 Frequency : frequent on X2 node
1147 Description: mdc_set_open_replay_data LBUG
1148 Details : Set replay data for requests that are eligible for replay.
1153 Description: lustre_mgs: operation 101 on unconnected MGS
1154 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1155 MGC, and late on MGC cannot successfully connect to MGS and a lot
1156 of the error messages complaining that MGS is not connected.
1159 Frequency : rare, depends on device drivers and load
1161 Description: MDS or OSS nodes crash due to stack overflow
1162 Details : Code changes in 1.6.4 increased the stack usage of some functions.
1163 In some cases, in conjunction with device drivers that use a lot
1164 of stack, the MDS (or possibly OSS) service threads could overflow
1165 the stack. One change which was identified to consume additional
1166 stack has been reworked to avoid the extra stack usage.
1168 Severity : enhancement
1170 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1172 Severity : enhancement
1174 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1176 Severity : enhancement
1178 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1180 Serverity : enhancement
1182 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1185 Frequency : rare on shutdown OST
1187 Description: Don't allow skipping OSTs if index has been specified.
1188 Details : Don't allow skipping OSTs if index has been specified, make locking
1189 in internal create lots better.
1194 Description: ASSERTION(!PageDirty(page)) failed
1195 Details : Wrong check could lead to an assertion failure under specific
1201 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1202 Details : access to bitfield in structure is always rounded to long
1203 and this produce problem with not atomic change any bit.
1208 Description: Lustre make rpms failed.
1209 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1210 builds Lustre rpms from tarball.
1212 Severity : enhancement
1214 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
1217 Frequency : rare on shutdown OST
1219 Description: If llog cancel was not send before clean_exports phase, this can
1220 produce deadlock in llog code.
1221 Details : If llog thread has last reference to obd and call class_import_put
1222 this produce deadlock because llog_cleanup_commit_master wait when
1223 last llog_commit_thread exited, but this never success because was
1224 called from llog_commit_thread.
1227 Frequency : only if OST index is skipped
1229 Description: NULL lov_tgts causing MDS oops
1230 Details : more safe checks for NULL lov_tgts for avoid oops.
1232 Severity : enhancement
1234 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1236 Severity : enhancement
1238 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1243 Description: make mgs_setparam() handle fsname containing dash
1244 Details : fsname containing a dash does not work with lctl conf_param
1246 Severity : enhancement
1248 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
1251 Frequency : rare, in recovery and (or) destroy lovobjid file.
1253 Description: rewrite lov objid code.
1254 Details : Cleanup for lov objid code, remove scability problems and wrong
1255 locking. Fix sending last_id into ost.
1257 Severity : enhancement
1259 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1261 Severity : enhancement
1263 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1264 Details : Use d_move_locked instead of __d_move.
1267 Frequency : rare, at shutdown
1269 Description: access already free / zero obd_namespace.
1270 Details : if client_disconnect_export was called without force flag set,
1271 and exist connect request in flight, this can produce access to
1272 NULL pointer (or already free pointer) when connect_interpret
1273 store ocd flags in obd_namespace.
1276 Frequency : only at startup
1278 Description: not alloc memory with spinlock held.
1279 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1280 if any spinlock held.
1285 Description: lfs find does not continue on file error
1286 Details : Continue other files processing when a file/dir is absent.
1290 Description: Inconsistent usage of lustre_pack_reply()
1291 Details : Standardize the usage of lustre_pack_reply() such that it
1292 always generate a CERROR on failure.
1295 Frequency : very rare
1297 Description: Fix replay if there is an un-replied request and open
1298 Details : In some cases, older replay request will revert the
1299 mcd->mcd_last_xid on MDS which is used to record the client's
1300 latest sent request.
1302 Severity : enhancement
1304 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1306 Severity : enhancement
1308 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1310 Severity : enhancement
1312 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1314 Severity : enhancement
1316 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
1318 Severity : enhancement
1320 Description: Build SLES10 patchless client fails
1321 Details : The configure was broken by run ./configure with
1322 --with-linux-obj=.... argument for patchless client. When the
1323 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1324 can't search header adequately. Use absolute path such as
1325 -I($LINUX)/include instead.
1327 Severity : enhancement
1329 Description: Lustre Page Accounting
1330 Details : New macros for page alloc and free which enable accounting
1331 of page allocation of Lustre. Use percpu counters to store memory
1332 and page statistics.
1335 Frequency : only if debugging is disabled
1337 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1338 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1339 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1340 these macros and replace them with their definitions instead.
1345 Description: interrupt oig_wait produce painc on resend.
1346 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1347 private set, and this produce situation when rq_ptlrpcd_data not
1348 copyed to new allocated request and triggered LBUG on assert
1349 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1350 ptlrpcd set if request is changed and can be safety changed to use
1353 Severity : enhancement
1355 Description: organize the server-side client stats on per-nid basis
1356 Details : Change the structure of stats under obdfilter and mds to
1364 The "uuid"s file would list the uuids of _active_ exports.
1365 And the clear entry is to clear all stats and stale nids.
1370 Description: Processes looping in ll_readdir() on Lustre clients finally causing
1371 a full node pseudo-hang
1372 Details : Concurrent access to the same directory from multiple clients with
1373 intensive file creation/removal can cause a client node to spin in
1374 ll_readdir(). i_version must be increased every time the lock is
1375 cancelled to ensure a revalidate is done.
1380 Description: touch file failed when fs is not full
1381 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1382 otherwise we can get ENOSP while fs is not full.
1385 Frequency : only for Cray XT3
1386 Bugzilla : 12829/13455
1387 Description: Changing primary group doesn't change the group lustre assigns to
1389 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1390 overridden with the primary group provided by the group upcall,
1391 whereas the supplied fsgid can be trusted if it is in the list of
1392 supplementary groups returned by the group upcall.
1394 Severity : enhancement
1396 Description: Root Squash Functionality
1397 Details : Implementation of NFS-like root squash capability. Specifically,
1398 don't allow someone with root access on a client node to be able
1399 to manipulate files owned by root on a server node.
1401 Severity : enhancement
1403 Description: Slow trucate/writes to huge files at high offsets.
1404 Details : Directly associate cached pages to lock that protect those pages,
1405 this allows us to quickly find what pages to write and remove
1406 once lock callback is received.
1411 Description: Too many locks accumulating on client during NFS usage
1412 Details : mds_open improperly used accmode to find out access mode to a
1413 file. Also mdc_intent_lock logic to find out if we already have
1414 lock similar to just received was flawed since introduction of
1415 skiplists - locks are now added to the front of the granted
1420 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
1422 Details : In obd setup failure handler, obd_stopping will not necessarily to
1423 be 1, and obd_set_up should also be checked to make sure whether
1424 obd is completely setup.
1426 Severity : enhancement
1428 Description: Allow masking D_WARNING, D_ERROR messages from console
1429 Details : Console messages can now be disabled via lnet.printk.
1434 Description: User code with malformed file open parameter crashes client node
1435 Details : Before packing join_file req, all the related reference should be
1436 checked carefully in case some malformed flags cause fake join_file
1442 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
1443 grain of quota doing harm to quota's accuracy
1444 Details : qunit size will be changed when quota limitation is too low/high;
1445 record the pending quota write in order to get more accureate
1446 quota; delete the patch for bug12588, which is unnecessary when
1447 this patch is landed. This bug also contains fixes for bug 14526,
1448 14299, 14601 and 13794.
1452 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1453 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1454 mode is equal to the granted mode. However, it should also check
1455 that LDLM_FL_CP_REQD is not set.
1459 Description: Heavy nfs access might result in deadlocks
1460 Details : After ELC code landed, it is now improper to enqueue any mds
1461 locks under och_sem, because enqueue might want to decide to
1462 cancel open locks for same inode we are holding och_sem for.
1466 Description: 35% write performance drop with ldiskfs2 when quotas are on
1467 Details : Enable ext3 journalled quota by default to improve performance
1468 when quotas are turned on.
1472 Description: Client eviction while running blogbench
1473 Details : A lot of unlink operations with concurrent I/O can lead to a
1474 deadlock causing evictions. To address the problem, the number of
1475 oustanding OST_DESTROY requests is now throttled to
1476 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1477 callbacks are processed in priority.
1480 Frequency : RHEL4 only
1482 Description: mkfs is very slow on IA64/RHEL4
1483 Details : A performance regression has been discovered in the MPT Fusion
1484 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1485 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1486 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1489 Frequency : PPC/PPC64 only
1491 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1492 Details : fix duplicated definitions between asm-ppc64/types.h and
1493 lustre_types.h on PPC.
1496 Frequency : PPC/PPC64 only
1498 Description: asm-ppc/segment.h does not exist
1499 Details : fix compile issue on PPC.
1503 Description: data checksumming impacts single node performance
1504 Details : add support for several checksum algorithms. Currently, CRC32 and
1505 Adler-32 are supported. The checksum type can be changed on the fly
1506 through /proc/fs/lustre/osc/*/checksum_type.
1510 Description: use adler32 for page checksums
1511 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1516 Description: better handle error messages in extents code
1518 Severity : enhancement
1520 Description: SNMP support enhancement
1521 Details : Adding total number of sampled request for an MDS node in snmp
1524 Severity : enhancement
1526 Description: Optimize ldlm waiting list processing for PR extent locks
1527 Details : When processing waiting list for read extent lock and meeting read
1528 lock that is same or wider to it that is not contended, skip
1529 processing rest of the list and immediatelly return current
1530 status of conflictness, since we are guaranteed there are no
1531 conflicting locks in the rest of the list.
1535 Description: Time out and refuse to reconnect
1536 Details : When the failover node is the primary node, it is possible
1537 to have two identical connections in imp_conn_list. We must
1538 compare not conn's pointers but NIDs, otherwise we
1539 can defeat connection throttling.
1543 Description: Client not clear own cache if answer to reconnect is lost.
1544 Details : Client gets evicted from server. Now client also thinks it is
1545 disconnected (or gets ENOTCONN on its operation) and decides to
1546 reconnect. Server receives reconnect message, but cannot find
1547 export. New export is created that is fully valid (new cookie!),
1548 but reply is lost and not reported to client. Client reconnects
1549 again and gets back a just-created connection, but it is not new
1550 so client thinks it was not evicted and does not do recovery.
1554 Description: Detect stride IO mode in read-ahead
1555 Details : When a client does stride read, read-ahead should detect that and
1556 read-ahead pages according to the detected stride pattern.
1560 Description: build for x2 fails
1561 Details : fix compile issue on Cray systems.
1563 Severity : enhancement
1565 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
1566 Details : This greatly improves speed of NFS writes on 2.6 kernels.
1569 Frequency : only on PPC/SLES10
1571 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1572 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1573 be defined. Add a hack in mkfs_lustre.c to work around this compile
1578 Description: LASSERT on MDS when client holding flock lock dies
1579 Details : ldlm pool logic depends on number of granted locks equal to
1580 number of released locks which is not true for flock locks, so
1581 just exclude such locks from consideration.
1585 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1586 Details : Use fsfilt_sync() for both the whole filesystem sync and
1587 individual file sync to eliminate dangerous inode locking
1588 with I_LOCK that can lead to a deadlock.
1592 Description: Update an obsolete wirecheck.c generator
1593 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
1597 Description: Client can panic on open sometimes
1598 Details : It is possible that we try to free already freed request in
1599 ll_file_open in some error cases when we send request from
1604 Description: performance in 1.6.3
1605 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1606 sure that 1MB requests are not fragmented by the block layer.
1610 Description: LDLM soft lockups - improvement
1611 Details : It is be possible to send the lock handle along with each read
1612 or write request because the client is already doing a lock match
1613 itself so there isn't any reason the OST should have to re-do that
1619 Description: lfs quota fails with deactivated OSTS
1620 Details : With this patch, three improvements are included:
1621 1. detete the softlimit in mds and osts when use "lfs quota".
1622 2. display the inaccurate data in the output of "lfs quota".
1623 3. try to get quota info when "lfs quota" is executed.
1628 Description: Extent locks not granted with no conflicts sometimes.
1629 Details : When race occurs in glimpse handler and nothing is returned,
1630 we do not reprocess the queue after lock cancel, and that leads
1631 to a stall until next activity on a resource
1634 Frequency : failover with quotaon
1636 Description: during mds failovers with quota on, OSTs got into deadlock state
1637 and causing dumpstack.
1638 Details : for every quota slave, at any time, there is only one quota req
1639 is sent to quota master for every uid/gid. Before that quota req
1640 returns, all the thread relative to the same uid/gid will wait.
1641 So if the quota req is lost because mds failovers or any other
1642 reasons, this bug will be hit. Now, dqacq_interpret() will handle
1643 quota reqs who time out.
1645 Severity : enhancement
1648 Description: when quota slave checks if quota is enough, there is an unnecessary
1650 Details : place this wait on necessary place instead of always waiting.
1652 --------------------------------------------------------------------------------
1654 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
1656 * Support for kernels:
1657 2.6.5-7.286 (SLES 9),
1658 2.6.9-55.0.9.EL (RHEL 4),
1659 2.6.16.53-0.8 (SLES 10),
1660 2.6.18-8.1.14.el5 (RHEL 5),
1661 2.6.18.8 vanilla (kernel.org)
1662 * Client support for unpatched kernels:
1663 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1664 2.6.16 - 2.6.22 vanilla (kernel.org)
1665 * Due to recently discovered recovery problems, we do not recommend
1666 using patchless RHEL 4 clients with this or any earlier release.
1667 * Recommended e2fsprogs version: 1.40.2-cfs1
1668 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1671 Frequency : occasional
1673 Description: MDS hang or stay in waiting lock
1674 Details : If client receive lock with CBPENDING flag ldlm need send lock
1675 cancel as separate rpc, to avoid situation when cancel request
1676 can't processed due all i/o threads stay in wait lock.
1679 Frequency : occasional
1681 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1682 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1683 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1684 oscc_has_objects and see OSC has no objests and call
1685 oscc_internal_create to resend create request.
1687 Severity : enhancement
1689 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1690 Details : Update which_patch & target file for SLES10 latest kernel.
1692 Severity : enhancement
1694 Description: add --type and --size parameters to lfs find
1695 Details : Enhance lfs find by adding filetype and filesize parameters. Also
1696 multiple OBDs can now be specified for the --obd option.
1698 Severity : enhancement
1700 Description: eliminate client locks in face of contention
1701 Details : file contention detection and lockless i/o implementation
1702 for contended files.
1704 Severity : enhancement
1706 Description: Remove client patches from SLES 10 kernel.
1707 Details : This causes SLES 10 clients to behave as patchless clients
1708 even on a Lustre-patched (server) kernel.
1710 Severity : enhancement
1712 Description: use i_size_read and i_size_write in 2.6 port
1713 Details : replace inode->i_size access with i_size_read/write()
1716 Frequency : when removing large files
1718 Description: scheduling issue during removal of large Lustre files
1719 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1720 It causes scheduling issues when removing large files (17TB in the
1726 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
1727 Details : Impossible (EPROTO is returned) to access a directory that has a
1728 non-default striping and ACLs.
1731 Frequency : only on ppc
1733 Description: /proc/fs/lustre/devices broken on ppc
1734 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1735 We should make sure the type of 'index' is loff_t and then cast
1736 explicitly as needed below. Do not assign an explicitly cast
1740 Frequency : only for rhel5
1742 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1743 Details : Modify the target file & which_kernel.
1746 Frequency : if the uninit_groups feature is enabled on ldiskfs
1748 Description: e2fsck reports "invalid unused inodes count"
1749 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
1750 feature and only a single inode is created in a group then the
1751 "bg_unused_inodes" count is incorrectly updated. Creating a
1752 second inode in that group would update it correctly.
1757 Description: buffer overruns could theoretically occur
1758 Details : llapi_semantic_traverse() modifies the "path" argument by
1759 appending values to the end of the origin string, and a buffer
1760 overrun may occur. Adding buffer overrun check in liblustreapi.
1762 Severity : enhancement
1764 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
1767 Frequency : only if filesystem is inconsistent
1769 Description: handle "serious error: objid * already exists" more gracefully
1770 Details : If LAST_ID value on disk is smaller than the objects existing in
1771 the O/0/d* directories, it indicates disk corruption and causes an
1772 LBUG(). If the object is 0-length, then we should use the existing
1773 object. This will help to avoid a full fsck in most cases.
1775 Severity : enhancement
1777 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1778 Details : Modify vm-tunables-rhel4.patch.
1780 Severity : enhancement
1782 Description: Kernel config for 2.6.18-vanilla.
1783 Details : Modify targets/2.6-vanilla.target.in.
1784 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1785 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1786 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1787 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1790 Frequency : occasional
1792 Description: improve handling recoverable errors
1793 Details : If request processed with error which can be recoverable on server
1794 request should be resend, otherwise page released from cache and
1800 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1801 Details : Modify target file & which_patch.
1802 A flaw was found in the IA32 system call emulation provided
1803 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1804 value could be stored in the %RAX register, which could trigger an
1805 out-of-bounds system call table access. An untrusted local user
1806 could exploit this flaw to run code in the kernel
1807 (ie a root privilege escalation). (CVE-2007-4573).
1811 Description: change order of libsysio includes
1812 Details : '#include sysio.h' should always come before '#include xtio.h'
1814 Severity : enhancement
1816 Description: adapt the lustre_config script to support the upgrade case
1817 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1818 server targets to 1.6 in parallel.
1823 Description: To avoid grant space > avaible space when the disk is almost
1824 full. Without this patch you might see the error "grant XXXX >
1825 available" or some LBUG about grant, when the disk is almost
1827 Details : In filter_check_grant, for non_grant cache write, we should
1828 check the left space by if (*left > ungranted + bytes), instead
1829 of (*left > ungranted), because only we are sure the left space
1830 is enough for another "bytes", then the ungrant space should be
1831 increase. In client, we should update cl_avail_grant only there is
1832 OBD_MD_FLGRANT in the reply.
1837 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1838 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1839 described in CVE-2007-4573. This problem could allow untrusted
1840 local users to gain root access.
1843 Frequency : when using O_DIRECT and quotas
1845 Description: Incorrect file ownership on O_DIRECT output files
1846 Details : block usage reported by 'lfs quota' does not take into account
1847 files that have been written with O_DIRECT.
1851 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
1852 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
1853 window to 0. If it does hit cache miss, and read-pattern is still
1854 stride-io mode, does not reset the stride window, but also does not
1855 increase the stride window length in this case.
1857 --------------------------------------------------------------------------------
1859 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
1861 * Support for kernels:
1862 2.6.5-7.286 (SLES 9),
1863 2.6.9-55.0.2.EL (RHEL 4),
1864 2.6.16.46-0.14 (SLES 10),
1865 2.6.18-8.1.8.el5 (RHEL 5),
1866 2.6.18.8 vanilla (kernel.org)
1867 * Client support for unpatched kernels:
1868 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1869 2.6.16 - 2.6.21 vanilla (kernel.org)
1870 * Due to recently discovered recovery problems, we do not recommend
1871 using patchless RHEL 4 clients with this or any earlier release.
1872 * Recommended e2fsprogs version: 1.40.2-cfs1
1873 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1877 Description: Fix errors in lfs documentation
1878 Details : Fixes man pages
1880 Severity : enhancement
1882 Description: Adaptive timeouts
1883 Details : RPC timeouts adapt to changing server load and network
1884 conditions to reduce resend attempts and improve recovery time.
1886 Severity : enhancement
1888 Description: llapi_file_create() does not allow some changes
1889 Details : add llapi_file_open() that allows specifying the file creation
1890 mode and open flags, and also returns an open file handle.
1892 Severity : enhancement
1894 Description: df doesn't work properly if diskfs blocksize != 4K
1895 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1898 Frequency : always on directories with default striping set
1900 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1901 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1903 Severity : enhancement
1905 Description: merge and cleanup kernel patches.
1906 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1908 Severity : enhancement
1910 Description: RedHat Update kernel for RHEL5
1911 Details : Kernel config file for RHEL5.
1913 Severity : enhancement
1915 Description: OSS needs mutliple precreate threads
1916 Details : Add ability to start more than one create thread per OSS.
1919 Frequency : only with quota on the root user
1921 Description: mds_obd_create error creating tmp object
1922 Details : When the user sets quota on root, llog will be affected and can't
1923 create files and write files.
1926 Frequency : Always on ia64 patchless client, and possibly others.
1928 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1929 Details : This allows the patchless client to be loaded on architectures
1930 without this export.
1935 Description: disorder of journal start and llog_add cause deadlock.
1936 Details : in llog_origin_connect, journal start should happen before
1937 llog_add keep the same order as other functions to avoid
1940 Severity : enhancement
1942 Description: RedHat Update kernel for RHEL5
1943 Details : Modify the kernel config file more closer RHEL5.
1945 Severity : enhancement
1947 Description: Build failure against Centos5 (RHEL5)
1948 Details : Define PAGE_SIZE when it isn't present.
1951 Frequency : occasionally when using NFS
1953 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1954 Details : with NFS, the anon dentry's parent was set to itself in
1955 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1956 corresponding dentry other than use rec->ur_name.
1958 Severity : enhancement
1960 Description: enable data checksumming by default
1961 Details : enable checksum by default, allow --disable-checksum
1962 configure option and "-o nochecksum" mount option. Checksums
1963 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1966 Frequency : Occasionally with failover
1968 Description: Client eviction due to failover config
1969 Details : after a connection loss, the lustre client should attempt to
1970 reconnect to the last active server first before trying the
1971 other potential connections.
1973 Severity : enhancement
1975 Description: client-side metadata stat-ahead during readdir(directory readahead)
1976 Details : perform client-side metadata stat-ahead when the client detects
1977 readdir and sequential stat of dir entries therein
1982 Description: when mds and osts use different quota unit(32bit and 64bit),
1983 quota will be released repeatly.
1984 Details : void sending multiple quota reqs to mds, which will keep the
1985 status between the reqs.
1988 Frequency : only with liblustre clients on XT3
1990 Description: evictions taking too long
1991 Details : allow llrd to evict clients directly on OSTs
1995 Description: osts not allocated evenly to files
1996 Details : change the condition to increase offset_idx
1998 Severity : enhancement
2000 Description: self-adjustable client's lru lists
2001 Details : use adaptive algorithm for managing client cached locks lru
2002 lists according to current server load, other client's work
2003 pattern, memory activities, etc. Both, server and client
2004 side namespaces provide number of proc tunables for controlling
2009 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
2010 Details : rewrite inherited code (uuid parsing code from ext2 utils and
2011 readdir code from ext3) from scratch preserving functionality.
2015 Description: Only those disconnect error should be returned by rq_status.
2016 Details : In open/enqueue processs, Some errors, which will cause client
2017 disconnected, should be returned by rq_status, while other
2018 errors should still be returned by intent, then mdc or llite will
2021 Severity : enhancement
2023 Description: Tune the kernel for good SCSI performance.
2024 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2025 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2029 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2030 Bugzilla : 13547, 13627
2031 Description: Data corruption for OSTs that are formatted larger than 2TB
2033 Details : When generating the bio request for lustre file writes the
2034 sector number would overflow a temporary variable before being
2035 used for the IO. The data reads correctly from Lustre (which
2036 will overflow in a similar manner) but other file data or
2037 filesystem metadata may be corrupted in some cases.
2041 Description: "lfs find -obd UUID" prints directories
2042 Details : "lfs find -obd UUID" will return all directory names instead
2043 of just file names. It is incorrect because the directories
2044 do not reside on the OSTs.
2047 Bugzilla : 13596, 13614
2048 Description: MDS hang after unclean shutdown of lots of clients
2049 Details : Never resend AST requests.
2053 Frequency : Always, for kernels after 2.6.16
2054 Description: Fix warning idr_remove called for id=.. which is not allocated.
2055 Details : Last kernels save old s_dev before kill super and not allow
2056 to restore from callback - restore it before call kill_anon_super.
2060 Description: Fix error on 'ls .' at the top of the Lustre mount.
2061 Details : Don't revalidate dentry if it is a root dentry.
2063 --------------------------------------------------------------------------------
2065 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
2067 * Support for kernels:
2068 2.6.5-7.286 (SLES 9),
2069 2.6.9-55.0.2.EL (RHEL 4),
2070 2.6.16.46-0.14 (SLES 10),
2071 2.6.18-8.1.8.el5 (RHEL 5),
2072 2.6.18.8 vanilla (kernel.org)
2073 * Client support for unpatched kernels:
2074 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2075 2.6.16 - 2.6.22 vanilla (kernel.org)
2076 * Due to recently discovered recovery problems, we do not recommend
2077 using patchless RHEL 4 clients with this or any earlier release.
2078 * Recommended e2fsprogs version: 1.39.cfs8
2079 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2084 Description: block reactivating mgc import until all deactivates complete
2085 Details : Fix race when failing back MDT/MGS to itself (testing)
2087 Severity : enhancement
2089 Description: lfs setstripe enhancement
2090 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2093 Frequency : mds/oss recovery
2095 Description: llog ctxt is refrenced after it has been freed.
2096 Details : llog ctxt refcount was added to avoide the race between ctxt free
2097 and llog recovery process. Each llog user must hold ctxt refcount
2098 before it access the llog. And the llog ctxt can only be freed
2099 when its refcount is zero.
2101 Severity : enhancement
2103 Description: randomly memory allocation failure util
2104 Details : Make lustre randomly failed allocating memory for testing purpose.
2107 Frequency : only for SLES10
2109 Description: Update kernel patch for SLES10 SP1
2110 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
2112 Severity : enhancement
2114 Description: omit set fsid for export NFS
2115 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
2116 as FS_REQUIRES_DEV for avoid problems with generate fsid.
2119 Frequency : after network failures
2121 Description: Add sync option to mount_lustre.c
2122 Details : Client loses data written to lustre after a network interruption.
2124 Severity : enhancement
2126 Description: Error message improvement.
2127 Details : Merging of two LCONSOLE_ERROR_MSG into one.
2129 Severity : enhancement
2131 Description: don't use GFP_* in generic Lustre code.
2132 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
2134 Severity : enhancement
2136 Description: obdclass is limited by single OBD_ALLOC(idarray)
2137 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
2139 Severity : enhancement
2141 Description: updated patchess for new RHEL4 kernel
2142 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
2144 Severity : enhancement
2146 Description: warnings with build patchless client with vanila 2.6.19 and up
2147 Details : change old ctl_table style and replace ctl_table/ctl_table_header
2148 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
2150 Severity : enhancement
2152 Description: O_DIRECT bypasses client statistics.
2153 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2154 read_ahead_stats were updated. Copied stats section from
2155 osc_send_oap_rpc() into async_internal().
2158 Frequency : only for Cray XT3
2160 Description: peer credits not enough on many OST per OSS systems.
2161 Details : Use new lnet way to add credits as we need those for pings and ASTs
2164 Frequency : only with liblustre
2166 Description: Liblustre is not releasing flock locks on file close.
2167 Details : Release flock locks on file close.
2170 Frequency : only for RHEL4
2172 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
2173 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
2174 Update target file and kernel config.
2178 Description: lustre support for RHEL5
2179 Details : Add support for RHEL5.
2184 Description: ASSERTION(export != NULL) failed in target_handle_connect
2185 Details : Assetion hit is result of rare race between disconnect and connet
2186 to same nid. target_handle_connect found old connect cockie and
2187 tried to reconnect, but can't find export for this cockie.
2192 Description: umount blocks forever on error
2193 Details : In result of wrong using obd_no_recov and obd_force flags client
2194 can hand if cancel or some other requests is lost.
2197 Frequency : Only for SLES
2199 Description: sanity_quota fail test_1
2200 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
2201 file, which makes TSTID[2] inunique.
2203 Severity : enhancement
2205 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2206 Details : Update target/ChangeLog/which_patch .
2208 Severity : enhancement
2210 Description: jbd statistics
2211 Details : Port older jbd statistics patch for sles10
2216 Description: lvbo_init failed for resource with missing objects.
2217 Details : Fix returning error if we do stat for file with missing/corrupted
2218 objects and i_size set to all sum of size of all avaible objects.
2219 if we to truncate/write to missing object - it is recreated.
2224 Description: Oops in read and write path when failing to allocate lock.
2225 Details : Check if lock allocation failed and return error back.
2228 Frequency : When flocks are used.
2230 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2231 Details : Flock locks might destroy just granted lock if it could be merged
2232 with another existing flock, this is done in completion handler,
2233 so teach ldlm_cli_enquque_fini that this is a valid case for
2239 Description: reply_lock_interpret crash due to race with it and lock cancel.
2240 Details : Do not replay locks that are being cancelled. Do not reference
2241 locks by their address during replay, just by their handle.
2244 Frequency : only with deactivated OSTs
2246 Description: lstripe command fails for valid OST index
2247 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
2248 instead of lov->desc.ld_active_tgt_count.
2250 Severity : enhancement
2252 Description: Build failure against Centos5 (RHEL5)
2253 Details : Use getpagesize() instead of PAGE_SIZE.
2255 --------------------------------------------------------------------------------
2257 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
2259 * Support for kernels:
2260 2.6.5-7.283 (SLES 9),
2261 2.6.9-55.EL (RHEL 4),
2262 2.6.16.46-0.14 (SLES 10),
2263 2.6.18.8 vanilla (kernel.org)
2264 * Client support for unpatched kernels:
2265 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2266 2.6.16 - 2.6.22 vanilla (kernel.org)
2267 * Due to recently discovered recovery problems, we do not recommend
2268 using patchless RHEL 4 clients with this or any earlier release.
2269 * Recommended e2fsprogs version: 1.39.cfs8
2270 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2271 * Starting with this release, the ldiskfs backing filesystem required
2272 by Lustre is now in its own package, lustre-ldiskfs. This package
2273 should be installed. It is versioned separately from Lustre and
2274 may be released separately in future.
2279 Description: server LBUG when shutting down
2280 Details : Block umount forever until the mount refcount is zero rather
2281 than giving up after an arbitrary timeout.
2283 Severity : enhancement
2285 Description: add optional extra BUILD_VERSION info
2286 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2287 to override the lustre version.
2290 Frequency : 2.6.18 servers only
2292 Description: ll_kern_mount() doesn't release the module reference
2293 Details : The ldiskfs module reference count never drops down to 0
2294 because ll_kern_mount() doesn't release the module reference.
2299 Description: server LBUG when using old ost_num_threads parameter
2300 Details : Accept the old ost_num_threads parameter but warn that it
2301 is deprecated, and fix an off-by-one error that caused an LBUG.
2306 Description: Transient SCSI error results in persistent IO issue
2307 Details : iobuf->dr_error is not reinitialized to 0 between two
2311 Frequency : sometimes when underlying device returns I/O errors
2313 Description: OSTs not going read-only during write failures
2314 Details : OSTs are not remounted read-only when the journal commit threads
2315 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2316 instead of the ext3 wrappers.
2319 Frequency : only with 10000 clients or more
2321 Description: poor connect scaling with increasing client count
2322 Details : Don't run filter_grant_sanity_check for more than 100 exports
2323 to improve scaling for large numbers of clients.
2326 Frequency : SLES10 only
2328 Description: sanity-quota.sh quotacheck failed: rc = -22
2329 Details : Quotas cannot be enabled on SLES10.
2332 Frequency : liblustre clients only
2334 Description: getdirentries does not give error when run on compute nodes
2335 Details : getdirentries does not fail when the size specified as an argument
2336 is too small to contain at least one entry
2338 Severity : enhancement
2340 Description: Add LNET router traceability for debug purposes
2341 Details : If a checksum failure occurs with a router as part of the
2342 IO path, the NID of the last router that forwarded the bulk data
2343 is printed so it can be identified.
2348 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2349 Details : Due to a race condition, liblustre clients were occasionally
2350 evicted incorrectly.
2352 Severity : enhancement
2354 Description: lfs setstripe use optional parameters instead of postional
2357 Severity : enhancement
2359 Description: Nanosecond timestamp support for ldiskfs
2360 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2361 resolution timestamps. There is not yet support for this at
2362 the Lustre filesystem level.
2365 Frequency : during server recovery
2367 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2368 Details : request with rq_no_resend flag not awake l_wait_event if they get a
2372 Frequency : nfs export on patchless client
2374 Description: connectathon hang when test nfs export over patchless client
2375 Details : Disconnected dentry cannot be found with lookup, so we do not need
2376 to unhash it or make it invalid
2379 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2380 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2381 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2386 Description: open req refcounting wrong on reconnect
2387 Details : If reconnect happened between getting open reply from server and
2388 call to mdc_set_replay_data in ll_file_open, we will schedule
2389 replay for unreferenced request that we are about to free.
2390 Subsequent close will crash in variety of ways.
2391 Check that request is still eligible for replay in
2392 mdc_set_replay_data().
2397 Description: disable writes to filesystem when reading health_check file
2398 Details : the default for reading the health_check proc file has changed
2399 to NOT do a journal transaction and write to disk, because this
2400 can cause reads of the /proc file to hang and block HA state
2401 checking on a healthy but otherwise heavily loaded system. It
2402 is possible to return to the previous behaviour during configure
2403 with --enable-health-write.
2405 Severity : enhancement
2407 Description: 64-bit inode version
2408 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2409 to the inode. This will be required for version-based recovery.
2414 Description: MDS fails to start if a duplicate client export is detected
2415 Details : in some rare cases it was possible for a client to connect to
2416 an MDS multiple times. Upon recovery the MDS would detect this
2417 and fail during startup. Handle this more gracefully.
2419 Severity : enhancement
2421 Description: Add -o localflock option to simulate old noflock behaviour.
2422 Details : This will achieve local-only flock/fcntl locks coherentness.
2427 Description: log_commit_thread vs filter_destroy race leads to crash
2428 Details : Take import reference before releasing llog record semaphore
2433 Description: Wrong request locking in request set processing
2434 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2435 imp_delayed_list, in this place should be used imp_lock.
2438 Frequency : when reconnecting
2440 Description: Grant leak when OSC reconnect to OST
2441 Details : When osc reconnect ost, OST(filter) should check whether it
2442 should grant more space to client by comparing fed_grant and
2443 cl_avail_grant, and return the granted space to client instead
2444 of "new granted" space, because client will call osc_init_grant
2445 to update the client grant space info.
2448 Frequency : when client reconnects to OST
2450 Description: Grant leak when OSC does a resend and replays bulk write
2451 Details : When osc reconnect to OST, OST(filter) should clear grant info of
2452 bulk write request, because the grant info while be sync between
2453 OSC and OST when reconnect, and we should ignore the grant info
2454 these of resend/replay write req.
2459 Description: Grant space more than avaiable space sometimes.
2460 Details : When then OST is about to be full, if two bulk writing from
2461 different clients came to OST. Accord the avaliable space of the
2462 OST, the first req should be permitted, and the second one
2463 should be denied by ENOSPC. But if the seconde arrived before
2464 the first one is commited. The OST might wrongly permit second
2465 writing, which will cause grant space > avaiable space.
2468 Frequency : when client is evicted
2470 Description: Grant might be wrongly erased when osc is evicted by OST
2471 Details : when the import is evicted by server, it will fork another
2472 thread ptlrpc_invalidate_import_thread to invalidate the
2473 import, where the grant will be set to 0. While the original
2474 thread will update the grant it got when connecting. So if
2475 the former happened latter, the grant will be wrongly errased
2476 because of this race.
2481 Description: Checking Stale with correct fid
2482 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2483 to check whether it is stale, because sometimes, we want the
2484 enqueue happened anyway, and op_data.fid2 will not be initialized.
2486 Severity : enhancement
2488 Description: update patchless client
2489 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2492 Frequency : only with 2.4 kernel
2494 Description: random memory corruption
2495 Details : size of struct ll_inode_info is to big for union inode.u and this
2496 can be cause of random memory corruption.
2501 Description: Memory leak in recovery
2502 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2503 It should also check obd_fail before fsfilt_start, otherwise if
2504 fsfilt_start return -EROFS,(failover mds during mds recovery).
2505 then the req will return with repmsg->transno = 0 and rc = EROFS.
2506 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2507 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2508 be freed no matter whether fsfilt_commit success or not.
2511 Frequency : only with large numbers of cores on a single node
2513 Description: Prevents from taking the superblock lock in llap_from_page for
2515 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2516 for llap_from_page call in ll_removepage() prevents from taking
2517 the superblock lock for a soon died page.
2522 Description: Not check open intent error before release open handle
2523 Details : in some rare cases, the open intent error is not checked before
2524 release open handle, which may cause
2525 ASSERTION(open_req->rq_transno != 0), because it tries to release
2526 the failed open handle.
2531 Description: Set cat log bitmap only after create log success.
2532 Details : in some rare cases, the cat log bitmap is set too early. and it
2533 should be set only after create log success.
2537 Description: Accessing a block bevice can re-enable I/O when Lustre is
2538 tearing down a device.
2539 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2544 Description: service threads may hog cpus when there are a lot of requests
2545 Details : Insert cond_resched to give other threads a chance to use some CPU
2550 Description: the cat log was not initialized in recovery
2551 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2552 unlink log on mds will not be initialized until mds post
2553 recovery. And also in mds post recovery, the unlink log will
2554 initialization will be done asynchronausly, so there will be race
2555 between add unlink log and unlink log initialization.
2559 Description: brw_stats were being printed incorrectly
2560 Details : brw_stats were being printed as log2 but all of them were not
2561 recorded as log2. Also remove some code duplication arising from
2562 filter_tally_{read,write}.
2566 Frequency : rare, only in recovery.
2567 Description: ASSERTION(req->rq_type != LI_POISON) failed
2568 Details : imp_lock should be held while iterating over imp_sending_list for
2569 prevent destroy request after get timeout in ptlrpc_queue_wait.
2573 Description: replay-single.sh test 52 fails
2574 Details : A lock's skiplist need to be cleanup when it being unlinked
2575 from its resource list.
2580 Description: Short directio read returns full requested size rather than
2582 Details : Direct I/O operations should return actual amount of bytes
2583 transferred rather than requested size.
2585 Severity : enhancement
2587 Description: metadata RPC reduction (e.g. for rm performance)
2588 Details : decrease the amount of synchronous RPC between clients and servers
2589 by canceling conflicing lock before the operation on the client
2590 and packing thier handles into the main operation RPC to server.
2592 Severity : enhancement
2594 Description: add #ifdef HAVE_KERNEL_CONFIG_H
2595 Details : kernels from 2.6.19 not need include linux/config.h, but add
2596 include linux/autoconf.h in commpiler command line.
2598 Severity : enhancement
2600 Description: patchless client support for 2.6.22 kernel
2601 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
2602 removed. In this case we need drop using os depended interface to
2603 kmem_cache and use cfs_mem_cache API.
2608 Description: fix mal-formatted messages
2609 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
2613 Frequency : always in liblustre
2614 Description: wrong IS_ERR implementation in liblustre.h
2615 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
2620 Description: Correct condition for output debug message.
2621 Details : inode i_nlink equal zero is not enough for output message about
2622 disk corruption, i_ctime and i_mode should be also checked.
2626 Frequency : always in patchless client
2627 Description: add configure check for truncate_complete_page
2628 Details : improve checks for exported symbols. This allow run check without
2629 sources, but with Module.symvers shipped with kernel distribution.
2630 add check for truncate_complete_page used by patchless client.
2634 Description: sanity.sh test_77h fails with "test_77h file compare failed"
2635 Details : test_77h uses a file which was messed by other test case.
2639 Description: Not Check whether lov_tgts is NULL in some lov functions
2640 Details : Checking whether lov_tgts is NULL in some functions.
2644 Description: replace obdo_alloc() with OBDO_ALLOC macro
2645 Details : nothing special is done in obdo_alloc() function, and for
2646 debugging purpose, it needs to be replaced with macros.
2650 Description: bad return value and errno from fcntl call
2651 Details : In liblustre API, errno should be a negative value if error
2656 Description: ptlrpc_check_set() LBUG
2657 Details : In case of positive reply from server and failed client bulk
2658 callback after bulk transfer shouldn't LBUG, but process this
2659 request as erroneous.
2661 Severity : enhancement
2663 Description: VFS operations stats tool.
2664 Details : Tool which collects stats by tracking value written in pid,
2665 ppid, gid and uses llstat to generate output to plot graph using
2667 Updated lustre/utils/Makefile.am
2668 Added lustre/utils/ltrack_stats.c
2670 Severity : enhancement
2672 Description: 2.6.18 server support (lustre 1.6.1)
2673 Details : Support for 2.6.18 kernels on the server side.
2678 Description: ASSERTION(imp->imp_conn_current) failed
2679 Details : an assertion failure is hit if a client node boots and attempts to
2680 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
2683 Frequency : only for i686
2685 Description: 1.4.11 RC1 build fails for RHEL 4, i686
2686 Details : Fixed config variable for build.
2691 Description: Updated patchess for new RHEL4 kernel
2692 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
2693 Updated patch jbd-stats-2.6.9.patch
2694 Updated patch qsnet-rhel4-2.6.patch
2695 Updated patch quota-deadlock-on-pagelock-core.patch
2696 Updated patch vfs_intent-2.6-rhel4.patch
2697 Updated patch vfs_races-2.6-rhel4.patch
2698 Updated series file 2.6-rhel4-titech.series
2699 Updated series file 2.6-rhel4.series
2700 Updated kernel config files
2705 Description: lquota slave complains LBUG when reconnecting with mds
2707 Details : quota slave depends qctxt->lqc_import to send its quota request.
2708 This pointer will be invalid if mds did failover or broke its
2709 connect to osts, which leads to LBUG.
2712 Frequency : when qunit size is too small(less than 20M)
2714 Description: write is stopped by improper -EDQUOT
2715 Details : If the master is busy and qunit size is small enough(let's say 1M),
2716 the slave can not get quota from master on time, which will lead
2717 slave to trigger a -EQUOTA to client.
2722 Description: Deadlock during metadata tests
2723 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
2724 called with the per-dentry lock held.
2727 Frequency : SLES9 only
2729 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
2730 Details : set CONFIG_NR_CPUS to 128 instead of 8.
2732 Severity : enhancement
2734 Description: remove fs_prep_san_write operation and related patches
2735 Details : remove the ext3-san-jdike patches which are no longer useful.
2740 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
2741 Details : a disk corruption can cause the mballoc code to assert on a
2742 double free or other extent corruptions. Handle these with
2743 ext3_error() instead of assertions.
2746 Frequency : only with mballoc3 code and deep extent trees
2748 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
2749 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
2750 can cause the filesystem to be remounted read-only.
2755 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
2756 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
2757 few external users of this ioctl number is lfsck's call to
2758 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
2759 Add the old ioctl number to the handler so both old and new
2764 Description: parallel lock callbacks
2765 Details : Instead of sending blocking and completion callbacks as separated
2766 requests, adding them to a set and sending in parallel.
2770 Description: Disable most debugging by default
2771 Details : To improve performance, disable most logging (for debug purposes)
2772 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
2773 default, and HA includes fewer messages.
2777 Frequency : only run on patchless client.
2778 Description: use do_facet on sanity.sh for test handling recoverables errors
2779 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
2783 Description: hash tables for lists of nids, connections and uuids
2784 Details : Hash tables noticeably help when a lot of clients connect to a
2785 server, to faster identify duplicate connections or reconnects,
2786 also to faster find export to evict in manual eviction case.
2790 Description: Sometimes, when the server evict a client, and the client will
2791 not be evicted as soon as possible.
2792 Details : In enqueue req, the error was returned by intent, instead of
2793 rq_status which make ptlrpc layer not detect this error, and
2794 does not evict the client. So enqueue error should be returned
2798 Frequency : only at startup
2800 Description: Delay client connections to MDT until fist MDT->OST connect
2801 Details : If a client tried to create a new file before the MDT had
2802 connected to any OSTs, the create would return EIO. Now
2803 the client will simply block until the MDT connects to the
2804 first OST and the create can succeed.
2807 Frequency : only for SLES9
2809 Description: Routinely utilize latest Quadrics drivers in CFS releases
2810 Details : Update patch qsnet-suse-2.6.patch.
2813 Frequency : only for sles10
2815 Description: Update patches for SLES 10 SP1 kernel.
2816 Details : Update the patch vfs_intent-2.6-sles10.patch.
2821 Description: Routinely utilize latest Quadrics drivers in CFS releases
2822 Details : Update patch qsnet-rhel4-2.6.patch.
2825 Frequency : at statup only
2827 Description: mds_lov_synchronize race leads to various problems
2828 Details : simultaneous MDT->OST connections at startup can cause the
2829 sync to abort, leaving the OSC in a bad state.
2834 Description: Using wrong pointer in osc_brw_prep_request
2835 Details : Access to array[-1] can produce panic if kernel compiled with
2836 CONFIG_PAGE_ALLOC enabled
2838 Severity : enhancement
2840 Description: Async OSC create to avoid the blocking unnecessarily.
2841 Details : If a OST has no remain object, system will block on the creating
2842 when need to create a new object on this OST. Now, ways use
2843 pre-created objects when available, instead of blocking on an
2844 empty osc while others are not empty. If we must block, we block
2845 for the shortest possible period of time.
2849 Frequency : only in recovery
2850 Description: Mark OST as early accessible if his start SYNC.
2851 Details : osc_precreate return flag early accessible if oscc marked as
2852 OSCC_FLAG_SYNC_IN_PROGRESS.
2857 Description: Sometimes precreate code can triger create object on wrong ost
2858 Details : Wrong protected or not not restored variables aftre precreate loop
2859 can produce creation object on wrong ost.
2862 Frequency : oss recovery
2864 Description: llog_commit_thread cleanup should sync with llog_commit_thread
2866 Details : llog_commit_thread_count should be synced between llog_commit
2867 start and cleanup, so new llog_commit thread should not be started
2868 when llog_commit threads being stopped to avoid accessing some
2871 Severity : enhancement
2873 Description: Add printing inode info into message about error in writepage.
2875 --------------------------------------------------------------------------------
2877 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2882 Frequency : on some architectures
2884 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2885 Details : Uninitialized flags sometimes cause configuration commands to
2889 Frequency : patchless clients only
2891 Description: missing __iget() symbol export
2892 Details : The __iget() symbol export is missing. To avoid the need for
2893 this on patchless clients the deathrow inode reaper is turned
2894 off, and we depend on the VM to clean up old inodes. This
2895 dependency was during via the fix for bug 12181.
2900 Description: sanity.sh fail: test_52b
2901 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
2902 incorrect inode's flags to client.
2904 --------------------------------------------------------------------------------
2906 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2908 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2909 INTEROPERATE with older versions automatically. In many cases a
2910 special upgrade step is needed. Please read the
2911 user documentation before upgrading any part of a 1.4.x system.
2912 * WARNING: Lustre configuration and startup changes are required with
2913 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2915 * Support for kernels:
2916 2.4.21-47.0.1.EL (RHEL 3),
2917 2.6.5-7.283 (SLES 9),
2918 2.6.9-42.0.10.EL (RHEL 4),
2919 2.6.12.6 vanilla (kernel.org),
2920 2.6.16.27-0.9 (SLES10)
2921 * Client support for unpatched kernels:
2922 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
2923 2.6.16 - 2.6.19 vanilla (kernel.org),
2924 2.6.9-42.0.8EL (RHEL 4)
2925 * Recommended e2fsprogs version: 1.39.cfs6
2926 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2929 Severity : enhancement
2931 Description: Async OSC create to avoid the blocking unnecessarily.
2932 Details : If a OST has no remain object, system will block on the creating
2933 when need to create a new object on this OST. Now, ways use
2934 pre-created objects when available, instead of blocking on an
2935 empty osc while others are not empty. If we must block, we block
2936 for the shortest possible period of time.
2938 Severity : enhancement
2940 Description: MountConf
2941 Details : Lustre configuration is now managed via mkfs and mount
2942 commands instead of lmc and lconf. New obd types (MGS, MGC)
2943 are added for dynamic configuration management. See
2944 https://mail.clusterfs.com/wikis/lustre/MountConf for
2947 Severity : enhancement
2949 Description: dynamic OST addition
2950 Details : OSTs can now be added to a live filesystem
2952 Severity : enhancement
2954 Description: startup order invariance
2955 Details : MDTs and OSTs can be started in any order. Clients only
2956 require the MDT to complete startup.
2958 Severity : enhancement
2960 Description: parallel, asynchronous orphan cleanup
2961 Details : orphan cleanup is now performed in separate threads for each
2962 OST, allowing parallel non-blocking operation.
2964 Severity : enhancement
2966 Description: optimized stripe assignment
2967 Details : stripe assignments are now made based on ost space available,
2968 ost previous usage, and OSS previous usage, in order to try
2969 to optimize storage space and networking resources.
2971 Severity : enhancement
2973 Description: Permanently set tunables
2974 Details : All writable /proc/fs/lustre tunables can now be permanently
2975 set on a per-server basis, at mkfs time or on a live
2978 Severity : enhancement
2980 Description: Lustre message v2
2981 Details : Add lustre message format v2.
2983 Severity : enhancement
2985 Description: client OST exclusion list
2986 Details : Clients can be started with a list of OSTs that should be
2987 declared "inactive" for known non-responsive OSTs.
2991 Description: ENOENT returned for valid filehandle during dbench.
2992 Details : Check if a directory has children when invalidating dentries
2993 associated with an inode during lock cancellation. This fixes
2994 an incorrect ENOENT sometimes seen for valid filehandles during
2995 testing with dbench.
2998 Frequency : SFS test only (otherwise harmless)
3000 Description: SPEC SFS validation failure on NFS v2 over lustre.
3001 Details : Changes the blocksize for regular files to be 2x RPC size,
3002 and not depend on stripe size.
3004 Severity : enhancement
3006 Description: fine-grained SMP locking inside DLM
3007 Details : Improve DLM performance on SMP systems by removing the single
3008 per-namespace lock and replace it with per-resource locks.
3010 Severity : enhancement
3012 Description: don't hold multiple extent locks at one time
3013 Details : To avoid client eviction during large writes, locks are not
3014 held on multiple stripes at one time or for very large writes.
3015 Otherwise, clients can block waiting for a lock on a failed OST
3016 while holding locks on other OSTs and be evicted.
3018 Severity : enhancement
3020 Description: Multiple MD RPCs in flight.
3021 Details : Further unserialise some read-only MDT RPCs - learn about intents.
3022 To avoid overly-overloading MDT, introduce a limit on number of
3023 MDT RPCs in flight for a single client and add /proc controls
3024 to adjust this limit.
3026 Severity : enhancement
3028 Description: client read/write statistics
3029 Details : Add client read/write call usage stats for performance
3030 analysis of user processes.
3031 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3032 file access. extents_stats shows chunk size distribution.
3033 extents_stats_per_process show chunk size distribution per
3036 Severity : enhancement
3038 Description: per-client statistics on server
3039 Details : Add ldlm and operations statistics for each client in
3040 /proc/fs/lustre/mds|obdfilter/*/exports/
3042 Severity : enhancement
3044 Description: improved MDT statistics
3045 Details : Add detailed MDT operations statistics in
3046 /proc/fs/lustre/mds/*/stats
3048 Severity : enhancement
3050 Description: VFS operations stats
3051 Details : Add client VFS call stats, trackable by pid, ppid, or gid
3052 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
3057 Description: Fix client-side osc byte counters
3058 Details : The osc read/write byte counters in
3059 /proc/fs/lustre/osc/*/stats are now working
3062 Frequency : always as root on SLES
3064 Description: Failure of copying files with lustre special EAs.
3065 Details : Client side always return success for setxattr call for lustre
3066 special xattr (currently only "trusted.lov").
3071 Description: Refcount LNET uuids
3072 Details : The global LNET uuid list grew linearly with every startup;
3073 refcount repeated list entries instead of always adding to
3076 Severity : enhancement
3078 Description: Dynamic service threads
3079 Details : Within a small range, start extra service threads
3080 automatically when the request queue builds up.
3083 Frequency : mixed-endian client/server environments
3085 Description: mixed-endian crashes
3086 Details : The new msg_v2 system had some failures in mixed-endian
3089 Severity : enhancement
3091 Description: Easy OST removal
3092 Details : OSTs can be permanently deactivated with e.g. 'lctl
3093 conf_param lustre-OST0001.osc.active=0'
3095 Severity : enhancement
3097 Description: MGS proc entries
3098 Details : Added basic proc entries for the MGS showing what filesystems
3101 Severity : enhancement
3103 Description: provide MGS failover
3104 Details : Added config lock reacquisition after MGS server failover.
3106 Severity : enhancement
3108 Description: add Linux 2.4 support
3109 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3113 Description: a large application tries to do I/O to the same resource and dies
3114 in the middle of it.
3115 Details : Check the req->rq_arrival time after the call to
3116 ost_brw_lock_get(), but before we do anything about
3117 processing it & sending the BULK transfer request. This
3118 should help move old stale pending locks off the queue as
3119 quickly as obd_timeout.
3122 Frequency : when an incorrect nid is specified during startup
3124 Description: ptlrpc connect to non-existant node causes kernel crash
3125 Details : LNET can't be re-entered from an event callback, which
3126 happened when we expire a message after the export has been
3127 cleaned up. Instead, hand the zombie cleanup off to another
3130 Severity : enhancement
3132 Description: plain/inodebits lock performance improvement
3133 Details : Grouping plain/inodebits in granted list by their request modes
3134 and bits policy, thus improving the performance of search through
3138 Frequency : only if OST filesystem is corrupted
3140 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3141 Details : for a short time RPCs with bulk IO are in the replay list,
3142 but replay of bulk IOs is unimplemented. If the OST filesystem
3143 is corrupted due to disk cache incoherency and then replay is
3144 started it is possible to trip an assertion. Avoid putting
3145 committed RPCs into the replay list at all to avoid this issue.
3148 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3150 Description: System hang on startup
3151 Details : This bug allowed the liblustre (e.g. catamount) client to
3152 return to the app before handling all startup RPCs. This
3153 could leave the node unresponsive to lustre network traffic
3154 and manifested as a server ptllnd timeout.
3156 Severity : enhancement
3158 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3159 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3160 boolean to control whether to print peer debug info when a
3161 client's RPC times out.
3164 Frequency : only for kernels with patches from Lustre below 1.4.3
3166 Description: Remove old rdonly API
3167 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3170 Frequency : only for devices with external journals
3172 Description: Set external device read-only also
3173 Details : During a commanded failover stop, we set the disk device
3174 read-only while the server shuts down. We now also set any
3175 external journal device read-only at the same time.
3178 Frequency : when upgrading from 1.4 while trying to change parameters
3180 Description: The wrong (new) MDC name was used when setting parameters for
3181 upgraded MDT's. Also allows changing of OSC (and MDC)
3182 parameters if --writeconf is specified at tunefs upgrade time.
3185 Frequency : when setting specific ost indicies
3187 Description: QOS code breaks on skipped indicies
3188 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3189 created with --index need not be sequential.
3191 Severity : enhancement
3193 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3194 Details : The uninit_groups feature works in conjunction with the kernel
3195 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3196 up the pass1 processing of e2fsck. This is a read-only feature
3197 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3198 mount filesystems that have had this feature enabled.
3200 Severity : enhancement
3202 Description: Improve multi-block allocation algorithm to avoid fragmentation
3203 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3204 allocation locality and avoid filesystem fragmentation.
3206 ------------------------------------------------------------------------------
3208 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3210 * Support for kernels:
3211 2.4.21-47.0.1.EL (RHEL 3)
3212 2.6.5-7.283 (SLES 9)
3213 2.6.9-42.0.10.EL (RHEL 4)
3214 2.6.12.6 vanilla (kernel.org)
3215 2.6.16.27-0.9 (SLES 10)
3216 * Recommended e2fsprogs version: 1.39.cfs5
3218 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3222 Frequency : occasional, depends on client load and configuration
3223 Bugzilla : 12181, 12203
3224 Description: data loss for recently-modified files
3226 Details : In some cases it is possible that recently written or created
3227 files may not be written to disk in a timely manner (this should
3228 normally be within 30s unless client IO load is very high).
3229 The problem appears as zero-length files or files that are a
3230 multiple of 1MB in size after a client crash or client eviction
3231 that are missing data at the end of the file.
3233 This problem is more likely to be hit on clients where files are
3234 repeatedly created and unlinked in the same directory, clients
3235 have a large amount of RAM, have many CPUs, the filesystem has
3236 many OSTs, the clients are rebooted frequently, and/or the files
3237 are not accessed by other nodes after being written.
3239 The presence of the problem can be detected by looking at
3240 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3241 smaller than the second (nr_unused) then dirty files will not
3242 be flushed automatically to disk. "sync; sleep 10" should be
3243 run several times on the node before unmounting it to update
3244 Lustre (this is also safe to run on nodes without this problem).
3246 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3247 that can cause this same problem, so customers using that kernel
3248 also need to update the kernel in addition to Lustre. In order
3249 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3251 It is normal that files written just before a client crash (less
3252 than 30s) may not yet have been flushed to disk, even for local
3256 Frequency : frequent on thin XT3 nodes
3258 Description: UUID collision on thin XT3 Linux nodes
3259 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3260 randomly, since we relied on an insufficiently-seeded PRNG.
3265 Description: OSS hangs after "All ost request buffers busy"
3266 Details : A deadlock between quota and journal operations caused OSS
3267 hangs after printing "All ost request buffers busy."
3270 Frequency : always on liblustre builds
3272 Description: Cleanup compiler warnings on liblustre
3275 Frequency : always on liblustre builds on XT3
3277 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3281 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3282 > 32000 subdirectories
3283 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3284 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3285 subdirectory count crosses 32000. This will aid e2fsck to
3286 correctly handle more than 32000 subdirectories.
3289 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3291 Description: System hang on startup
3292 Details : This bug allowed the liblustre (e.g. catamount) client to
3293 return to the app before handling all startup RPCs. This
3294 could leave the node unresponsive to lustre network traffic
3295 and manifested as a server ptllnd timeout.
3297 Severity : enhancement
3299 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3300 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3301 boolean to control whether to print peer debug info when a
3302 client's RPC times out.
3307 Description: make O_SYNC working on 2.6 kernels
3308 Details : 2.6 kernels use different method for mark pages for write,
3309 so need add a code to lustre for O_SYNC work.
3314 Description: Failure to close file and release space on NFS
3315 Details : Put inode details into lock acquired in ll_intent_file_open.
3316 Use mdc_intent_lock in ll_intent_open to properly
3317 detect all kind of errors unhandled by mdc_enqueue.
3322 Description: proc file read during shutdown sometimes raced obd removal,
3324 Details : Add lock to prevent obd access after proc file removal.
3327 Frequency : Only for files larger than 4GB on 32-bit clients.
3329 Description: improperly doing page alignment of locks
3330 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3331 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3336 Description: RPCs being resent when they shouldn't be
3337 Details : Some RPCs that should not be resent are being resent. This
3338 can cause inconsistencies in the RPC state machine. Do not
3339 resend such requests.
3342 Frequency : rare, only with NFS export
3344 Description: Crash on NFS re-export node
3345 Details : under very unusual load conditions an assertion is hit in
3346 ll_intent_file_open()
3349 Frequency : only if OST filesystem is corrupted
3351 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3352 Details : for a short time RPCs with bulk IO are in the replay list,
3353 but replay of bulk IOs is unimplemented. If the OST filesystem
3354 is corrupted due to disk cache incoherency and then replay is
3355 started it is possible to trip an assertion. Avoid putting
3356 committed RPCs into the replay list at all to avoid this issue.
3361 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3362 Details : Large single O_DIRECT read and write calls can fail to allocate
3363 a sufficiently large buffer to process the request. In case of
3364 allocation failure the allocation is retried with a smaller
3365 buffer and broken into smaller requests.
3367 Severity : enhancement
3369 Description: Add -o localflock option to simulate old noflock behaviour.
3370 Details : This will achieve local-only flock/fcntl locks coherentness.
3375 Description: versioning check is incomplete
3376 Details : Checking the version difference of client vs. server, report
3377 error if the gap is too big.
3382 Description: add support PG_writeback bit
3383 Details : add support for PG_writeback bit for Lustre, for more carefull
3384 work with page cache in 2.6 kernel. This also fix some deadlocks
3385 and remove hack for work O_SYNC with 2.6 kernel.
3387 Severity : enhancement
3389 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3390 Details : The uninit_groups feature works in conjunction with the kernel
3391 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3392 up the pass1 processing of e2fsck. This is a read-only feature
3393 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3394 mount filesystems that have had this feature enabled.
3396 Severity : enhancement
3398 Description: Improve multi-block allocation algorithm to avoid fragmentation
3399 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3400 allocation locality and avoid filesystem fragmentation.
3402 ------------------------------------------------------------------------------
3404 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3406 * Support for kernels:
3407 2.6.9-42.0.3.EL (RHEL 4)
3408 2.6.5-7.276 (SLES 9)
3409 2.4.21-47.0.1.EL (RHEL 3)
3410 2.6.12.6 vanilla (kernel.org)
3411 2.6.16.21-0.8 (SLES10)
3412 * Recommended e2fsprogs version: 1.39.cfs2-0
3414 * The backwards-compatible /proc/sys/portals symlink has been removed
3415 in this release. Before upgrading, please ensure that you change
3416 any configuration scripts or /etc/sysctl.conf files that access
3417 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3418 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3419 in advance of the upgrade on any system running Lustre 1.4.6 or
3420 newer, since /proc/sys/lnet was added in that version.
3421 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3425 Frequency : only when quota is used
3427 Description: avoid scanning export list for quota master
3428 Details : Change the algorithms to avoid scanning export list in order
3429 to improve the efficiency.
3432 Frequency : MDS failover only, very rarely
3434 Description: "went back in time" messages on mds failover
3435 Details : The greatest transno may be lost when the current operation
3436 finishes with an error (transno==0) and the client's last_rcvd
3437 record is over-written. Save the greatest transno in the
3438 mds_last_transno for this case.
3441 Frequency : always for specific kernels and striping counts
3443 Description: client may get "Matching packet too big" without ACL support
3444 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3445 when trying to access files in certain configurations. The
3446 clients should in fact be denied when mounting because they do
3447 not understand ACLs.
3450 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3452 Description: many clients connecting with IO in progress causes connect timeouts
3453 Details : Avoid synchronous journal commits to avoid delays caused by many
3454 clients connecting/disconnecting when bulk IO is in progress.
3455 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3456 OST_IO_PORTAL to avoid delays behind potentially many pending
3460 Frequency : occasionally with multiple writers to a single file
3462 Description: shared writes to file may result in wrong size reported by stat()
3463 Details : Allow growing of kms when extent lock is cancelled
3466 Frequency : always with random mmap IO to multi-striped file
3468 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3469 Details : Only if the hole is at the end of OST object so that kms is too
3470 small. Fix is to increase kms accordingly in ll_nopage.
3473 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3475 Description: writes to a missing object would leak memory on the OST
3476 Details : If there is an inconsistency between the MDS and OST filesystems,
3477 such that the MDS references an object that doesn't exist, writes
3478 to that object will leak memory due to incorrect cleanup in the
3479 error handling path, eventually running out of memory on the OST.
3484 Description: Creating too long symlink causes lustre errors
3485 Details : Check symlink and name lengths before sending requests to MDS.
3488 Frequency : only if flock is enabled (not on by default)
3490 Description: posix locks not released on fd closure on 2.6.9+
3491 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3492 kernels, this caused such locks not to be released on fd close and
3493 then assertions on fs unmount about still used locks.
3496 Frequency : MDS failover only, very rarely
3498 Description: clients may get ASSERTION(granted_lock != NULL)
3499 Details : When request was taking a long time, and a client was resending
3500 a getattr by name lock request. The were multiple lock requests
3501 with the same client lock handle and
3502 mds_getattr_name->fixup_handle_for_resent_request found one of the
3503 lock handles but later failed with ASSERTION(granted_lock != NULL).
3508 Description: handle->h_buffer_credits > 0, assertion failure
3509 Details : h_buffer_credits is zero after truncate, causing assertion
3510 failure. This patch extends the transaction or creates a new
3514 Frequency : NFS re-export or patchless client
3515 Bugzilla : 11179, 10796
3516 Description: Crash on NFS re-export node (__d_move)
3517 Details : We do not want to hash the dentry if we don't have a lock.
3518 But if this dentry is later used in d_move, we'd hit uninitialised
3519 list head d_hash, so we just do this to init d_hash field but
3520 leave dentry unhashed.
3523 Frequency : NFS re-export or patchless client
3525 Description: NFS exports has problem with symbolic link
3526 Details : lustre client didn't properly install dentry when re-exported
3527 to NFS or running patchless client.
3530 Frequency : NFS re-export or patchless client
3532 Description: Various nfs/patchless fixes.
3533 Details : fixes reuse disconected alias for lookup process - this fixes
3534 warning "find_exported_dentry: npd != pd",
3535 fix permission error with open files at nfs.
3536 fix apply umask when do revalidate.
3539 Frequency : occasional
3541 Description: Crash on NFS re-export node
3542 Details : calling clear_page() on the wrong pointer triggered oops in
3543 generic_mapping_read().
3546 Frequency : rarely, using O_DIRECT IO
3548 Description: unaligned directio crashes client with LASSERT
3549 Details : check for unaligned buffers before trying any requests.
3552 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3554 Description: stale data returned from RAID cache
3555 Details : If only a small amount of IO is done to the RAID device before
3556 reading it again it is possible to get stale data from the RAID
3557 cache instead of reading it from disk.
3560 Frequency : always for sles10 kernel
3562 Description: sles10 support
3563 Details : ll_follow_link: compile fixes and using of nd_set_link
3564 under newer kernels.
3567 Frequency : depends on arch, kernel and compiler version, always on sles10
3570 Description: recursive or deep enough symlinks cause stack overflow
3571 Details : getting rid of large stack-allocated variable in
3575 Frequency : depends on hardware
3577 Description: lustre write performance loss in the SLES10 kernel
3578 Details : the performance loss is caused by using of write barriers in the
3579 ext3 code. The SLES10 kernel turns barrier support on by
3580 default. The fix is to undo that change for ldiskfs.
3582 ------------------------------------------------------------------------------
3584 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3586 * Support for kernels:
3587 2.6.9-42.0.3EL (RHEL 4)
3588 2.6.5-7.276 (SLES 9)
3589 2.4.21-47.0.1.EL (RHEL 3)
3590 2.6.12.6 vanilla (kernel.org)
3594 Frequency : quota enabled and large files being deleted
3596 Description: releasing more than 4GB of quota at once hangs OST
3597 Details : If a user deletes more than 4GB of files on a single OST it
3598 will cause the OST to spin in an infinite loop. Release
3599 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3604 Description: statfs data retrieved from /proc may be stale or zero
3605 Details : When reading per-device statfs data from /proc, in the
3606 {kbytes,files}_{total,free,avail} files, it may appear
3607 as zero or be out of date.
3610 Frequency : systems with MD RAID1 external journal devices
3612 Description: lconf's call to blkid is confused by RAID1 journal devices
3613 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3614 of returning all block devices that match the journal UUID.
3617 Frequency : always, for aggregate stripe size over 4GB
3619 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3620 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3621 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3624 Frequency : always if "lfs find" used on a local file/directory
3626 Description: "lfs find" segfaults if used on a local file/directory
3627 Details : The case where a directory component was not specified wasn't
3628 handled correctly. Handle this properly.
3631 Frequency : always on ppc64
3633 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3634 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3635 perform endianness conversion.
3638 Frequency : rarely (truncate to non-zero file size after write under load)
3639 Bugzilla : 10730, 10687
3640 Description: Files padded with zeros to next 4K multiple
3641 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3642 it is possible that files that are truncated to a non-zero size
3643 immediately after being written are filled with zero bytes beyond
3644 the truncated size. No file data is lost.
3646 Severity : enhancement
3648 Description: Allow recovery/failover for liblustre clients.
3649 Details : liblustre clients were unaware of failover configurations until
3652 Severity : enhancement
3654 Description: user file locks should fail when not mounting with flock option
3655 Details : Set up an error-returning stub in ll_file_operations.lock field
3656 to prevent incorrect behaviour when client is mounted without
3657 flock option. Also, set up properly f_op->flock field for
3661 Frequency : always on ia64
3663 Description: "lfs df" loops on printing out MDS statfs information
3664 Details : The obd_ioctl_data was not initialized and in some systems
3665 this caused a failure during the ioctl that did not return
3666 an error. Initialize the struct and return an error on failure.
3669 Frequency : SLES 9 only
3671 Description: Error of copying files with lustre special EAs as root
3672 Details : Client side always return success for setxattr call for lustre
3673 special xattr (currently only "trusted.lov").
3676 Frequency : rarely on clusters with both ia64+i386 clients
3678 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3679 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3680 clients are doing concurrent writes to the same file it is
3681 possible that the ia64 clients may LASSERT because the OST
3682 extent locks are not PAGE_SIZE aligned. Ensure that grown
3683 locks are always aligned on the request boundary.
3686 Frequency : specific use, occasional
3688 Description: Overwriting in use executable truncates on-disk binary image
3689 Details : If one node attempts to overwrite an executable in use by
3690 another node, we now correctly return ETXTBSY instead of
3691 truncating the file.
3693 Severity : enhancement
3695 Description: Async OSC create to avoid the blocking unnecessarily.
3696 Details : If a OST has no remain object, system will block on the creating
3697 when need to create a new object on this OST. Now, ways use
3698 pre-created objects when available, instead of blocking on an
3699 empty osc while others are not empty. If we must block, we block
3700 for the shortest possible period of time.
3705 Description: chmod on Lustre root is propagated to other clients
3706 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3707 invalid by the follow_mount time.
3710 Frequency : liblustre clients only
3712 Description: Race in 'instant cancel' lock handling could lead to such locks
3713 never to be granted in case of SMP MDS
3714 Details : Do not destroy not yet granted but cbpending locks in
3718 Frequency : replay/resend of open
3720 Description: non null lock assetion failure in mds_intent_policy
3721 Details : Trying to replay/resend lockless open requests resulted in
3722 mds_open() returning 0 with no lock. Now it sets a flag if
3723 a lock is going to be returned.
3725 Severity : enhancement
3727 Description: Checksum enhancements
3728 Details : New checksum enhancements allow for resending RPCs that failed
3731 Severity : enhancement
3733 Description: Tunables on number of dirty pages in cacche
3734 Details : Allow to set limit on number of dirty pages cached.
3739 Description: client crash on unmount - lock still has references
3740 Details : In some error handling cases it was possible to leak a lock
3741 reference on a client while accessing a file. This was not
3742 harmful to the client during operation, but would cause the
3743 client to crash when the filesystem is unmounted.
3746 Frequency : specific case, rare
3748 Description: ETXTBSY on mds though file not in use by client
3749 Details : ETXTBSY is no longer incorrectly returned when attempting to
3750 chmod or chown a directory that the user previously tried to
3751 execute or a currently-executing binary.
3754 Frequency : extremely rare except on liblustre-based clients
3756 Description: Lustre space not freed when files are deleted
3757 Details : Clean up open-unlinked files after client eviction. Previously
3758 the unlink was skipped and the files remained as orphans.
3763 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3764 Details : In some cases it was possible to send a blocking callback to
3765 a client doing a glimpse, even though that client didn't get
3766 a lock granted. When the glimpse lock is cancelled on the OST
3767 the freed lock is left on the waiting list and corrupted the list.
3770 Frequency : all core dumps
3772 Description: Broke core dumps to lustre
3773 Details : Negative dentry may be unhashed if parent does not have UPDATE
3774 lock, but some callers, e.g. do_coredump, expect dentry to be
3775 hashed after successful create, hash it in ll_create_it.
3777 ------------------------------------------------------------------------------
3779 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3781 * Support for kernels:
3782 2.6.9-42.0.2.EL (RHEL 4)
3783 2.6.5-7.276 (SLES 9)
3784 2.4.21-40.EL (RHEL 3)
3785 2.6.12.6 vanilla (kernel.org)
3789 Frequency : always on RHEL 3
3791 Description: Number of open files grows over time
3792 Details : The number of open files grows over time, whether or not
3793 Lustre is started. This was due to a filp leak introduced
3794 by one of our kernel patches.
3796 ------------------------------------------------------------------------------
3798 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
3800 * Support for kernels:
3801 2.6.9-42.EL (RHEL 4)
3802 2.6.5-7.267 (SLES 9)
3803 2.4.21-40.EL (RHEL 3)
3804 2.6.12.6 vanilla (kernel.org)
3809 Bugzilla : 5719, 9635, 9792, 9684
3810 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3811 Details : If a server is under heavy load and cannot reply to new
3812 connection requests before the client resends the (re)connect,
3813 the connection handling code can behave badly if two service
3814 threads are concurrently handing separate (re)connections from
3815 the same client. Add better locking to the connection handling
3816 code, and ensure that only a single connection will be processed
3817 for a given client UUID, even if the lock is dropped.
3819 Severity : enhancement
3821 Description: add TCP zero-copy support to kernel
3822 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3823 sends if the hardware supports scatter-gather and checksumming.
3824 This allows socklnd to do client-write and server-read more
3825 efficiently and reduce CPU utilization from skbuf copying.
3828 Frequency : only if NFS exporting from client
3830 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3831 Details : If setting attributes on a file created under NFS that had
3832 never been opened it would be possible to oops the client
3833 if the file had no objects.
3837 Bugzilla : 9326, 10402, 10897
3838 Description: client crash in ptlrpcd_wake() thread when sending async RPC
3839 Details : It is possible that ptlrpcd_wake() dereferences a freed async
3840 RPC. In rare cases the ptlrpcd thread alread processed the RPC
3841 before ptlrpcd_wake() was called and the request was freed.
3844 Frequency : always for liblustre
3846 Description: liblustre client does MDS+OSTs setattr RPC for each write
3847 Details : When doing a write from a liblustre client, the client
3848 incorrectly issued an RPC to the MDS and each OST the file was
3849 striped over in order to update the timestamps. When writing
3850 with small chunks and many clients this could overwhelm the MDS
3851 with RPCs. In all cases it would slow down the write because
3852 these RPCs are unnecessary.
3854 Severity : enhancement
3856 Description: allow number of MDS service threads to be changed at module load
3857 Details : It is now possible to change the number of MDS service threads
3858 running. Adding "options mds mds_num_threads={N}" to the MDS's
3859 /etc/modprobe.conf will set the number of threads for the next
3860 time Lustre is restarted (assuming the "mds" module is also
3861 reloaded at that time). The default number of threads will
3862 stay the same, 32 for most systems.
3867 Description: OST crash if filesystem is unformatted or corrupt
3868 Details : If an OST is started on a device that has never been formatted
3869 or if the filesystem is corrupt and cannot even mount then the
3870 error handling cleanup routines would dereference a NULL pointer.
3875 Description: NULL pointer deref in llap_from_page.
3876 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3877 incorrect) mapping if the page was truncated/reclaimed while it was
3878 searched for. Check for this condition and skip such pages when
3879 doing readahead. Introduce extra check to llap_from_page() to
3880 verify page->mapping->host is non-NULL (so page is not anonymous).
3883 Frequency : Sometimes when using sys_sendfile
3885 Description: "page not covered by a lock" warnings from ll_readpage
3886 Details : sendfile called ll_readpage without right page locks present.
3887 Now we introduced ll_file_sendfile that does necessary locking
3888 around call to generic_file_sendfile() much like we do in
3892 Frequency : with certain MDS communication failures at client mount time
3894 Description: NULL pointer deref after failed client mount
3895 Details : a client connection request may delayed by the network layer
3896 and not be sent until after the PTLRPC layer has timed out the
3897 request. If the client fails the mount immediately it will try
3898 to clean up before the network times out the request. Add a
3899 reference from the request import to the obd device and delay
3900 the cleanup until the network drops the request.
3903 Frequency : occasionally during client (re)connect
3905 Description: assertion failure during client (re)connect
3906 Details : processing a client connection request may be delayed by the
3907 client or server longer than the client connect timeout. This
3908 causes the client to resend the connection request. If the
3909 original connection request is replied in this interval, the
3910 client may trip an assertion failure in ptlrpc_connect_interpret()
3911 which thought it would be the only running connect process.
3914 Frequency : only with obd_echo servers and clients that are rebooted
3916 Description: kernel BUG accessing uninitialized data structure
3917 Details : When running an obd_echo server it did not start the ping_evictor
3918 thread, and when a client was evicted an uninitialized data
3919 structure was accessed. Start the ping_evictor in the RPC
3920 service startup instead of the OBD startup.
3922 Severity : enhancement
3923 Bugzilla : 10193 (patchless)
3924 Description: Remove dependency on various unexported kernel interfaces.
3925 Details : No longer need reparent_to_init, exit_mm, exit_files,
3926 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3929 Frequency : rare (only users of deprecated and unsupported LDAP config)
3931 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3932 Details : LDAP apparently contains 'lustreName' attributes instead of
3933 'name'. A simple remapping of the name is sufficient.
3936 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3938 Description: waiting_locks_callback trips kernel BUG if client is evicted
3939 Details : Running with the dump_on_timeout debug flag turned on makes
3940 it possible that the waiting_locks_callback() can try to dump
3941 the Lustre kernel debug logs from an interrupt handler. Defer
3942 this log dumping to the expired_lock_main() thread.
3944 Severity : enhancement
3946 Description: Support NFS exporting on 2.6 kernels.
3947 Details : Implement non-rawops metadata methods for NFS server to use without
3948 changing NFS server code.
3951 Frequency : very rare (synthetic metadata workload only)
3953 Description: two racing renames might cause an MDS thread to deadlock
3954 Details : Running the "racer" program may cause one MDS thread to rename
3955 a file from being the source of a rename to being the target of
3956 a rename at exactly the same time that another thread is doing
3957 so, and the second thread has already enqueued these locks after
3958 doing a lookup of the target and is trying to relock them in
3959 order. Ensure that we don't try to re-lock the same resource.
3962 Frequency : only very large systems with liblustre clients
3964 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3965 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3966 all OSTs in parallel. This allows the eviction of stale liblustre
3967 clients to proceed much faster than if they were done in series,
3968 and also offers similar improvements for other set_info RPCs.
3973 Description: excessive CPU usage during initial read phase on client
3974 Details : During the initial read phase on a client, it would agressively
3975 retry readahead on the file, consuming too much CPU and impacting
3976 performance (since 1.4.5.8). Improve the readahead algorithm
3977 to avoid this, and also improve some other common cases (read
3978 of small files in particular, where "small" is files smaller than
3979 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3984 Description: MDS crash when receiving packet with unknown intent.
3985 Details : Do not LBUG in unknown intent case, just return -EFAULT
3987 Severity : enhancement
3988 Bugzilla : 9293, 9385
3989 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3990 Details : Do not serialize getattr (non-intent version) and statfs.
3993 Frequency : occasional, when OST network is overloaded/intermittent
3995 Description: client evicted by OST after bulk IO timeout
3996 Details : If a client sends a bulk IO request (read or write) the OST
3997 may evict the client if it is unresposive to its data GET/PUT
3998 request. This is incorrect if the network is overloaded (takes
3999 too long to transfer the RPC data) or dropped the OST GET/PUT
4000 request. There is no need to evict the client at all, since
4001 the pinger and/or lock callbacks will handle this, and the
4002 client can restart the bulk request.
4005 Frequency : Always when mmapping file with no objects
4007 Description: client crashes when mmapping file with no objects
4008 Details : Check that we actually have objects in a file before doing any
4009 operations on objects in ll_vm_open, ll_vm_close and
4015 Description: Request leak when working with deleted CWD
4016 Details : Introduce advanced request refcount tracking for requests
4017 referenced from lustre intent.
4019 Severity : Enhancement
4021 Description: Cache open file handles on client.
4022 Details : MDS now will return special lock along with openhandle, if
4023 requested and client is allowed to hold openhandle, even if unused,
4024 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4025 closing files for every read/write openration.
4027 Severity : Enhancement
4029 Description: Cache open negative dentries on client when possible.
4030 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4031 negative dentries on lock revocation.
4036 Description: Remounting a client read-only wasn't possible with a zconf mount
4037 Details : It wasn't possible to remount a client read-only with llmount.
4039 Severity : enhancement
4040 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4041 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4042 MPICH is included to add Lustre-specific ADIO interfaces.
4043 This is based closely on the UFS ADIO layer and only differs
4044 in file creation, in order to allow the OST striping to be set.
4045 This is user-contributed code and not supported by CFS.
4050 Description: extended inode attributes (immutable, append-only) work improperly
4051 when 2.4 and 2.6 kernels are used on client/server or vice versa
4052 Details : Introduce kernel-independent values for these flags.
4054 Severity : enhancement
4057 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4058 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4059 granularity. Now, Fractional MB granularity is be supported,
4060 this is very useful for low memory system.
4062 Severity : enhancement
4064 Description: Getattr by fid
4065 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4066 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4067 have locks and avoids getattr rpc requests for it on every stat.
4070 Frequency : Always, for filesystems larger than 2TB
4072 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4073 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4074 filesystems larger than 2TB to be mounted with mballoc enabled.
4077 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4079 Description: filesystem corruption for non-standard kernels and very large OSTs
4080 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4081 filesystems larger than 2TB is mounted then the kernel will
4082 silently corrupt the start of the filesystem. CONFIG_LBD is
4083 enabled for all CFS-supported kernels, but the possibility of
4084 this happening with a modified kernel config exists.
4086 Severity : enhancement
4088 Description: add client O_DIRECT support for 2.6 kernels
4089 Details : It is now possible to do O_DIRECT reads and writes to files
4090 in the Lustre client mountpoint on 2.6 kernel clients.
4092 Severity : enhancement
4094 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4095 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4096 parallel, not waiting for response from every OST before sending
4097 a rpc to the next OST.
4102 Description: setattr vs write race when updating file timestamps
4103 Details : Client processes that update a file timestamp into the past
4104 right after writing to the file (e.g. tar) it is possible that
4105 the updated file modification time can be reset to the current
4106 time due to a race between processing the setattr and write RPC.
4108 Severity : enhancement
4110 Description: Bring 'lfs find' closer in line with regular Linux find.
4111 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4112 -print0 options and obtains all the needed info through the lustre
4115 Severity : enhancement
4117 Description: support up to 1024 configured devices on one node
4118 Details : change obd_dev array from statically allocated to dynamically
4119 allocated structs as they are first used to reduce memory usage
4124 Description: Flush dirty partially truncated pages during truncate
4125 Details : Immediatelly flush partially truncated pages in filter_setattr,
4126 this way we completely avoid having any pages in page cache on OST
4127 and can retire ugly workarounds during writes to flush such pages.
4132 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4133 Details : protect inode from truncation within vfs_unlink() context
4134 just take a reference before calling vfs_unlink() and release it
4135 when parent's i_sem is free.
4138 Frequency : always, if extents are used on OSTs
4140 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4141 Details : OSTs using the extents format would not zero the high 16 bits of
4142 the index physical block number. This is not a problem for any
4143 OST filesystems smaller than 16TB, and no kernels support ext3
4144 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4145 new/modified files) and can be fixed for existing filesystems
4146 with e2fsprogs-1.39-cfs1.
4151 Description: import connection selection may be incorrect if timer wraps
4152 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4153 import connections to be ignored if the 32-bit jiffies counter
4154 wraps. Use a 64-bit jiffies counter.
4157 Frequency : during server recovery
4159 Description: crash after server is denying duplicate export
4160 Details : If clients are resending connect requests to the server, the
4161 server refuses to allow a client to connect multiple times.
4162 Fixed a bug in the handling of this case.
4165 Frequency : very large clusters immediately after boot
4167 Description: LNET request buffers exhausted under heavy short-term load
4168 Details : If a large number of client requests are generated on a service
4169 that has previously never seen so many requests it is possible
4170 that the request buffer growth cannot keep up with the spike in
4171 demand. Instead of dropping incoming requests, they are held in
4172 the LND until the RPC service can accept more requests.
4175 Frequency : Sometimes during replay
4177 Description: Assertion failure in ll_local_open after replay.
4178 Details : If replay happened on an open request reply before we were able
4179 to set replay handler, reply will become not swabbed tripping the
4180 assertion in ll_local_open. Now we set the handler right after
4181 recognising of open request
4184 Frequency : very rare
4186 Description: kernel reports "badness in vsnprintf"
4187 Details : Reading from the "recovery_status" /proc file in small chunks
4188 may cause a negative length in lprocfs_obd_rd_recovery_status()
4189 call to vsnprintf() (which is otherwise harmless). Exit early
4190 if there is no more space in the output buffer.
4192 Severity : enhancement
4194 Description: clear OBD RPC statistics by writing to them
4195 Details : It is now possible to clear the OBD RPC statistics by writing
4196 to the "stats" file.
4201 Description: Client mtime is not the same on different clients after utimes
4202 Details : In some cases, the client was using the utimes() syscall on
4203 a file cached on another node. The clients now validate the
4204 ctime from the MDS + OSTs to determine which one is right.
4209 Description: Inability to activate failout mode
4210 Details : lconf script incorrectly assumed that in python string's numeric
4211 value is used in comparisons.
4214 Frequency : always with multiple stripes per file
4216 Description: Inefficient object allocation for mutli-stripe files
4217 Details : When selecting which OSTs to stripe files over, for files with
4218 a stripe count that divides evenly into the number of OSTs,
4219 the MDS is always picking the same starting OST for each file.
4220 Return the OST selection heuristic to the original design.
4225 Description: mount failures may take full timeout to return an error
4226 Details : Under some heavy load conditions it is possible that a
4227 failed mount can wait for the full obd_timeout interval,
4228 possibly several minutes, before reporting an error.
4229 Instead return an error as soon as the status is known.
4231 ------------------------------------------------------------------------------
4233 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
4235 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4236 INTEROPERATE with older versions automatically. Please read the
4237 user documentation before upgrading any part of a live system.
4238 * WARNING: Lustre networking configuration changes are required with
4239 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4242 * Support for kernels:
4243 2.6.9-22.0.2.EL (RHEL 4)
4244 2.6.5-7.244 (SLES 9)
4245 2.6.12.6 vanilla (kernel.org)
4248 Severity : enhancement
4249 Bugzilla : 7981/8208
4250 Description: Introduced Lustre Networking (LNET)
4251 Details : LNET is new networking infrastructure for Lustre, it includes
4252 a reorganized network configuration mode (see the user
4253 documentation for full details) as well as support for routing
4254 between different network fabrics. Lustre Networking Devices
4255 (LNDs) for the supported network fabrics have also been
4256 created for this new infrastructure.
4258 Severity : enhancement
4259 Description: Introduced Access control lists
4260 Details : clients can set ACLs on files and directories in order to have
4261 more fine-grained permissions than the standard Unix UGO+RWX.
4262 The MDS must be started with the "-o acl" mount option.
4264 Severity : enhancement
4265 Description: Introduced filesystem quotas
4266 Details : Administrators may now establish per-user quotas on the
4269 Severity : enhancement
4271 Description: Configuration change for the XT3
4272 The PTLLND is now used to run Lustre over Portals on the XT3
4273 The configure option(s) --with-cray-portals are no longer used.
4274 Rather --with-portals=<path-to-portals-includes> is used to
4275 enable building on the XT3. In addition to enable XT3 specific
4276 features the option --enable-cray-xt3 must be used.
4281 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4282 Details : A race between allocating a new llcd and re-getting the llcd_lock
4283 allowed another thread to grab newly-allocated llcd.
4285 Severity : enhancement
4287 Description: 2.6 OST async journal commit and locking fix to improve performance
4288 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4289 2.6 kernels are now async as they already were in 2.4 kernels so
4290 that they can commit concurrently with the network bulk transfer.
4291 For block-allocated files the filter allocation semaphore is held
4292 to avoid filesystem fragmentation during allocation. BKL lock
4293 removed for 2.6 xattr operations where it is no longer needed.
4298 Description: lconf incorrectly determined whether two IP networks could talk
4299 Details : In some more complicated routing and multiple-network
4300 configurations, lconf will avoid trying to make a network
4301 connection to a disjoint part of the IP space. It was doing the
4302 math incorrectly for one set of cases.
4307 Description: Fix for potential infinite loop processing records in an llog.
4308 Details : If an llog record is corrupted/zeroed, it is possible to loop
4309 forever in llog_process(). Validate the llog record length
4310 and skip the remainder of the block on error.
4313 Frequency : occasional (liblustre only)
4315 Description: liblustre could not open files whose last component is a symlink
4316 Details : sysio_path_walk() would incorrectly pass the open intent to
4317 intermediate path components.
4320 Frequency : rare (liblustre only with non-standard tuning)
4321 Bugzilla : 7201 (7350)
4322 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4323 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4324 after it has been released, only do this for referenced locks
4326 Severity : enhancement
4328 Description: specifying an (invalid) directory default stripe_size of -1
4329 would reset the directory default striping
4330 Details : stripe_size -1 was used internally to signal directory stripe
4331 removal, now use "all default" to signal dir stripe removal
4332 as a directory striping of "all default" is not useful
4335 Frequency : common for large clusters running liblustre clients
4337 Description: doing an ls when liblustre clients are running is slow
4338 Details : sending a glimpse AST to a liblustre client waits for every AST
4339 to time out, as liblustre clients will not respond. Since they
4340 cannot cache data we refresh the OST lock LVB from disk instead.
4342 Severity : enhancement
4344 Description: doing an ls at the same time as file IO can be slow
4345 Details : enqueue and other "small" requests can be blocked behind many
4346 large IO requests. Create a new OST IO portal for non-IO
4347 requests so they can be processed faster.
4350 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4352 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4353 created file with mode 000
4354 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4355 allow an ftruncate() as a non-root user to a file with mode 000.
4356 Lustre now respects this flag to disable mode checks when
4357 truncating a file owned by the user
4360 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4363 Description: Revoking locks from clients that went dead or catatonic might take
4365 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4366 cancellation of such locks instant on servers without waiting for
4367 any reply from clients. Clients drops these locks when cancel
4368 notification from server is received without replying.
4371 Frequency : liblustre-only, when liblustre client dies or becomes busy
4373 Description: Doing ls on Linux clients can take a long time with active
4375 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4376 granting such locks to it in the first place if possible. Locks
4377 are taken by proxy on the OST during the read or write and
4378 dropped immediately afterward. Add connect flags handling, do
4379 not grant locks to liblustre clients for glimpse ASTs.
4381 Severity : enhancement
4383 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4384 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4385 sequential reads and behaves badly with applications doing random
4386 reads. Improve it by reading ahead at least read region, and
4387 avoiding excessive large RPC for small reads.
4389 Severity : enhancement
4391 Description: Creating more than 1000 files for a single job may cause a load
4392 imbalance on the OSTs if there are also a large number of OSTs.
4393 Details : qos_prep_create() uses an OST index reseed value that is an
4394 even multiple of the number of available OSTs so that if the
4395 reseed happens in the middle of the object allocation it will
4396 still utilize the OSTs as uniformly as possible.
4401 Description: OST or MDS may oops in ping_evictor_main()
4402 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4403 but doesn't restart at beginning of obd_exports_timed list
4406 Severity : enhancement
4408 Description: improve by-nid export eviction on the MDS and OST
4409 Details : allow multiple exports with the same NID to be evicted at one
4410 time without re-searching the exports list.
4413 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4415 Description: MDS may oops in groups_free()
4416 Details : in rare race conditions a newly allocated group_info struct is
4417 freed again, and this can be NULL. The 2.4 compatibility code
4418 for groups_free() checked for a NULL pointer, but 2.6 did not.
4421 Frequency : common for liblustre clients doing little filesystem IO
4422 Bugzilla : 9352, 7313
4423 Description: server may evict liblustre clients accessing contended locks
4424 Details : if a client is granted a lock or receives a completion AST
4425 with a blocking AST already set it would not reply to the AST
4426 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4429 Frequency : lfs setstripe, only systems with more than 160 OSTs
4431 Description: unable to set striping with a starting offset beyond OST 160
4432 Details : llapi_create_file() incorrectly limited the starting stripe
4433 index to the maximum single-file stripe count.
4436 Frequency : LDAP users only
4438 Description: lconf did not handle in-kernel recovery with LDAP properly
4439 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4441 Severity : enhancement
4443 Description: bind OST threads to NUMA nodes to improve performance
4444 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4445 node and do their allocations there to localize memory access
4447 Severity : enhancement
4449 Description: llmount can determine client NID directly from Myrinet (GM)
4450 Details : the client NID code from gmnalnid was moved directly into
4451 llmount, removing the need to use this or specifying the
4452 client NID explicitly when mounting GM clients with zeroconf
4455 Frequency : if client is started with down MDS
4457 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4458 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4459 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4460 ptlrpc_import_delayed_req() because we didn't check if the
4461 request was interrupted, and we also didn't break out of the
4462 event loop if there was no timeout
4467 Description: data loss during non-page-aligned writes to a single file from
4468 both multiple nodes and multiple threads on one node at same time
4469 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4470 inconsistency led to false short-reads, that were cached and later
4471 used by ->prepare_write() to fill in partially written page,
4472 leading to data loss.
4475 Frequency : always, if lconf --abort_recovery used
4477 Description: lconf --abort_recovery fails with 'Operation not supported'
4478 Details : lconf was attempting to abort recovery on the MDT device and not
4481 Severity : enhancement
4483 Description: remove cleanup logs
4484 Details : replace lconf-generated cleanup logs with lustre internal
4485 cleanup routines. Eliminates the need for client-cleanup and
4488 Severity : enhancement
4490 Description: add support for EAs (user and system) on lustre filesystems
4491 Details : it is now possible to store extended attributes in the Lustre
4492 client filesystem, and with the user_xattr mount option it
4493 is possible to allow users to store EAs on their files also
4495 Severity : enhancement
4497 Description: Add possibility (config option) to show minimal available OST free
4499 Details : When compiled with --enable-mindf configure option, statfs(2)
4500 (and so, df) will return least minimal free space available from
4501 all OSTs as amount of free space on FS, instead of summary of
4502 free spaces of all OSTs.
4504 Severity : enhancement
4506 Description: do not expand extent locks acquired on OST-side
4507 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4508 by server: they are not cached anyway.
4511 Frequency : when mmap is used/binaries executed from Lustre
4513 Description: Unmmap pages before throwing them away from read cache.
4514 Details : llap_shrink cache now attempts to unmap pages before discarding
4515 them (if unmapping failed - do not discard). SLES9 kernel has
4516 extra checks that trigger if this unmapping is not done first.
4521 Description: lconf didn't resolve symlinks before checking to see whether a
4522 given mountpoint was already in use
4525 Frequency : when migrating failover services
4526 Bugzilla : 6395, 9514
4527 Description: When migrating a subset of services from a node (e.g. failback
4528 from a failover service node) the remaining services would
4529 time out and evict clients.
4530 Details : lconf --force (implied by --failover) sets the global obd_timeout
4531 to 5 seconds in order to quickly disconnect, but this caused
4532 other RPCs to time out too quickly. Do not change the global
4533 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4535 Severity : enhancement
4536 Frequency : if MDS is started with down OST
4537 Bugzilla : 9439,5706
4538 Description: Allow startup/shutdown of an MDS without depending on the
4539 availability of the OSTs.
4540 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4541 Add appropriate locking and lov-osc refcounts for safe
4542 cleaning. Add osc abort_inflight calls in case the
4543 synchronize never started.
4546 Frequency : occasional (Cray XT3 only)
4548 Description: root not authorized to access files in CRAY_PORTALS environment
4549 Details : The client process capabilities were not honoured on the MDS in
4550 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4551 been accessed by an authorized user then root was able to access
4552 the file on the local client also. The root user capabilities
4553 are now allowed on the MDS, as this environment has secure UID.
4556 Frequency : occasional
4558 Description: ldiskfs "too long searching" message happens too often
4559 Details : A debugging message (otherwise harmless) prints too often on
4560 the OST console. This has been reduced to only happen when
4561 there are fragmentation problems on the filesystem.
4566 Description: Division by zero in statfs when all OSCs are inactive
4567 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4568 lov_statfs divides by value returned by lov_get_stripecnt().
4572 Bugzilla : 9489, 3273
4573 Description: First write from each client to each OST was only 4kB in size,
4574 to initialize client writeback cache, which caused sub-optimal
4575 RPCs and poor layout on disk for the first writen file.
4576 Details : Clients now request an initial cache grant at (re)connect time
4577 and so that they can start streaming writes to the cache right
4578 away and always do full-sized RPCs if there is enough data.
4579 If the OST is rebooted the client also re-establishes its grant
4580 so that client cached writes will be honoured under the grant.
4585 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4586 Details : Now I/O RPCs go to different portal number and (presumably) fast
4587 lock requests (and glimses) and other RPCs get their own service
4588 threads pool that should be able to service those RPCs
4591 Severity : enhancement
4593 Description: Ability to exchange lustre version between client and servers and
4594 issue warnings at client side if client is too old. Also for
4595 liblustre clients there is ability to refuse connection of too old
4597 Details : New 'version' field is added to connect data structure that is
4598 filled with version info. That info is later checked by server and
4602 Frequency : rare, liblustre only.
4603 Bugzilla : 9296, 9581
4604 Description: Two simultaneous writes from liblustre at offset within same page
4605 might proceed at the same time overwriting eachother with stale
4607 Details : I/O lock withing llu_file_prwv was released too early, before data
4608 actually was hitting the wire. Extended lock-holding time until
4609 server acknowledges receiving data.
4612 Frequency : extremely rare. Never observed in practice.
4614 Description: avoid generating lustre_handle cookie of 0.
4615 Details : class_handle_hash() generates handle cookies by incrementing
4616 global counter, and can hit 0 occasionaly (this is unlikely, but
4617 not impossible, because initial value of cookie counter is
4618 selected randonly). Value of 0 is used as a sentinel meaning
4619 "unassigned handle" --- avoid it. Also coalesce two critical
4620 sections in this function into one.
4622 Severity : enhancement
4624 Description: allow liblustre clients to delegate truncate locking to OST
4625 Details : To avoid overhead of locking, liblustre client instructs OST to
4626 take extent lock in ost_punch() on client's behalf. New connection
4627 flag is added to handle backward compatibility.
4629 Severity : enhancement
4630 Bugzilla : 4928, 7341, 9758
4631 Description: allow number of OST service threads to be specified
4632 Details : a module parameter allows the number of OST service threads
4633 to be specified via "options ost ost_num_threads={N}" in the
4634 OSS's /etc/modules.conf or /etc/modprobe.conf.
4638 Bugzilla : 6146, 9635, 9895
4639 Description: servers crash with bad pointer in target_handle_connect()
4640 Details : In rare cases when a client is reconnecting it was possible that
4641 the connection request was the last reference for that export.
4642 We would temporarily drop the export reference and get a new
4643 one, but this may have been the last reference and the export
4644 was just destroyed. Get new reference before dropping old one.
4646 Severity : enhancement
4647 Frequency : if client is started with failover MDS
4649 Description: Allow multiple MDS hostnames in the mount command
4650 Details : Try to read the configuration from all specified MDS
4651 hostnames during a client mount in case the "primary"
4654 Severity : enhancement
4656 Description: Stop sending data to evicted clients as soon as possible.
4657 Details : Check if the client we are about to send or are sending data to
4658 was evicted already. (Check is done every second of waiting,
4659 for which l_wait_event interface was extended to allow checking
4660 of exit condition at specified intervals).
4663 Frequency : rare, normally only when NFS exporting is done from client
4665 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4667 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4670 Frequency : always, if srand() is called before liblustre initialization
4672 Description: Liblustre uses system PRNG disturbing its usage by user application
4673 Details : Introduce internal to lustre fast and high-quality PRNG for
4674 lustre usage and make liblustre and some other places in generic
4675 lustre code to use it.
4677 Severity : enhancement
4678 Bugzilla : 9477, 9557, 9870
4679 Description: Verify that the MDS configuration logs are updated when xml is
4680 Details : Check if the .xml configuration logs are newer than the config
4681 logs stored on the MDS and report an error if this is the case.
4682 Request --write-conf, or allow starting with --old_conf.
4684 Severity : enhancement
4686 Description: Handle symlinks in the path when checking if Lustre is mounted.
4687 Details : Resolve intermediate symlinks when checking if a client has
4688 mounted a filesystem to avoid duplicate client mounts.
4693 Description: lconf can hit an error exception but still return success.
4694 Details : The lconf command catches the Command error exception at the top
4695 level script context and will exit with the associated exit
4696 status, but doesn't ensure that this exit status is non-zero.
4701 Description: failure of ptlrpc thread startup can cause oops
4702 Details : Starting a ptlrpc service thread can fail if there are a large
4703 number of threads or the server memory is very fragmented.
4704 Handle this without oopsing.
4707 Frequency : always, only if liblustre and non-default acceptor port was used
4709 Description: liblustre cannot connect to servers with non-default acceptor port
4710 Details : tcpnal_set_default_params() was not called and was therefore
4711 ignoring the environment varaible TCPNAL_PORT, as well as other
4712 TCPNAL_ environment variables
4717 Description: two objects could be created on the same OST for a single file
4718 Details : If an OST is down, in some cases it was possible to create two
4719 objects on a single OST for a single file. No problems other
4720 than potential performance impact and spurious error messages.
4724 Bugzilla : 5681, 9562
4725 Description: Client may oops in ll_unhash_aliases
4726 Details : Client dcache may become inconsistent in race condition.
4727 In some cases "getcwd" can fail if the current directory is
4733 Description: Inode refcounting problems in NFS export code
4734 Details : link_raw functions used to call d_instantiate without obtaining
4735 extra inode reference first.
4739 Bugzilla : 9942, 9903
4740 Description: Referencing freed requests leading to crash, memleaks with NFS.
4741 Details : We used to require that call to ll_revalidate_it was always
4742 followed by ll_lookup_it. Also with revalidate_special() it is
4743 possible to call ll_revalidate_it() twice for the same dentry
4744 even if first occurence returned success. This fix changes semantic
4745 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4746 reference on a request referred from the intent.
4747 ll_intent_release() then releases such a request.
4750 Frequency : rare, normally benchmark loads only
4752 Description: unlinked inodes were kept in memory on the client
4753 Details : If a client is repeatedly creating and unlinking files it
4754 can accumulate a lot of stale inodes in the inode slab cache.
4755 If there is no other client load running this can cause the
4756 client node to run out of memory. Instead flush old inodes
4757 from client cache that have the same inode number as a new inode.
4760 Frequency : SLES9 2.6.5 kernel and long filenames only
4761 Bugzilla : 9969, 10379
4762 Description: utime reports stale NFS file handle
4763 Details : SLES9 uses out-of-dentry names in some cases, which confused
4764 the lustre dentry revalidation. Change it to always use the
4768 Frequency : rare, unless heavy write-truncate concurrency is continuous
4769 Bugzilla : 4180, 6984, 7171, 9963, 9331
4770 Description: OST becomes very slow and/or deadlocked during object unlink
4771 Details : filter_destroy() was holding onto the parent directory lock
4772 while truncating+unlinking objects. For very large objects this
4773 may block other threads for a long time and slow overall OST
4774 responsiveness. It may also be possible to get a lock ordering
4775 deadlock in this case, or run out of journal credits because of
4776 the combined truncate+unlink. Solution is to do object truncate
4777 first in one transaction without parent lock, and then do the
4778 final unlink in a new transaction with the parent lock. This
4779 reduces the lock hold time dramatically.
4782 Frequency : rare, 2.4 kernels only
4784 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4785 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4786 before unmounting their filesystem in order to avoid deadlock.
4787 The kernel_locked() function in 2.4 kernels only checks whether
4788 the kernel lock is held, not whether it is this process that is
4789 holding it as 2.6 kernels do.
4794 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4795 Details : The client ptlrpc code may be trying to reconnect to a down
4796 server before a previous connection attempt has timed out.
4797 Increase the reconnect interval to be longer than the connection
4798 timeout interval to avoid sending duplicate connections to
4802 Frequency : echo_client brw_test command
4804 Description: fix echo_client to work with OST preallocated code
4805 Details : OST preallocation code (5137) didn't take echo_client IO path
4806 into account: echo_client calls filter methods outside of any
4807 OST thread and, hence, there is no per-thread preallocated
4808 pages and buffers to use. Solution: hijack pga pages for IO. As
4809 a byproduct, this avoids unnecessary data copying.
4813 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4814 Description: Client can oops in mdc_commit_close() after open replay
4815 Details : It was possible for the MDS to return an open request with no
4816 transaction number in mds_finish_transno() if the client was
4817 evicted, but without actually returning an error. Clients
4818 would later try to replay that open and may trip an assertion
4819 Simplify the client close codepath, and always return an error
4820 from the MDS in case the open is not successful.
4823 Frequency : rare, 2.6 OSTs only
4825 Description: OST may deadlock under high load on fragmented files
4826 Details : If there was a heavy load and highly-fragmented OST filesystems
4827 it was possible to have all the OST threads deadlock waiting on
4828 allocation of biovecs, because the biovecs were not released
4829 until the entire RPC IO was completed. Instead, release biovecs
4830 as soon as they are complete to ensure forward IO progress.
4832 Severity : enhancement
4834 Description: Support for specifying external journal device at mount
4835 Details : If an OST or MDS device is formatted with an external journal
4836 device, this device major/minor is stored in the ext3 superblock
4837 and may not be valid for failover. Allow detecting and
4838 specifying the external journal at mount time.
4843 Description: Mounting an MDS with pending unlinked files may cause oops
4844 Details : target_finish_recovery() calls mds_postrecov() which returned
4845 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4846 considers this an error and immediately begins cleaning up the
4847 lov, just after starting the mds_lov process
4849 Severity : enhancement
4851 Description: Implement 'lfs df' to report actual free space on per-OST basis
4852 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4853 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4854 usage of objects; '-h' to report in human readable format.
4856 ------------------------------------------------------------------------------
4858 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
4865 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4866 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4867 at startup during orphan unlinking. Instead, skip buddy update
4868 before setup, as it will be regenerated after recovery is complete.
4873 Description: If an OST is inactive, its locks might reference stale inodes.
4874 Details : lov_change_cbdata() must iterate over all namespaces, even if
4875 they are inactive to clear inode references from the lock.
4877 Severity : enhancement
4878 Frequency : occasional, if non-standard max_dirty_mb used
4880 Description: Client will block write RPCs if not enough grant
4881 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4882 then the client will block writes while waiting for another RPC
4883 to complete instead of consuming its dirty limit. With change
4884 we get improved performance when max_dirty_mb is small.
4886 Severity : enhancement
4887 Bugzilla : 3389, 6253
4888 Description: Add support for supplementary groups on the MDS.
4889 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4890 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4891 lookups for user supplementary groups into a cache.
4895 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4896 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4897 for liblustre O_EXCL is enforced by MDS.
4901 Description: readdir never returns NULL in liblustre.
4902 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4903 dentry in struct dirent.
4907 Frequency : liblustre only, depends on application IO pattern
4908 Description: liblustre clients evicted if not contacting servers
4909 Details : Don't put liblustre clients into the ping_evictor list, so
4910 they will not be evicted by the pinger ever.
4912 Severity : enhancement
4914 Description: Add ability to evict clients by NID from MDS.
4915 Details : By echoing "nid:$NID" string into
4916 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4917 $NID would be instantly evicted from this MDS and from all active
4918 OSTs connected to it.
4922 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4923 Details : lookup_it_finish() used to query file size from OSTs that was not
4928 Description: service threads change working directory to that of init
4929 Details : Starting lustre service threads may pin the working directory
4930 of the parent thread, making that filesystem busy. Threads
4931 now change to the working directory of init to avoid this.
4935 Frequency : during shutdown only
4936 Description: shutdown with a failed MDS or OST can cause unmount to hang
4937 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4942 Frequency : 2.6 only
4943 Description: chmod/chown may include an extra supplementary group
4944 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4945 supplementary group and if none is specified this is used.
4948 Bugzilla : 5479 (6816)
4950 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4951 Details : If lookup is in progress on a file that is unlinked we might try
4952 to revalidate the inode and fail in revalidate after lookup is
4953 complete and ll_file_open() enqueues the open again but
4954 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4957 Frequency : always, if lconf --abort_recovery used
4959 Description: lconf --abort_recovery fails with 'Operation not supported'
4960 Details : lconf was attempting to abort recovery on the MDT device and not
4963 ------------------------------------------------------------------------------
4965 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4970 Frequency : rare (only unsupported configurations with a node running as an
4972 Bugzilla : 6514, 5137
4973 Description: Mounting a Lustre file system on a node running as an OST could
4975 Details : OSTs now preallocates memory needed to write out data at
4976 startup, instead of when needed, to avoid having to
4977 allocate memory in possibly low memory situations.
4978 Specifically, if the file system is mounted on on OST,
4979 memory pressure could force it to try to write out data,
4980 which it needed to allocate memory to do. Due to the low
4981 memory, it would be unable to do so and the node would
4982 become unresponsive.
4984 Severity : enhancement
4986 Description: Addition of lconf --service command line option
4987 Details : lconf now accepts a '--service <arg>' option, which is
4988 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4990 Severity : enhancement
4992 Description: Failover mode is now the default for OSTs.
4993 Details : By default, OSTs will now run in failover mode. To return to
4994 the old behaviour, add '--failout' to the lmc line for OSTs.
4996 Severity : enhancement
4998 Description: Health checks are now provided for MDS and OSTs
4999 Details : Additional detailed health check information on MSD and OSTs
5000 is now provided through the procfs health_check value.
5003 Frequency : occasional, depends on IO load
5005 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5006 after numerous create/delete cycles
5007 Details : The ext3 inode allocation policy would not allocate new inodes
5008 very well on the OSTs because there are no new directories
5009 being created. Instead we look for groups with free space if
5010 the parent directories are nearly full.
5015 Description: Network or server problems during mount may cause partially
5016 mounted clients instead of returning an error.
5017 Details : The config llog parsing code may overwrite the error return
5018 code during mount error handling, returning success instead
5024 Description: MDS can fail to allocate large reply buffers
5025 Details : After long uptimes the MDS can fail to allocate large reply
5026 buffers (e.g. zconf client mount config records) due to memory
5027 fragmentation or consumption by the buffer cache. Preallocate
5028 some large reply buffers so that these replies can be sent even
5029 under memory pressure.
5033 Frequency : rare (liblustre)
5034 Description: fsx running with liblustre complained that using truncate() to
5035 extend the file doesn't work. This patch corrects that issue.
5036 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5037 ATTR_SIZE and lsm use in llu_setattr_raw.
5041 Frequency : rare, only 2.6 kernels
5042 Description: Unusual file access patterns on the MDS may result in inode
5043 data being lost in very rare circumstances.
5044 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5045 code in the 2.6 kernel caused the inode and/or EA data not to
5046 be read from disk, causing single-file corruption.
5050 Frequency : rare, only 2.6 filesystems using extents
5051 Description: Heavy concurrent write and delete load may cause data corruption.
5052 Details : It was possible under high-load situations to have an extent
5053 metadata block in the block device cache from a just-unlinked
5054 file overwrite a newly-allocated data block. We now unmap any
5055 metadata buffers that alias just-allocated data blocks.
5059 Frequency : filesystems with default stripe_count larger than 77
5060 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5061 Details : lconf specifies an inode size of 4096 bytes when the default
5062 stripe_count is larger than 77. This conflicts with the default
5063 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5064 this case to avoid pinning too much memory for large EAs.
5066 ------------------------------------------------------------------------------
5068 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5073 Frequency : rare (extremely heavy IO load with hundreds of clients)
5075 Description: Client is evicted, gets IO error writing to file
5076 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5077 caused clients to be evicted for AST timeouts. The fixes in
5078 bug 5192 mean we no longer need to have such short AST timeouts
5079 so ldlm_timeout has been increased.
5082 Frequency : occasional during --force or --failover shutdown under load
5083 Bugzilla : 5949, 4834
5084 Description: Server oops/LBUG if stopped with --force or --failover under load
5085 Details : a collection of import/export refcount and cleanup ordering
5086 issues fixed for safer force cleanup
5089 Frequency : only filesystems larger than 120 OSTs
5090 Bugzilla : 5990, 6223
5091 Description: lfs getstripe would oops on a very large filesystem
5092 Details : lov_getconfig used kfree on vmalloc'd memory
5095 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5096 Bugzilla : 6242, 6243
5097 Description: reading from files that had been truncated to a non-zero size
5098 but never opened returned no data
5099 Details : ll_file_read() reads zeros from no-object files to EOF
5104 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5105 Details : There's little harm in aborting MDS/OSS recovery and letting it
5106 try again, so I removed the LASSERT and return an error instead.
5108 Severity : enhancement
5110 Description: New debugging infrastructure for tracking down data corruption
5111 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5112 (b) cover more of the client-side code path, and (c) try to narrow
5113 down where problems occurred
5117 Bugzilla : 3819, 4364, 4397, 6313
5118 Description: Racing close and eviction MDS could cause assertion in mds_close
5119 Details : It was possible to get multiple mfd references during close and
5120 client eviction, leading to one thread referencing a freed mfd.
5122 Severity: : enhancement
5123 Bugzilla : 3262, 6359
5124 Description: Attempts to reconnect to servers are now more aggressive.
5125 Details : This builds on the enhanced upcall-less recovery that was added
5126 in 1.4.2. When trying to reconnect to servers, clients will
5127 now try each server in the failover group every 10 seconds. By
5128 default, clients would previously try one server every 25 seconds.
5133 Description: After recovery, certain operations trigger a failed
5134 assertion on a client.
5135 Details : Failing over an mds, using lconf -d --failover, while a
5136 client was doing a readdir() call would cause the client to
5137 LBUG after recovery completed and the readdir() was resent.
5139 Severity : enhancement
5141 Description: Default groups are now added by lconf
5142 Details : You can now run lconf --group <servicename> without having to
5143 manually add groups with lmc.
5146 Frequency : occasional
5148 Description: Nodes with an elan id of 0 trigger a failed assertion
5151 Frequency : always when accessing e.g. tty/console device nodes
5153 Description: tty and some other devices nodes cannot be used on lustre
5154 Details : file's private_data field is used by device data and lustre
5155 values in there got lost. New field was added to struct file to
5156 store fs-specific private data.
5159 Frequency : when exporting Lustre via NFS
5161 Description: NFSD failed occasionally when looking up a path component
5162 Details : NFSD is looking up ".." which was broken in ext3 directories
5163 that had grown large enough to become hashed.
5166 Frequency : Clusters with multiple interfaces not on the same subnet
5168 Description: Nodes will repeatedly try to reconnect to an interface which it
5169 cannot reach and report an error to the log.
5170 Details : Extra peer list entries will be created by lconf with some peers
5171 unreachable. lconf now validates the peer before adding it.
5174 Frequency : Only if a default stripe is set on the filesystem root.
5176 Description: Setting a default stripe on the filesystem root prevented the
5177 filesystem from being remounted.
5178 Details : The client was sending extra request flags in the root getattr
5179 request and did not allocate a reply buffer for the dir EA.
5182 Frequency : occasional, higher if lots of files are accessed by one client
5183 Bugzilla : 6159, 6097
5184 Description: Client trips assertion regarding lsm mismatch/magic
5185 Details : While revalidating inodes the VFS looks up inodes with ifind()
5186 and in rare cases can find an inode that is being freed.
5187 The ll_test_inode() code will free the lsm during ifind()
5188 when it finds an existing inode and then the VFS later attaches
5189 this free lsm to a new inode.
5193 Bugzilla : 6422, 7030
5194 Description: MDS deadlock between mkdir and client eviction
5195 Details : Creating a new file via mkdir or mknod (starting a transaction
5196 and getting the ns lock) can deadlock with client eviction
5197 (gets ns lock and trying to finish a synchronous transaction).
5200 Frequency : occasional
5201 Description: While starting a server, the fsfilt_ext3 module could not be
5203 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5204 kernels. Previously, lconf would still use the ext3 name
5205 when trying to load modules. Now, it will correctly use
5206 ext3 on 2.4 and ldiskfs on 2.6.
5208 Severity : enhancement
5209 Description: The default stripe count has been changed to 1
5210 Details : The interpretation of the default stripe count (0, to lfs
5211 or lmc) has been changed to mean striping across a single
5212 OST, rather than all available. For general usage we have
5213 found a stripe count of 1 or 2 works best.
5215 Severity : enhancement
5216 Description: Add support for compiling against Cray portals.
5217 Details : Conditional compiling for some areas that are different
5221 Frequency : occasional
5222 Bugzilla : 6409, 6834
5223 Description: Creating files with an explicit stripe count may lead to
5224 a failed assertion on the MDS
5225 Details : If some OSTs are full or unavailable, creating files may
5226 trigger a failed assertion on the MDS. Now, Lustre will
5227 try to use other servers or return an error to the
5231 Frequency : occasional
5233 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5234 caused lock timeout during readahead (since 1.4.2).
5235 Details : Processes doing readahead might match a lock that hasn't been
5236 granted yet if there are overlapping and conflicting lock
5237 requests. The readahead process waits on ungranted lock
5238 (original lock is CBPENDING), while OST waits for that process
5239 to cancel CBPENDING read lock and eventually evicts client.
5241 Severity : enhancement
5243 Description: Initial enabling of flock support for clients
5244 Details : Implements fcntl advisory locking and file status functions.
5245 This feature is provided as an optional mount flag (default
5246 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5247 locking are implemented yet, and those that are are not guaranteed
5248 to be completely correct in production environments.
5249 mount -t lustre -o [flock|noflock] ...
5252 Frequency : occasional
5254 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5255 assertion in the ext3 JBD (journaling) layer.
5256 Details : The b_committed_data struct is protected by the big kernel lock
5257 in 2.4 kernels, serializing journal_commit_transaction() and
5258 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5259 access to this struct. In 2.6 kernels there is finer grained
5260 locking to improve SMP performance of the JBD layer.
5264 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5266 -----------------------------------------------------------------------------
5268 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5270 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5271 versions of Lustre. Please update all servers and clients to
5272 version 1.4.2 or later at the same time. You must also run
5273 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5274 to update the configuration logs.
5276 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5277 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5278 - swab llog records of type '0' so we get proper header size/idx (5861)
5279 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5280 - fix rename of one directory over another leaking an inode (5953)
5281 - avoid SetPageDirty on 2.6 (5981)
5282 - don't re-add just-being-destroyed locks to the waiting list (5653)
5283 - when creating new directories, inherit the parent's custom
5284 striping settings if present parent (3048)
5285 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5286 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5287 - don't hold client locks on temporary worklist from l_lru (5666)
5288 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5289 - automatically evict dead clients (5921)
5290 - Update file size properly in create+truncate+fstat case (6196)
5291 - Do not unhash mountpoint dentries, do not allow removal of
5293 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5294 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5295 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5296 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5297 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5298 - Correctly handle reads of files with no objects (6243)
5299 - lctl recover will also mark a device active if deactivate used (5933)
5301 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5302 - allow --write-conf on an MDS with different nettype than client (5619)
5303 - don't write config llogs to MDS for mounts not from that MDS (5617)
5304 - lconf should create multiple TCP connections from a client (5201)
5305 - init scripts are now turned off by default; run chkconfig --on
5306 lustre and chkconfig --on lustrefs to use them
5307 - upcalls are no longer needed for clients to recover to failover
5309 - add --abort-recovery option to lconf to abort recovery on device
5311 - add support for an arbitrary number of OSTs (3026)
5312 - Quota support protocol changes.
5313 - forward compatibility changes to wire structs (6007)
5314 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5315 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5316 - improved "lustre status" script
5317 - initialize blocksize for non-regular files (6062)
5318 - added --disable-server and --disable-client configure options (5782)
5319 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5320 - Vanilla 2.4.29 support
5321 - increase maximum number of obd devices to 520 (6242)
5322 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5323 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5324 SLES 9 2.6.5 kernels
5325 - limit stripes per file to 160 (the maximum EA size) (6093)
5327 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5330 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5331 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5332 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5333 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5334 - small ext3 extents cleanups and fixes (5733)
5335 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5336 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5337 - avoid needless client->OST connect, fix handle mismatch (5317)
5338 - fix DLM error path that led to out-of-sync client, long delays (5779)
5339 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5340 - fix several locking issues related to i_size (5492,5624,5654,5672)
5341 - don't move pending lock onto export if it is already evicted (5683)
5342 - fix kernel oops when creating .foo in unlinked directory (5548)
5343 - fix deadlock in obdfilter statistics vs. object create (5811)
5344 - use time_{before,after} to avoid timer jiffies wrap (5882)
5345 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5346 - Do not leak request if server was not able to process it (5154)
5347 - If mds_open unable to find parent dir, make that negative lookup(5154)
5348 - don't create new directories with extent-mapping (5909, 5936)
5350 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5351 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5352 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5353 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5354 - support for RHEL4 kernel on i686 (5773)
5355 - provide error messages when incompatible logs are encountered (5898)
5357 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5358 * version 1.4.0.10 (1.4.1 release candidate 1)
5360 - don't keep a lock reference when lock is not granted (4238)
5361 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5362 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5363 - drop import inflight refcount on signal_completed_replay error (5255)
5364 - unlock page after async write error during send (3677)
5365 - handle missing objects in filter_preprw_read properly (5265)
5366 - no transno return for symlink open, don't save no-trasno open (3440)
5367 - don't try to complete elan receive that already failed (4012)
5368 - free RPC server reply state on error (5406)
5369 - clean up thread from ptlrpc_start_thread() on error (5160)
5370 - readahead could read extra page into cache that wasn't ejected (5388)
5371 - prevent races in class_attach/setup/cleanup/detach (5260)
5372 - don't dereference de->d_inode after l_dput of de (5458)
5373 - use "int" for stripe value returned from lock_to_stripe (5544)
5374 - mballoc allocation and error-checking fixes in 2.6 (5504)
5375 - block device patches to fix I/O request sizes in 2.6 (5482)
5376 - look up hostnames for IB nals (5602)
5377 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5378 - don't start multiple acceptors for the same port (5277)
5379 - fix incorrect LASSERT in mds_getattr_name (5635)
5380 - export a proc file for general "ping" checking (5628)
5381 - fix "lfs check" to not block when the MDS is down (5628)
5383 - service request history (4965)
5384 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5385 - create an "evict_client" /proc entry on OSTs, like the MDS has
5386 - fix mount usage message, return errors per mount(8) (5168)
5387 - change grep [] to grep "[]" in tests so they work in more UMLs
5388 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5389 - remove ancient LOV_MAGIC_V0 EA support (5047)
5390 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5391 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5392 - allow readahead allocations to fail when low on memory (5383)
5393 - mmap locking landed again, after considerable improvement (2828)
5394 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5396 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5399 - send OST transaction number in read/write reply to free req (4966)
5400 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5401 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5402 - the watchdog thread now runs as interruptible (5246)
5403 - flock/lockf fixes (but it's still disabled, pending 5135)
5404 - don't use EXT3 constants in llite code (5094)
5405 - memory shortage at startup could cause assertion (5176)
5407 - reorganization of lov code
5408 - single portals codebase
5410 - add extents/mballoc support (5025)
5411 - direct I/O reads in the obdfilter (4048)
5412 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5414 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5417 - send OST transaction number in read/write reply to free req (4966)
5418 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5419 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5420 - the watchdog thread now runs as interruptible (5246)
5421 - handle missing objects in filter_preprw_read properly (5265)
5422 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5423 - drop import inflight refcount on signal_completed_replay error (5255)
5424 - unlock page after async write error during send (3677)
5425 - return original error code on reconstructed replies (3761)
5426 - no transno return for symlink open, don't save no-trasno open (3440)
5428 - add pid to ldlm debugging output (4922)
5429 - bump the watchdog timeouts -- we can't handle 30sec yet
5430 - extra debugging for orphan dentry/inode bug (5259)
5432 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5435 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5436 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5437 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5438 - fix NULL dereference in /proc/sys/portals/routes (4827)
5439 - allow failed mdc_close() operations to be interrupted (4561)
5440 - stop precreate on OST before MDS would time out on it (4778)
5441 - don't send partial-page writes before EOF from client (4410)
5442 - discard client grant for sub-page writes on large-page clients (4520)
5443 - don't free dentries not owned by NFS code, check generation (4806)
5444 - fix lsm leak if mds_create_objects() fails (4801)
5445 - limit debug_daemon file size, always print CERROR messages (4789)
5446 - use transno after validating reply (3892)
5447 - process timed out requests if import state changes (3754)
5448 - update mtime on OST during writes, return in glimpse (4829)
5449 - add mkfsoptions to LDAP (4679)
5450 - use ->max_readahead method instead of zapping global ra (5039)
5451 - don't interrupt __l_wait_event() during strace
5453 - add software watchdogs to catch hung threads quickly (4941)
5454 - make lustrefs init script start after nfs is mounted
5455 - fix CWARN/ERROR duplication (4930)
5456 - return async write errors to application if possible (2248)
5457 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5458 - print NAL number in %x format (4645)
5459 - update barely-supported suse-2.4.21-171 series (4842)
5460 - support for sles 9 %post scripts
5461 - support for building 2.6 kernel-source packages
5462 - support for sles km_* packages
5464 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5467 - ignore -ENOENT errors in osc_destroy (3639)
5468 - notify osc create thread that OSC is being cleaned up (4600)
5469 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5470 - reconstruct ost_handle() like mds_handle() (4657)
5471 - create a new thread to do import eviction to avoid deadlock (3969)
5472 - let lconf resolve symlinked-to devices (4629)
5473 - don't unlink "objects" from directory with default EA (4554)
5474 - hold socknal file ref over connect in case target is down (4394)
5475 - allow more than 32000 subdirectories in a single directory (3244)
5476 - fix blocks count for O_DIRECT writes (3751)
5477 - OST returns ENOSPC from object create when no space left (4539)
5478 - don't send truncate RPC if file size isn't changing (4410)
5479 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5480 - bind to privileged port in socknal and tcpnal (3689)
5482 - rate limit CERROR/CWARN console message to avoid overload (4519)
5483 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5484 - basic mmap support (3918)
5485 - kernel patch series update from b1_4 (4711)
5487 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5490 - avoid crash during MDS cleanup with OST shut down (2775)
5491 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5492 - don't use bad inodes on the MDS (3744)
5493 - dynamic object preallocation to improve recovery speed (4236)
5494 - don't hold spinlock over lock dumping or change debug flags (4401)
5495 - don't zero obd_dev when it is force cleaned (3651)
5496 - print grants to console if they go negative (4431)
5497 - "lctl deactivate" will stop automatic recovery attempts (3406)
5498 - look for existing locks in ldlm_handle_enqueue() (3764)
5499 - don't resolve lock handle twice in recovery avoiding race (4401)
5500 - revalidate should check working dir is a directory (4134)
5502 - don't always mark "slow" obdfilter messages as errors (4418)
5504 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5507 - don't close LustreDB during write_conf until it is done (3860)
5508 - fix typo in lconf for_each_profile (3821)
5509 - allow dumping logs from multiple threads at one time (3820)
5510 - don't allow multiple threads in OSC recovery (3812)
5511 - fix debug_size parameters (3864)
5512 - fix mds_postrecov to initialize import for llog ctxt (3121)
5513 - replace config semaphore with spinlock (3306)
5514 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5515 - don't allow enqueue to complete on a destroyed export (3822)
5516 - down write_lock before checking llog header bitmap (3825)
5517 - recover from lock replay timeout (3764)
5518 - up llog sem before sending rpc (3652)
5519 - reduce ns lock hold times when setting kms (3267)
5520 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5521 - fix NULL deref and obd_dev leak on setup error (3312)
5522 - replace some LBUG about llog ops with error handling (3841)
5523 - don't match INVALID dentries from d_lookup and spin (3784)
5524 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5525 - fix invalid assertion in ptlrpc_set_wait (3880)
5527 - add libwrap support for the TCP acceptor (3996)
5528 - add /proc/sys/portals/routes for non-root route listing (3994)
5529 - allow setting MDS UUID in .xml (2580)
5530 - print the stack of a process that LBUGs (4228)
5532 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5535 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5536 - make sure to unset replay flag from failed open requests (3440)
5537 - if default stripe count is 0, use OST count for inode size (3636)
5538 - update parent mtime/ctime on client for create/unlink (2611)
5539 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5540 - free recovery state on server during a forced cleanup (3571)
5541 - unregister_reply for resent reqs (3063)
5542 - loop back devices mounting and status check on 2.6 (3563)
5543 - fix resource-creation race that can provoke i_size == 0 (3513)
5544 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5545 - more debugging for page-accounting assertion (3746)
5546 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5547 - avoid deadlock after precreation failure (3758)
5548 - fix race and lock order deadlock in orphan handling (3450, 3750)
5549 - add validity checks when grabbing inodes from l_ast_data (3599)
5551 - add /proc/.../recovery_status to obdfilter (3428)
5552 - lightweight CDEBUG infrastructure, debug daemon (3668)
5553 - change default OSC RPC parameters to be better on small clusters
5554 - turn off OST read cache for files smaller than 32MB
5555 - install man pages and include them in rpms (3100)
5556 - add new init script for (un)mounting lustre filesystems (2593)
5557 - run chkconfig in %post for init scripts (3701)
5558 - drop scimac NAL (unmaintained)
5560 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5563 - clean kiobufs before and after use (3485)
5564 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5565 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5566 - change init script to fail more clearly if not run as root (1528)
5567 - allow clients to reconnect during replay (1742)
5568 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5569 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5570 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5571 - fix cond_resched (3554)
5572 - only evict unfinished clients after recovery (3515)
5573 - allow bulk resend, prevent data loss (3570)
5574 - dynamic ptlrpc request buffer allocation (2102)
5575 - don't allow unlinking open directory if it isn't empty (2904)
5576 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5577 - remove extraneous obd dereference causing LASSERT failure (3334)
5578 - don't use get_cycles() when creating temp. files on the mds (3156)
5579 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5580 - handle EEXIST for set-stripe, set proper directory name (3336)
5582 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5583 - fix ksocknal_fmb_callback() error messages (2918)
5585 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5588 - don't copy lvb into (possibly NULL) reply on error (2983)
5589 - don't deref dentry after dput, don't free lvb on error (2922)
5590 - use the kms to determine writeback rpc length (2947)
5591 - increment oti_logcookies when osc is inactive (2948)
5592 - update client's i_blocks count via lvb messages (2543)
5593 - handle intent open/close of special files properly (1557)
5594 - mount MDS with errors=remount-ro, like obdfilter (2009)
5595 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5596 - don't use cancelling-locks' kms values (2947)
5597 - use highest lock extent for kms, not last one (2925)
5598 - don't dereference ERR_PTR() dentry in error handling path (3107)
5599 - fix thread race in portals_debug_dumplog() (3122)
5600 - create lprocfs device entries at setup instead of at attach (1519)
5601 - common AST error handler, don't evict client on completion race (3145)
5602 - zero nameidata in detach_mnt in 2.6 (3118)
5603 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5604 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5605 - initialize RPC timeout timer earlier for 2.6 (3219)
5606 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5607 - print nal/nid for unknown nid (3258)
5608 - additional checks for oscc recovery before doing precreate (3284)
5609 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5610 - don't crash in mdc_close for bad permissions on open (3285)
5611 - zero i_rdev for non-device files (3147)
5612 - clear page->private before handing to FS, better assertion (3119)
5613 - tune the read pipeline (3236)
5614 - fix incorrect decref of invalidated dentry (2350)
5615 - provide read-ahead stats and refine rpc in flight stats (3328)
5616 - don't hold journal transaction open across create RPC (3313)
5617 - update atime on MDS at close time (3265)
5618 - close LDAP connection when recovering to avoid server load (3315)
5619 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5620 - don't leak open file on MDS after open resend (3325)
5621 - serialize filter_precreate and filter_destroy_precreated (3329)
5622 - loop device shouldn't call sync_dev() for nul device (3092)
5623 - clear page cache after eviction (2766)
5624 - resynchronize MDS->OST in background (2824)
5625 - refuse to mount the same filesystem twice on same mountpoint (3394)
5626 - allow llmount to create routes for mounting behind routers (3320)
5627 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5628 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5629 - fix rare problem with rename on htree directories (3417)
5631 - allow default OST striping configuration per directory (1414)
5632 - fix compilation for qswnal for 2.6 kernels (3125)
5633 - increase maximum number of MDS request buffers for large systems
5634 - change liblustreapi to be useful for external progs like lfsck (3098)
5635 - increase local configuration timeout for slow disks (3353)
5636 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5638 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5641 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5642 - don't overwrite extent policy data in reply if lock was blocked (2901)
5643 - drop filter export grants atomically with removal from device (2663)
5644 - del obd_self_export from work_list in class_disconnect_exports (2908)
5645 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5646 - swab reply message in mdc_close, other PPC fixes (2464)
5647 - fix destroying of named logs (2325)
5648 - overwrite old logs when running lconf --write_conf (2264)
5649 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5650 - fix race in target_handle_connect (2898)
5651 - mds_reint_create() should take same inode create lock (2926)
5652 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5653 - don't close files for self_export to avoid uninitialized obd (2936)
5654 - allow MDS with the same name as client node (2939)
5655 - hold dentry reference for closed log files for unlink (2325)
5656 - reserve space for all logs during transactions (2059)
5657 - don't evict page beyond end of stripe extent (2925)
5658 - don't oops on a deleted current working directory (2399)
5659 - handle hard links to targets without a parent properly (2517)
5660 - don't dereference NULL lock when racing during eviction (2867)
5661 - don't grow lock extents when lots of conflicting locks (2919)
5663 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5666 - account for cache space usage on clients to avoid data loss (974)
5667 - lfsck support in lustre kernel code (2349)
5668 - reduce journal credits needed for BRW writes (2370)
5669 - orphan handling to avoid losing space on client/server crashes
5670 - ptlrpcd can be blocked, stopping ALL progress (2477)
5671 - use lock value blocks to assist in proper KMS, faster stat (1021)
5672 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5673 - recovery for initial connections (2355)
5674 - fixes for mds_cleanup_orphans (1934)
5675 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5676 - block all file creations until orphan recovery completes (1901)
5677 - client remove rq_connection from request struct (2423)
5678 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5679 - recovery timer race (2670)
5680 - mdc_close recovey bug (2532)
5681 - ptlrpc cleanup bug (2710)
5682 - mds timeout on local locks (2588)
5683 - namespace lock held during RPCs (2431)
5684 - handle interrupted sync write properly (2503)
5685 - don't try to handle a message that hasn't been replied to (2699)
5686 - client assert failure during cleanup after abort recovery (2701)
5687 - leak mdc device after failed mount (2712)
5688 - ptlrpc_check_set allows timedout requests to complete (2714)
5689 - wait for inflight reqs when ptlrpcd finishes (2710)
5690 - make sure unregistered services are removed from the srv_list
5691 - reset bulk XID's when resending them (caught by 1138 test)
5692 - unregister_bulk after timeout
5693 - fix lconf error (2694)
5694 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5695 - readahead locks pages, leaves pending causing memory pressure (2673)
5696 - increase OST request buffers to 4096 on large machines (2729)
5697 - fix up permission of existing directories in simple_mkdir (2661)
5698 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5699 - don't assign transno to errored transactions (2742)
5700 - don't delete objects on OST if given a bogus objid from MDS (2751)
5701 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5702 - if rq_no_resend, then timeout request after recovery (2432)
5703 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5704 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5705 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5706 - MDS recovery shouldn't delete objects that it has given out (2730)
5707 - if enqueue arrives after completion, don't clobber LVB (2819)
5708 - don't unlock pages twice when trigger_group_io returns error (2814)
5709 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5710 - don't write pages to disk if there was an error (1450)
5711 - don't ping imports that have recovery disabled (2676)
5712 - take buffered bytes into account when balancing socknal conn (2817)
5713 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5714 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5715 - remove little-endian swabbing of llog records (1987)
5716 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5717 - retry reposting request buffers if they fail (1191)
5718 - grow extent at grant time to avoid granting a revoked lock (2809)
5719 - lock revoke doesn't evict page if covered by a second lock (2765)
5720 - disable VM readahead to avoid reading outside lock extents (2805)
5722 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5723 - updated kernel patches for hp-2.4.20 kernel (2681)
5725 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5728 - fix truncated write corruption (2366)
5729 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5731 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5732 - ensure len(uuid) < 37 in lmc (1171)
5733 - fix ia64 OOPS in llog_test (2255)
5734 - zero end of page at obdfilter for partial page writes (2648)
5735 - don't leave stale dentries around after renames (bug 2428)
5736 - fix timeouts when evicting a client with a single lock held (2642)
5737 - set deadline for the initial HELLO message to drain (2634)
5738 - print out dotted-quad IP addresses in the socknal (2302)
5740 - additional debugging for MDS client eviction problem (2443)
5741 - fix mkfsoptions support for osts (2603, 2604)
5743 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5746 - add series for the vanilla 2.6.0 kernel
5747 - add series for the vanilla 2.4.24 kernel
5748 - add series for a cray x86/64 UL kernel drop
5749 - fix xattr patches for the vanilla 2.4.19 series
5751 - generate true UUIDs in lmc (1171)
5752 - have portals stack dumping break in UML (2466)
5753 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5754 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5755 - wake callback waiting threads on client eviction (2460)
5756 - Add --ptldebug and --subsystem to lmc (1719)
5757 - update assertion to allow safe interrupt allocation
5758 - set rq_no_resend for cancel requests (2432)
5759 - recalculate ptlrpcd timeout after resend (2494)
5760 - call vfs_rmdir when removing pending directories (2368)
5761 - fix renaming a file to itself (2429)
5762 - lmc creates a default one-stripe lov (2454)
5763 - expand procfs space to handle large clusters (2326)
5764 - increase UML stack to avoid overflow
5765 - update lconf's list of debug and subsystem masks
5766 - fix lfs find --obd (2510)
5767 - /proc tunable for disabling filter read caching (2591)
5768 - stop rpm packages from altering slapd.conf (2301)
5769 - disable nagle in the socknal under 0conf (2578)
5770 - choose mds inode size based on stripe count (2572)
5771 - fix kernel-source rpm problems (2516)
5773 - add --disable-doc to avoid pdf generation (2421)
5774 - update documentation, tests, type-os, comments
5775 - avoid format warnings on ia64
5776 - remove the TOE NAL
5777 - tiny code cleanups by removing unused fields
5779 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5782 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5783 - remove the most blatant lies from BUILDING file (2371)
5784 - change default debug level to reasonable production setting
5785 - reduce client side cache size to reduce cache flush time
5786 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5787 - make TCP zerocopy and pinger support enabled by default (2476)
5788 - sync writes completed after process exits caused crashes (2319)
5789 - maintain correct mount count on the MDS (2356)
5790 - backout 1557, because 2316 wasn't really fixed
5791 - better file I/O statistics gathering in /proc
5792 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5793 - another kernel patch to fix zero-copy TCP function export
5794 - don't take duplicate lock when processing re-sent getattr (2420)
5795 - lctl uses obd_self_export instead of creating new conn (2353)
5796 - MDS/OST recovery case which requires object creation asserted (2425)
5797 - move lfs from /usr/sbin to /usr/bin in packages
5798 - fix race between mds_client_add and mds_client_free (2417)
5799 - use kmalloc instead of slabs in portals (2430)
5800 - don't create duplicate records when a failover MDS is present (2442)
5801 - remove unnecessary mount age check (2332)
5802 - don't remove directory inodes from locks prematurely (2451)
5803 - don't break if MDS service name is the same as hostname (2103)
5804 - fix races in client write RPC generation when cache full (2482)
5806 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5809 - remove now-unused request->rq_obd (278)
5810 - if an allocation fails, print out how much memory we've used (1933)
5811 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5812 - add the "configurable stack size" patch to most series files (1256)
5813 - ability to write large log records, for 100+ OST configs (2306)
5814 - fix NULL deref when filter_prep fails (2314)
5815 - fix operator precedence error in filter_sync
5816 - dynamic allocation of socknal TX descriptors (2315)
5817 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5818 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5819 - ensure that CWARN messages in Portals always get to the syslog
5820 - __init/__exit are not for prototype decls (ldlm_init/exit)
5821 - x86-64 compile warning fixes
5822 - fix gateway LMC keyword conflict (2318)
5823 - fix MDS lock inversions in getattr/reint paths (1844)
5824 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5825 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5826 - DLM race condition prevented some lock evictions (2328)
5827 - ENOMEM detection and retry on socknal sends (2230)
5828 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5829 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5830 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5831 - better error messages when a client is rejected during recovery (1505)
5832 - avoid cancelling locks which were never granted, after failure (2330)
5833 - fix i_sem/journal inversion in mds_client_add (2333)
5834 - fix truncate/getattr lock cycle deadlock (2334)
5835 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5836 - fix two rare exit paths which could leak an l_lock() ref (2321)
5837 - fencepost error in MDS/OST orphan recovery (2226)
5838 - make log record alignment 8 bytes (1988)
5839 - lstripe now fails when requested offset > ost_count (2237)
5840 - ensure that all kernel series have a complete list.h (1607)
5841 - fix crashes in special-file operations (2316)
5842 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5844 - allow configurable automake binary, for testing new versions
5845 - small update to the lfs documentation
5847 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5849 * fix negative export reference count in fsfilt_sync (2312)
5851 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5852 * release candidate 0.9.1
5854 - orphans are moved into the PENDING directory for possible recovery
5855 - replayed opens now open by fid for orphan/rename safety (1042)
5856 - last close of an orphan inode generates a transno (683)
5857 - chdir() and mount() now pin the directory entry (1020)
5858 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5859 - discard very old requests without processing them (1502)
5860 - remove some common, well-understood CERRORs (1505)
5861 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5862 - clear "grant" flags in OST replies until OST grant code lands (1644)
5863 - fix read performance by not clobbering i_blksize on client (1598)
5864 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5865 - make LRU size a /proc tunable, clears locks when reduced (707)
5866 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5867 - server locks take references on exports now (1558)
5868 - build fixes for 2.4.20-rh trees (1663)
5869 - return an error from lov_create if all OSCs are inactive (1751)
5870 - fix import levels when a reconnect happens without a timeout (1597)
5871 - exit early from mds_open if we get a lookup error (1749)
5872 - partial page read at EOF wouldn't wait for disk before sending (1642)
5873 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5874 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5875 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5876 - write/truncate lock inversion (1639)
5877 - Don't auto-load obdclass, portals modules during cleanup (1495)
5878 - fix timestamps from jumping to "now" (1763)
5879 - extra journal assertions (1648)
5880 - add an extra multiunlink test (1771)
5881 - fix read_record/write_record API (1776)
5882 - fix leak of offset_extent, possible incorrect i_size later (1772)
5883 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5884 - Debugging for the kqswnal_get_idle_tx problems (1820)
5885 - Allow recovery to be attempted multiple times (1536)
5886 - Write out MDS last_rcvd file after it is first created (1600)
5887 - Fix tx_descriptor leak in failed transmit situations (1827)
5888 - ext3 journaling fixes for assertion failure after IO error (1871)
5889 - class_export_put() on freed export after completion AST error (1896)
5890 - Fix revalidate looping in VFS (1322)
5891 - Don't access a freed export during MDS_REINT timeout (1521)
5892 - Add open-unlink recovery support on the MDS (1673,1764)
5893 - Return an error if no MDS data was read from last_rcvd (1946)
5894 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5895 - Don't setup a disk device that doesn't match exported UUID (317)
5896 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5897 - avoid committing NULL handle in force close
5898 - local.sh is now a one-stripe LOV configuration
5899 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5900 - fix ext3 htree duplicate directory entry corruption (1516)
5901 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5902 - update to new LOV EA format (2097)
5903 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5905 2003-06-15 Phil Schwan <phil@clusterfs.com>
5908 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5909 - per-import/export recovery handling (958, 931, 959)
5910 - multiple last-rcvd slots, for serving multiple FSes (949)
5911 - connections are again shared between multiple imp/exports (963, 964)
5912 - "umount -f" would hang if any requests needed to be sent (393, 978)
5913 - avoid pinning large req buffer by copying for queued messages (989)
5914 - add "uuid" to "lctl device" command to help upcalls (991)
5915 - "open" RPCs with transnos would confuse recovery counters (1037)
5916 - do proper endian conversion of all wire messages (288, 340, 891)
5917 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5918 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5919 - fix ptlrpc_connection leak in target_handle_connect (1174)
5920 - fix import refcounting bug in OST and MDS cleanup (1134)
5921 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5922 - fix very unlikely obd_types race condition (501)
5923 - remove osc_open hack for echo_client (1187)
5924 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5925 - a failure in read_inode2 leads to deadlock (1139)
5926 - cancel ack-locks as soon as transaction is committed (1072)
5927 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5928 - make sure to commitrw after any preprw to avoid deadlock (1162)
5929 - failing to execute a file in a lustre FS would lock inode (1203)
5930 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5931 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5932 - obd_uuid2tgt would walk off the end of the list (1255)
5933 - on IA64 the osc would give portals incorrect bulk size (1258)
5934 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5935 - fix lock inversion caused by new llite matching code (1282)
5936 - limit the number of dirty pages on a client to 10MB (1286)
5937 - timed out locks were not being corrected cancelled (1289)
5938 - fix O_DIRECT above 4GB on IA-32 (1292)
5939 * major user-visible changes
5940 - fail out/fail over policy now controlled by the upcall (993)
5942 - add OBD_PING to check server availability and failure (954)
5943 - lustre messages are now sent in sending host order (288, 340, 891)
5944 - add eadatalen to MDS getattr reply (340)
5945 - OST read replies may contain second buffer, with per-page status (593)
5947 2003-03-11 Phil Schwan <phil@clusterfs.com>
5950 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5951 - failed open()s could cause deadlock; fixed (867, 869)
5952 - stop cancelling OST locks when files are closed (481)
5953 - overlapping XID spaces caused network corruption (851, 853)
5954 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5955 - setattr_raw added, to do single-RPC, server-side setattrs
5956 - lmc/lconf syntax change for OST UUIDs
5957 - fix crashy race condition between ptlrpc_free_req and osc_close
5958 - don't use request in mdc_enqueue if we hit a timeout (889)
5959 - don't set the inode i_size for regular files from the MDS (896)
5960 - handle out of order completion AST (842)
5961 - don't LBUG if a lock request times out after receiving AST (913)
5962 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5963 - if a bad lock AST arrives, send an error instead of dropping entirely
5964 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5965 - fix leak in bulk IO when only partially completed (899, 900, 926)
5966 - fix O_DIRECT for ia64 (55)
5967 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5968 - C-z after timeout could hang a process forever; fixed (977)
5970 - client-side I/O cache (678, 924, 929, 941, 970)
5972 - READPAGE and SETATTRs which don't take server-side locks get
5975 2003-02-11 Phil Schwan <phil@clusterfs.com>
5978 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5979 - Fix fsfilt_extN_readpage() to read a full page of directory
5980 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5981 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5982 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5983 - Most intent operations were removed, in favour of a new RPC mode
5984 that does a single RPC to the server and bypasses most of the VFS
5985 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5986 - Aggressively cancel local locks on DLM servers
5987 - mds_reint_unlink sends EA to the client if it's the last nlink.
5988 client uses that EA to unlink OST objects.
5989 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5990 - recursive symlinks were fixed (439)
5991 - fixed NULL deref in DEBUG_REQ
5992 - filter_update_lastobjid no longer calls sync, which annoyed extN
5993 - fixed multi-client small-writes to a single file problem (445)
5994 - fixed mtime updates during file writes (607)
5995 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5996 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5997 - fixed Linux OST size reporting problem (444, 656)
5998 - OST now updates object mtime with writes or setattr (607, 619)
5999 - client verifies file size before zeroing page past EOF (445)
6000 - OST now writes last allocated objid to disk with allocation (108)
6001 - LOV on echo now works (409)
6003 - mds_reint_unlink sends a new buffer, with the EA included. this
6004 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6005 set if a regular file was being unlinked, and it was the last link
6006 - use PtlGet from the target for bulk writes (315)
6007 - OST now updates object mtime with writes or setattr (607, 619)
6008 - LDLM now has a grant-time callback to revalidate locked items, if
6010 - Many MDS operations were reorganized to combat race conditions
6012 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6013 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6015 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6018 - Fully reactivate OST imports after reconnection (512, others)
6019 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6020 - More graceful error handling for truncating on dead OST (515)
6021 - Don't error out unless we're actually accessing dead stripes (474)
6022 - Fix garbage sizes when stripes are missing (410)
6023 - LRU counters were broken, causing constant lock purge (433, 432)
6024 - garbage on read from stripes with failed OSTs (441)
6025 - mark OSCs as active before reconnecting during recovery (438)
6026 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6027 - lfind did not preserve OST order in output (443)
6028 - symlinks cause hung clients, incorrect data (439)
6029 - stop dereferencing request after dropping refcount (457)
6030 - don't LASSERT(spin_is_locked) on non-SMP (455)
6031 - fixes for many rename() bugs
6032 - fstat didn't correctly synchronize attributes (399)
6033 - server must handle lock cancellation during blocking AST prep (487)
6034 - bulk descriptors were free()d too soon (511)
6035 - fix paths in lconf, which would load incorrect modules (451, 507)
6036 - fix confusing lconf 'host not found' error message (386)
6037 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6038 - fix race condition in mdc_blocking_ast() for inode access (526)
6039 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6040 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6041 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6042 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6043 - fix lstripe in several ways (kernel side) (527)
6044 - fix request leak in ldlm_cli_enqueue (262)
6045 - incorrect OSC was marked inactive after OST failure
6046 - call mds_fs_cleanup before unmounting filesystem (524)
6047 - fix races between taking ns_lock and ldlm_lock_change_resource
6048 - fix races updating LOV export open file list
6049 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6050 - fix recovery NULL deref in ldlm_cli_cancel_unused
6051 - fix some DLM races by using new hash table for lock handles (419)
6052 - permit the client to specify desired inodes, at replay
6053 - duplicate requests when we queue them for replay reintegration
6054 - fix last_rcvd offset calculation
6055 - sync after each recovered transaction, so we always make progress
6056 - never, not always, ERESTART requests without transnos
6057 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6059 - skip replay if the MDS says that the client is already connected
6060 - don't check for a recovery-enabled export to match lctl's UUID
6061 - don't INC_USE_COUNT for phantom exports
6062 - don't crash when cleaning up phantom exports (567)
6063 - don't double-finish or set replay data for errored mdc_open requests
6064 - abort requests when they time out, so we don't get old replies
6065 - send/receive replies for AST messages again
6066 - if the client says that it doesn't have the lock, cancel it on the
6068 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6069 mark it as destroyed, it will all work out when decref is called
6070 - fix module use counts (22, 581)
6072 - ASTs now expect a reply (server cancels lock on error reply)
6074 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6077 - fix many simultaneous client startup (392)
6078 - fix dentry->d_it clobbering
6079 - credentials weren't being shipped for readdir/getattr operations
6080 - remove invalid assertions triggered during some concurrent MD
6082 - proper Lustre versions added (336, 389)
6083 - fix memory leak for create error case (398)
6084 - fix LOV locking bug that would get cli/srv out of sync
6085 - fix echo client over LOV (409)
6086 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6087 - fix double-O_EXCL intent crash (424)
6088 - avoid sending multiple lock CANCELs (352)
6090 - MDS can do multi-client recovery (modulo bugs in new code)
6092 - many updates, edits, cleanups
6094 2002-11-18 Phil Schwan <phil@clusterfs.com>
6097 - fix null d_it dereference (346)
6098 - fix full OST/dbench hang (333)
6099 - fix permission problem with file removal (286)
6100 - fix removal of OSCs from LOV when they fail
6101 - fix NULL deref during bulk timeout (214)
6102 - fix problems related to multiple filesystems on one MDS (241)
6103 - fixed serious subtle metadata locking bugs
6104 - free locks on clients when inodes are removed due to memory
6106 - fix inode pointer in lock data (285)
6107 - partial support for multiple MDS on a single host (241)
6108 - data locks weren't cancelled at clear_inode time (290, 311)
6109 - intent locks could lead to unbounded lock growth (205)
6110 - added a maximum lock count, an LRU list, and a flusher
6111 - fix multiple rename (365)
6112 - properly abstracted the echo client
6113 - OSC locked 1 byte too many; fixed
6114 - rewrote brw callback code:
6115 - fixed recovery bugs related to LOVs (306)
6116 - fixed too-many-pages-in-one-write crash (191)
6117 - fixed (again) crash in sync_io_timeout (214)
6118 - probably fixed callback-related race (385)
6120 - Add capability to MDS protocol
6121 - LDLM cancellations and callbacks on different portals
6123 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6126 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6127 - defer open object destruction to close time (601981, 138)
6128 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6129 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6130 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6131 - don't call obd_setattr in ll_file_release for destroyed objects
6132 * protocol change to lustre_msg: move |version| and add |flags|
6133 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6134 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6135 * added replay of create, unlink, link and rename operations during
6136 MDS failover; recovery should be much more robust now
6137 * remove failed OSCs from LOVs (only lov_create uses this so far)
6138 * the lustre-HOWTO was brought (more) up to date (582544)
6140 2002-10-23 Phil Schwan <phil@clusterfs.com>
6143 - in-use dentries weren't being reused properly (617851)
6144 - prevent multiple LDLM setup (599178)
6145 - fix LOV size calculations for truncate (617853)
6146 - fix client handling of MDS intent errors (POSIX)
6147 - fix permission bug in lovstripe.c test (624321)
6148 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6149 - truncate past end of file could corrupt data
6150 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6151 - a race in recovery could return ETIMEDOUT to apps (623947)
6152 - building outside the source directory was fixed
6153 * the lustre-HOWTO was brought (more) up to date (582544)
6154 * major progress was made on recovery functionality
6156 2002-10-10 Phil Schwan <phil@clusterfs.com>
6159 - recovery deadlock fix
6160 - rm -rf causes LBUG fix (617817)
6161 - file open by multiple tasks fix (618962)
6162 - directory permissions bugs (602707 and 620007)
6163 - journal_stop fixed with locking (611313)
6164 - O_APPEND failures resolved (618273, perhaps 614459)
6165 - lconf PATH fix (619770)
6166 - IA64 build fix (621450)
6167 - RPC buffer sizes scale with amount of memory
6169 2002-10-01 Phil Schwan <phil@clusterfs.com>
6172 - locks would be cancelled without throwing away data pages,
6173 resulting in inconsistent data (605627)
6174 - inode attributes were not always being refreshed (605627, 612449)
6175 - lconf now continues to cleanup after lctl reports an error
6176 - MDS now enforces user permissions (602707)
6177 - lprocfs cleanup fixed, but not yet enabled (614157)
6178 - fixed infinite server hang, should a client not respond to an AST
6179 - avoid going into recovery if user calls readlink() with a buffer
6180 that's too small (613941)
6181 - AST RPCs no longer require replies (614867) -- this may be changed
6182 - don't crash server if client sends an IOV that's too big (611336)
6183 - fixed lock conversion deadlock (611892)
6184 - fixed the following of symlinks (614622)
6185 * recovery: the server can remove locks from a client that dies, other
6186 clients can make progress
6187 * more extN patch fixes
6188 * compile-time configurable ptlrpc buffer allocations
6190 - collaborative read cache document
6191 - Lustre Lite Performance CDR document-in-progress
6193 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6196 - fix typo in patch-2.4.18
6198 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6201 - clear ptlrpc request each time in handle_incoming_request()
6202 - unlink of files now destroys the object on the OST
6204 2002-09-19 Peter Braam <braam@clusterfs.com>
6206 * add hard link support
6207 * change obdfile creation method
6208 * kernel patch changed
6210 2002-09-19 Peter Braam <braam@clusterfs.com>
6213 - stack overflow bug in extN fixed
6215 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6217 * documentation updates
6218 - add man pages for config tools
6219 - update tests/README to describe testing with new config tools
6220 - finish metadata API descriptions
6221 * bug fixes and cleanups
6222 - statfs workaround for 16TB limit
6223 - LOV stripe allocation improved, can stripe on subset of OSTs
6224 - LOV file size/IO offset was wrong for files > 4GB in size
6225 - object EA data was being dropped, caused files to be unreadable
6226 - memory overflow with non-LOV OST caused memory corruption
6227 - fixed regression tests to work with new config tools, obdfilter
6228 - fixed bug when directory size became larger than 1 block
6229 - fixed bug (for single client case) when PWD was deleted
6230 - invalidate local directory pages when doing intent-based ops
6231 - avoid LDLM oops when lock callback contained bad data
6233 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6235 * documentation updates
6236 * bug fixes and cleanups
6237 - configuration tools
6240 - 64-bit compile warnings
6241 - 64-bit internal statfs data
6243 * test_brw on persistent OST devices
6245 * lprocfs (disabled)
6247 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6249 * documentation updates
6250 * bug fixes and cleanups
6251 * configuration tools
6253 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6256 * many small fixes to 0_5_4
6257 * io/network handling
6258 * thinkos in MDS operations
6260 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6263 * crucial basic fixes to 0.5.3
6264 * IOR, Iozone work over Elan
6267 2002-08-07 Phil Schwan <phil@clusterfs.com>
6268 * version 0_5_3, our first alpha
6269 * we use the new Portals iovs
6270 * documentation updates
6271 * bug fixes and cleanups
6272 * small changes in the DLM wire protocol
6274 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6275 * version 0_5_1 with some initial stability,
6276 * locking on MD and file I/O.
6277 * documentation updates
6278 * several bug fixes since 0.5.0
6279 * small changes in wire protocol
6281 2002-07-18 Phil Schwan <phil@clusterfs.com>
6283 * delivered as Lustre Light Alpha
6284 * fixed a crash after handling invalid MDS requests
6285 * fixed directory pages for architectures with non-4k pages sizes
6287 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6288 * release version v0_4_4
6289 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6291 * Fixes a number of interruption problems with OST operations.
6292 * Update documentation for portals header changes
6293 * Move all wire protocol structs/defines to lustre_idl.h
6294 * Fixes symlink length bug.
6295 * Add tcpdump to repository.
6297 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6298 * release version v0_4_3
6299 * Fixes statfs for inodes on extN.
6300 * Fixes bug in runtests which would delete /etc/hosts.
6301 * Use 64-bit object IDs wherever possible (not into VFS though)
6302 Remove ost_get_info, which is unused by lustre, and out of date.
6304 2002-07-03 Peter Braam <braam@clusterfs.com>
6305 * release version v0_4_2 Fixes a lookup error (type not passed)
6306 * move forward to head of Portals
6307 * move forward to latest Lustre kernel
6309 2002-06-25 Peter Braam <braam@clusterfs.com>
6310 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob