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 : always with long access acl
37 Descriptoin: mds can't pack reply with long acl.
38 Details : mds don't control size of acl but they limited by reint/getattr
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.
74 -------------------------------------------------------------------------------
75 2008-12-31 Sun Microsystems, Inc.
77 * Support for kernels:
79 2.6.9-67.0.22.EL (RHEL 4),
80 2.6.16.60-0.31 (SLES 10),
81 2.6.18-92.1.17.el5 (RHEL 5),
82 2.6.22.14 vanilla (kernel.org)
83 * Client support for unpatched kernels:
84 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
85 2.6.16 - 2.6.22 vanilla (kernel.org)
86 * Client support for unpatched kernels:
87 we do not recommend using patchless RHEL4 clients with kernels
88 prior to 2.6.9-55EL (RHEL4U5).
89 * Recommended e2fsprogs version: 1.40.11-sun1
90 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
91 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
92 removed cwd "./" (refer to Bugzilla 14399).
93 * A new quota file format has been introduced in 1.6.5.
94 The format conversion from prior releases is handled transparently,
95 but releases older than 1.4.12/1.6.5 don't understand this new
96 format. The automatic format conversion can be avoided by running
97 the following command on the MDS:
98 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
99 For more information, please refer to bugzilla 13904.
100 * A new quota file format was introduced in 1.6.6/1.8.0.
101 The format conversion from prior releases is handled transparently,
102 but releases older than 1.6.6/1.8.0 don't understand this new
103 format. The automatic format conversion can be avoided by running
104 the following commands on the MDS and OSS servers (for
105 pre 1.4.12-1.6.5 quota files):
106 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV',
107 'tunefs.lustre --param="ost.quota_type=ug1" $MDTDEV'
108 or (for 1.4.12/1.6.5 quota files)
109 'tunefs.lustre --param="mdt.quota_type=ug2" $MDTDEV',
110 'tunefs.lustre --param="ost.quota_type=ug2" $MDTDEV'
111 For more information, please refer to bugzilla 13904.
112 * Output of lfs quota has been made less detailed by default,
113 old (verbose) output can be obtained by using -v option.
114 * File join has been disabled in this release, refer to Bugzilla 16929.
115 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
116 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
117 Lustre filesystem with 4K stack may cause a stack overflow. For more
118 information, please refer to bugzilla 17630.
121 Frequency : on remount
123 Description: external journal device not working after the remount
124 Details : clear dev_rdonly flag for external journal devices in
130 Description: shutdown vs evict race
131 Details : client_disconnect_export vs connect request race.
132 if client will evicted at this time - we start invalidate
133 thread without referece to import and import can be freed
139 Description: shrink LOV EAs before replying
140 Details : correctly adjust LOV EA buffer for reply.
145 Description: don't skip ost target if they assigned to file
146 Details : Drop slow OSCs if we can, but not for requested start idx.
147 This means "if OSC is slow and it is not the requested
148 start OST, then it can be skipped, otherwise skip it only
149 if it is inactive/recovering/out-of-space.
151 Severity : enhancement
153 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
156 Frequency : rare, need acl's on inode.
158 Description: client can't handle ost additional correctly
159 Details : if ost was added after client connected to mds client can have
160 hit lnet_try_match_md ... to big messages to wide striped files.
161 in this case need teach client to handle config events about add
162 lov target and update client max ea size at that event.
164 Severity : enhancement
166 Description: Update to sles9 kernel-2.6.5-7.314.
168 Severity : enhancement
170 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
173 Frequency : Create a symlink file with a very long name
175 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
176 Details : If there is no extra space in the request for early cancels,
177 ldlm_req_handles_avail() returns 0 instead of a negative value.
182 Description: mds is deadlocked
183 Details : in rare cases, inode in catalog can have i_no less than have parent
184 i_no, this produce wrong order for locking during open, and
185 parallel unlink can be lock open. this need teach mds_open to grab
186 locks in resource id order, not at parent -> child order.
188 Severity : enhancement
190 Description: Add /proc entry for import status
191 Details : The mdc, osc, and mgc import directories now have
192 an import directory that contains useful import data for debugging
195 Severity : enhancement
197 Description: Re-disable certain /proc logging
198 Details : Enable and disable client's offset_stats, extents_stats and
199 extents_stats_per_process stats logging on the fly.
202 Frequency : Only on FC kernels 2.6.22+
204 Description: oops in statahead
205 Details : Do not drop reference count for the dentry from VFS when lookup,
206 VFS will do that by itself.
208 Severity : enhancement
210 Description: Generic /proc file permissions
211 Details : Set /Proc file permissions in a more generic way to enable non-
212 root users operate on some /proc files.
216 Description: Hitting mdc_commit_close() ASSERTION
217 Details : Properly handle request reference release in
218 ll_release_openhandle().
222 Frequency : only patchless client
223 Description: add workaround for race between add/remove dentry from hash
225 Severity : enhancement
227 Description: Allow OST glimpses to return PW locks
231 Description: LBUG when llog conf file is full
232 Details : When llog bitmap is full, ENOSPC should be returned for plain
237 Description: Prevent import from entering FULL state when server in recovery
241 Description: service mount cannot take device name with ":"
242 Details : Only when device name contains ":/" will mount treat it as
248 Description: replace ptlrpcd with the statahead thread to interpret the async
249 statahead RPC callback
253 Frequency : on recovery
254 Description: I/O failures after umount during fail back
255 Details : if client reconnected to restarted server we need join to recovery
256 instead of find server handler is changed and process self eviction
257 with cancel all locks.
259 Severity : enhancement
261 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
265 Description: Kernel BUG tries to release flock
266 Details : Lustre does not destroy flock lock before last reference goes
267 away. So always drop flock locks when client is evicted and
268 perform unlock regardless of successfulness of speaking to MDS.
270 Severity : enhancement
272 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
274 Severity : enhancement
276 Description: Upcall on Lustre log has been dumped
277 Details : Allow for a user mode script to be called once a Lustre log has
278 been dumped. It passes the filename of the dumped log to the
279 script, the location of the script can be specified via
280 /proc/sys/lnet/debug_log_upcall.
285 Description: avoid messages about idr_remove called for id that is not allocated
286 Details : Move assigment s_dev for clustered nfs to end of initialization,
287 for avoid problem with error handling.
292 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
293 Details : When connection is reused this not moved from CONN_UNUSED_HASH into
294 CONN_USED_HASH and this prodice warning when put connection again
300 Description: avoid ASSERTION(client_stat->nid_exp_ref_count == 0) failed
301 Details : release reference to stats when client disconnected, not
302 when export destroyed for avoid races when client destroyed
303 after main ost export.
307 Description: more cleanup in mds_lov
308 Details : add workaround for get valid ost count for avoid warnings about
309 drop too big messages, not init llog cat under semphore which
310 can be blocked on reconnect and break normal replay, fix access
313 Severity : enhancement
315 Description: File striping can now be set to use an arbitrary pool of OSTs.
317 Severity : enhancement
319 Description: Export bytes_read/bytes_write count on OSC/OST.
323 Description: Early reply size mismatch, MGC loses connection
324 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
325 the connect flags are properly negotiated.
329 Description: Properly propagate oinfo flags from lov to osc for statfs
330 Details : restore missing copy oi_flags to lov requests.
334 Description: exports in /proc are broken
335 Details : recreate /proc entries for clients when they reconnect.
337 Severity : enhancement
339 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
340 : l_getgroups(8), lst(8), routerstat(8)
341 Details : included man pages for llobdstat(8), llstat(8),
342 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
344 Severity : enhancement
346 Description: Implement lustre ll_show_options method.
348 Severity : enhancement
350 Description: Update to SLES9 kernel-2.6.5-7.312.
352 Severity : enhancement
354 Description: Update to RHEL4 kernel-2.6.9-67.0.22.EL.
358 Description: exports in /proc are broken
359 Details : recreate /proc entries for clients when they reconnect.
363 Description: don't fail open with -ERANGE
364 Details : if client connected until mds will be know about real ost count
365 get LOV EA can be fail because mds not allocate enougth buffer
370 Description: Resolve device initialization race
371 Details : Prevent proc handler from accessing devices added to the
372 obd_devs array but yet be intialized.
374 Severity : enhancement
376 Description: configure's --enable-quota should check the
377 : kernel .config for CONFIG_QUOTA
378 Details : configure is terminated if --enable-quota is passed but
379 : no quota support is in kernel
381 Severity : enhancement
383 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
385 Severity : enhancement
387 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
391 Frequency : rare, on PPC clients
392 Description: don't swab ost objects in response about directory, because
394 Details : bug similar bug 14856, but in different function.
396 Severity : enhancement
398 Description: lfs quota tool enhancement
399 Details : added units specifiers support for setquota, default to
400 current uid/gid for quota report, short quota stats by
401 default, nonpositional parameters for setquota, added
402 llapi_quotactl manual page.
404 Severity : enhancement
406 Description: *optional* service tags registration
407 Details : if the "service tags" package is installed on a Lustre node
408 When the filesystem is mounted, a local-node service tag will
409 be created. See http://inventory.sun.com/ for more information
410 about the Service Tags asset management system.
414 Description: Client runs out of low memory
415 Details : Consider only lowmem when counting initial number of llap pages
418 Description: Kernel BUG tries to release flock
419 Details : Lustre does not destroy flock lock before last reference goes
420 away. So always drop flock locks when client is evicted and
421 perform unlock regardless of successfulness of speaking to MDS.
424 Frequency : occasional
426 Description: add refcount for osc callbacks, so avoid panic on shutdown
428 Severity : enhancement
430 Description: Update to RHEL4 kernel-2.6.9-67.0.20.
433 Frequency : testing only
435 Description: sanity test 65a fails if stripecount of -1 is set
436 Details : handle -1 striping on filesystem in ll_dirstripe_verify
439 Frequency : only in unusual configurations
441 Description: Kernel panic with find ost index.
442 Details : lov_obd have panic if some OST's have sparse indexes.
444 Severity : enhancement
446 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
449 Frequency : rarely, if filesystem is mounted with -o flock
451 Description: do not process already freed flock
452 Details : flock can possibly be freed by another thread before it reaches
453 to ldlm_flock_completion_ast.
456 Frequency : rarely, if filesystem is mounted with -o flock
458 Description: LBUG during stress test
459 Details : Need properly lock accesses the flock deadlock detection list.
462 Frequency : rarely, if binaries are being run from Lustre
464 Description: oops in page fault handler
465 Details : kernel page fault handler can return two special 'pages' in
466 error case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
469 Frequency : rarely, during shutdown
471 Description: timeout with invalidate import.
472 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
473 should be handled by ptlrpcd. This produce long age waiting and
474 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
480 Description: ASSERTION(CheckWriteback(page,cmd)) failed
481 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
485 Frequency : only with broken builds/installations
487 Description: no LBUG if lquota.ko and fsfilt_ldiskfs.ko are different versions
488 Details : just return an error to a user, put a console error message
490 Severity : enhancement
492 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
494 Severity : enhancement
496 Description: Update to RHEL4 kernel-2.6.9-67.0.15.
498 Severity : enhancement
500 Description: enable MGS and MDT services start separately
501 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
502 a co-located MGS without starting the MGS, which is a complement
503 to 'nosvc' mount option.
506 Frequency : always, on big-endian systems
508 Description: cleanup in ptlrpc code, related to PPC platform
509 Details : store magic in native order avoid panic's in recovery on PPC
510 node and forbid from this error in future. Also fix posibily
511 of twice swab data. Fix get lov striping to userpace.
514 Frequency : rarely, if replay get lost on server
516 Description: server incorrectly drop resent replays lead to recovery failure.
517 Details : do not drop replay according to msg flags, instead we check the
518 per-export recovery request queue for duplication of transno.
521 Frequency : after recovery
523 Description: precreate to many object's after del orphan.
524 Details : del orphan st in oscc last_id == next_id and this triger growing
525 count of precreated objects. Set flag LOW to skip increase count
526 of precreated objects.
529 Frequency : rare, on clear nid stats
531 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
532 Details : when clean nid stats sometimes try destroy live entry,
533 and this produce panic in free.
536 Frequency : occasionally since 1.6.4
538 Description: Stack overflow during MDS log replay
539 Details : ease stack pressure by using a thread dealing llog_process.
542 Frequency : very rare
544 Description: MDT cannot be unmounted, reporting "Mount still busy"
545 Details : Mountpoint references were being leaked during open reply
546 reconstruction after an MDS restart. Drop mountpoint reference
547 in reconstruct_open() and free dentry reference also.
552 Description: wait until IO finished before start new when do lock cancel.
553 Details : VM protocol want old IO finished before start new, in this case
554 need wait until PG_writeback is cleared until check dirty flag
555 and call writepages in lock cancel callback.
560 Description: mds_mfd_close() ASSERTION(rc == 0)
561 Details : In mds_mfd_close(), we need protect inode's writecount change
562 within its orphan write semaphore to prevent possible races.
565 Frequency : rare, on shutdown ost
567 Description: don't hit live lock with umount ost.
568 Details : shrink_dcache_parent can be in long loop with destroy dentries,
569 use shrink_dcache_sb instead.
572 Frequency : only when echo_client is used
574 Description: don't panic with use echo_client
575 Details : echo client pass NULL as client nid pointer and this produce NULL
579 Frequency : Always on 32-bit PowerPC systems
581 Description: fix build on PPC32
582 Details : compile code with -m64 flag produce wrong object file for PPC32.
587 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
588 Details : In reconstruct_* functions, LASSERTs on both the data supplied
589 by a client, and the data on disk are dangerous and incorrect.
590 Change them with client eviction.
592 Severity : enhancement
594 Description: skiplist implementation simplification
595 Details : skiplists are used to group compatible locks on granted list
596 that was implemented as tracking first and last lock of each lock
597 group the patch changes that to using doubly linked lists
601 Description: delete compatibility for 32bit qdata
602 Details : as planned, when lustre is beyond b1_8, lquota won't support 32bit
603 qunit. That means servers of b1_4 and servers of b1_8 can't be
604 used together if users want to use quota.
607 Frequency : only with administrator action
609 Description: mount failure if config log has invalid conf_param setting
610 Details : If administrator specified an incorrect configuration parameter
611 with "lctl conf_param" this would cause an error during future
612 client mounts. Instead, ignore the bad configuration parameter.
615 Frequency : blocks per group < blocksize*8 and uninit_groups is enabled
617 Description: ldiskfs error: XXX blocks in bitmap, YYY in gd
618 Details : If blocks per group is less than blocksize*8, set rest of the
622 Frequency : Application do stride read on lustre
624 Description: The read performance will drop a lot if the application does
626 Details : Because the stride_start_offset are missing in stride read-ahead,
627 it will cause clients read a lot of unused pages in read-ahead,
628 then the read-performance drops.
632 Description: more ldlm soft lockups
633 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
634 starve other threads from the resource lock for a long time in
635 case of long waiting queue, so change the debug level from
636 D_OTHER to the less frequently used D_INFO.
638 Severity : enhancement
640 Description: add -gid, -group, -uid, -user options to lfs find
642 Severity : enhancement
644 Description: ll_recover_lost_found_objs - recover objects in lost+found
645 Details : OST corruption and subsequent e2fsck can leave objects in the
646 lost+found directory. Using the "ll_recover_lost_found_objs"
647 tool, these objects can be retrieved and data can be salvaged
648 by using the object ID saved in the fid EA on each object.
653 Description: this bug _only_ happens when inode quota limitation is very low
654 (less than 12), so that inode quota unit is 1 at initialization.
655 Details : if remaining quota equates 1, it is a sign to demonstate that quota
656 is effective now. So least quota qunit should be 2.
660 Description: Hung threads in invalidate_inode_pages2_range
661 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
662 one is completed. As a result, some RPCs are stuck in the queue
667 Description: Procfs and llog threads access destoryed import sometimes.
668 Details : Sync the import destoryed process with procfs and llog threads by
669 the import refcount and semaphore.
673 Description: mds fails to respond, threads stuck in ldlm_completion_ast
674 Details : Sort source/child resource pair after updating child resource.
679 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
680 Details : If insertion of an extent fails, then discard the inode
681 preallocation and free data blocks else it can lead to duplicate
686 Description: don't always update ctime in ext3_xattr_set_handle()
687 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle()
688 In some cases the ctime should not be updated, for example for
689 2.0->1.8 compatibility it is necessary to delete an xattr and it
690 should not update the ctime.
694 Description: add quota statistics
695 Details : 1. sort out quota proc entries and proc code.
696 2. add quota statistics
701 Description: quotas are not honored with O_DIRECT
702 Details : all writes with the flag O_DIRECT will use grants which leads to
703 this problem. Now using OBD_BRW_SYNC to guard this.
707 Bugzilla : 15713/16362
708 Description: Assertion in iopen_connect_dentry in 1.6.3
709 Details : looking up an inode via iopen with the wrong generation number can
710 populate the dcache with a disconneced dentry while the inode
711 number is in the process of being reallocated. This causes an
712 assertion failure in iopen since the inode's dentry list contains
713 both a connected and disconnected dentry.
717 Description: assertion failure in ldlm_handle2lock()
718 Details : fix a race between class_handle_unhash() and class_handle2object()
719 introduced in lustre 1.6.5 by bug 13622.
721 Severity : enhancement
723 Description: superblock lock contention with many SMP cores on one client
724 Details : several client filesystem locks were highly contended on SMP
725 NUMA systems with 8 or more cores. Per-CPU datastructures
726 and more efficient locking implemented to reduce contention.
731 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
732 Details : remove the limit of 256 scsi disks in the sd_iostat patch
737 Description: oops in sd_iostats_seq_show()
738 Details : unloading/reloading the scsi low level driver triggers a kernel
739 bug when trying to access the sd iostat file.
744 Description: Kernel panics during QLogic driver reload
745 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
746 patch, causing memory corruption.
751 Description: journal_dev option does not work in b1_6
752 Details : pass mount option during pre-mount.
754 Severity : enhancement
756 Description: Add a FIEMAP(FIle Extent MAP) ioctl for ldiskfs
757 Details : FIEMAP ioctl will allow an application to efficiently fetch the
758 extent information of a file. It can be used to map logical blocks
759 in a file to physical blocks in the block device.
762 Frequency : only with adaptive timeout enabled
764 Description: DEBUG_REQ() bad paging request
765 Details : ptlrpc_at_recv_early_reply() should not modify req->rq_repmsg
766 because it can be accessed by reply_in_callback() without the
770 Frequency : only on Cray X2
772 Description: X2 build failures
773 Details : fix build failures on Cray X2.
777 Description: xid & resent requests
778 Details : Initialize RPC XID from clock at startup (randomly if clock is
783 Description: quota recovery deadlock during mds failover
784 Details : This patch includes att18982, att18236, att18237 in bz14840.
786 1. fix osts hang when mds does failover with quotaon
787 2. prevent watchdog storm when osts threads wait for the
792 Description: kernel panic on racer
793 Details : Do not access dchild->d_inode when IS_ERR(dchild) is true.
795 Severity : enhancement
797 Description: Add lustre_start utility to start or stop multiple Lustre servers
802 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
803 Details : In case of memory pressure, list_del() can be called twice on
804 req->rq_history_list, causing a kernel oops.
808 Description: (ptllnd_peer.c:557:kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
809 Details : fix stack overflow in the distributed lock manager by defering
810 export eviction after a failed ast to the elt thread instead of
811 handling it in the dlm interpret routine.
813 Severity : enhancement
815 Description: More exported tunables for mballoc
816 Details : Add support for tunable preallocation window and new tunables for
821 Description: Detect corruption of block bitmap and checking for preallocations
822 Details : Checks validity of on-disk block bitmap. Also it does better
823 checking of number of applied preallocations. When corruption is
824 found, it turns filesystem readonly to prevent further corruptions.
828 Frequency : only for big-endian servers
829 Description: Check for big-endian system while mounting fs with extents feature
830 Details : Mounting a filesystem with extents feature will fail on big-endian
831 systems since ext3-based ldiskfs is not supported on big-endian
832 systems. Can be over-ridden with "bigendian_extents" mount option.
836 Description: Excessive recovery window
837 Details : With AT enabled, the recovery window can be excessively long (6000+
838 seconds). To address this problem, we no longer use
839 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
840 timeout no longer depends on the service time, it is set to
841 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
842 time via pb_service_time.
846 Description: Watchdog triggered on MDS failover
847 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
848 the OSTs know that the MDS "UUID" can be reused for the same export
849 from a different NID, so we do not need to wait for the export to
852 -------------------------------------------------------------------------------
854 2008-05-26 Sun Microsystems, Inc.
856 * Support for kernels:
857 2.6.5-7.311 (SLES 9),
858 2.6.9-67.0.7.EL (RHEL 4),
859 2.6.16.54-0.2.5 (SLES 10),
860 2.6.18-53.1.14.el5 (RHEL 5),
861 2.6.22.14 vanilla (kernel.org)
862 * Client support for unpatched kernels:
863 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
864 2.6.16 - 2.6.22 vanilla (kernel.org)
865 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
866 we do not recommend using patchless RHEL4 clients with kernels
867 prior to 2.6.9-55EL (RHEL4U5).
868 * Recommended e2fsprogs version: 1.40.7-sun1
869 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
870 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
871 removed cwd "./" (refer to Bugzilla 14399).
872 * A new quota file format has been introduced in 1.6.5.
873 The format conversion from prior releases is handled transparently,
874 but releases older than 1.4.12/1.6.5 will not understand this new
875 format. The automatic format conversion can be avoided by running
876 the following command on the MDS before upgrading:
877 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
878 For more information, please refer to bugzilla 13904.
882 Description: quota performance fix
883 Details : quota data is written in journalled mode instead of ordered to
888 Description: lfs support for human-readable quota grace time strings
889 Details : lfs setquota -t and lfs quota -t represent quota grace times
890 in "XXwXXdXXhXXmXXs" format instead of large values in seconds
893 Frequency : always with o2ib 1.3 and sles10
895 Description: fix build with SLES10 and o2ib v3.
896 Details : sles10 uses diffrent name for Module.symver file but configure
897 assume this file has same name on RHEL/SLES/vanila kernels.
900 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
902 Description: MDS may lose file striping (and hence file data) in some cases
903 Details : If there are additional extended attributes stored on the MDS,
904 in particular ACLs, SELinux, or user attributes (if user_xattr
905 is specified for the client mount options) then there is a risk
906 of attribute loss. Additionally, the Lustre file striping
907 needs to be larger than default (e.g. striped over all OSTs),
908 and an additional attribute must be stored initially in the
909 inode and then increase in size enough to be moved to the
910 external attribute block (e.g. ACL growing in size) for file
913 Severity : enhancement
915 Description: add message levels for liblustreapi
920 Description: MDT cannot be unmounted, reporting "Mount still busy"
921 Details : Mountpoint references were being leaked during open reply
922 reconstruction after an MDS restart. Drop mountpoint reference
923 in reconstruct_open() and free dentry reference also.
928 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
929 Details : Move the 'good_osts' check before the 'total_bavail' check. This
930 will result in an -EAGAIN and in the exit call path we call
931 alloc_rr() which will with increasing aggressiveness attempt to
932 aquire precreated objects on the minimum number of required OSCs.
936 Description: Use old size assignment to avoid deadlock
937 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
938 the scheduling while holding a spinlock. We do not need locking
939 for size in ll_update_inode() because size is only updated from
940 the MDS for directories or files without objects, so there is no
941 other place to do the update, and concurrent access to such inodes
942 are protected by the inode lock.
946 Description: Use __u64 instead of int for valid bits
950 Description: resolve "_IOWR redefined" build error on SLES10
954 Description: dump the memory debugging after all modules are unloaded to
955 suppress false negative in conf_sanity test 39
959 Description: the recovery timer never expires
960 Details : for new client connect request, the recovery timer should not be
961 reset, otherwise recovery timer will never expired, if the old
962 client never come. Only old client connect and first connection
963 req should trigger recovery timer reset.
967 Description: the min numbers of lproc stats are wrong
968 Details : adding a new constant LC_MIN_INIT and use it for initialization
972 Frequency : always with interactive lfs
974 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
977 Frequency : with multiple concurrent readdir processes in same directory
978 Bugzilla : 15406, 15169, 15175
979 Description: misc fixes for directory readahead.
980 Details : prevent previous statahead async RPC callback from processing the
981 current "statahead_info", race condition between async RPC callback
982 add dentry into dentry hash table and "ls" thread revalidate such
983 dentry, statahead his/miss control for hidden items, and so on.
985 Severity : enhancement
987 Description: build kernel-ib packages for OFED 1.3 in our release cycle
991 Description: incore types cleaning in quota code (with respect to 64-bit limits)
992 Details : several u32 variables declarations are replaced with u64 declarations
997 Description: fix SLES kernel versioning
998 Details : the kernel version for our SLES 10 kernel did not include a "-"
999 before the "smp" at the end. while this was not a problem in
1000 general, it did mean that software trying to use the kernel
1001 version to try to detect a vendor specific kernel would fail.
1002 this was most evident by the OFED build scripts.
1007 Description: Don't update lov_desc members until making sure they are valid
1008 Details : When updating lov_desc members via proc fs, need fix their
1009 validities before doing the real update.
1012 Frequency : very rare
1014 Description: don't put request into delay list while invalidate in flight.
1015 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1016 import in flight. this produce timeout for invalidate and sometimes
1017 can cause stale data.
1019 Severity : enhancement
1021 Description: Update kernel to SLES9 2.6.5-7.311.
1023 Severity : enhancement
1025 Description: Update kernel to RHEL4 2.6.9-67.0.7.
1030 Frequency : on PPC only
1031 Description: not convert ost objects for directory because it's not exist.
1032 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1034 Severity : enhancement
1036 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
1037 __u64 as usingied long long type.
1040 Frequency : rare, on shutdown
1042 Description: race process ast vs remove callback
1043 Details : removing callback before disconnect import open race with
1044 processing callback.
1046 Severity : enhancement
1048 Description: Update kernel to SLES9 2.6.5-7.311.
1050 Severity : enhancement
1052 Description: Files open for execute are not marked busy on SLES10
1053 Details : Add FMODE_EXEC to SLES10 SP1 server kernel series.
1055 Severity : enhancement
1057 Description: Add server support for vanilla-2.6.22.14.
1060 Frequency : occasional
1062 Description: Avoid lov_create() getting stuck in obd_statfs_rqset()
1063 Details : If an OST is down the MDS will hang indefinitely in
1064 obd_statfs_rqset() waiting for the statfs data. While for
1065 MDS QOS usage of statfs, it should not stuck in waiting.
1067 Severity : enhancement
1069 Description: Disable adaptive timeouts by default
1072 Frequency : on network error
1074 Description: panic with double free request if network error
1075 Details : mdc_finish_enqueue is finish request if any network error occuring,
1076 but it's true only for synchronus enqueue, for async enqueue
1077 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1081 Frequency : rare, on recovery
1083 Description: read procfs can produce deadlock in some situation
1084 Details : Holding lprocfs lock with send rpc can produce block for destroy
1085 obd objects and this also block reconnect with -EALREADY.
1086 This isn't fix all lprocfs bugs - but make it rare.
1088 Severity : enhancement
1090 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1093 Frequency : frequent on X2 node
1095 Description: mdc_set_open_replay_data LBUG
1096 Details : Set replay data for requests that are eligible for replay.
1101 Description: lustre_mgs: operation 101 on unconnected MGS
1102 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1103 MGC, and late on MGC cannot successfully connect to MGS and a lot
1104 of the error messages complaining that MGS is not connected.
1107 Frequency : rare, depends on device drivers and load
1109 Description: MDS or OSS nodes crash due to stack overflow
1110 Details : Code changes in 1.6.4 increased the stack usage of some functions.
1111 In some cases, in conjunction with device drivers that use a lot
1112 of stack, the MDS (or possibly OSS) service threads could overflow
1113 the stack. One change which was identified to consume additional
1114 stack has been reworked to avoid the extra stack usage.
1116 Severity : enhancement
1118 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1120 Severity : enhancement
1122 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1124 Severity : enhancement
1126 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1128 Serverity : enhancement
1130 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1133 Frequency : rare on shutdown OST
1135 Description: Don't allow skipping OSTs if index has been specified.
1136 Details : Don't allow skipping OSTs if index has been specified, make locking
1137 in internal create lots better.
1142 Description: ASSERTION(!PageDirty(page)) failed
1143 Details : Wrong check could lead to an assertion failure under specific
1149 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1150 Details : access to bitfield in structure is always rounded to long
1151 and this produce problem with not atomic change any bit.
1156 Description: Lustre make rpms failed.
1157 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1158 builds Lustre rpms from tarball.
1160 Severity : enhancement
1162 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
1165 Frequency : rare on shutdown OST
1167 Description: If llog cancel was not send before clean_exports phase, this can
1168 produce deadlock in llog code.
1169 Details : If llog thread has last reference to obd and call class_import_put
1170 this produce deadlock because llog_cleanup_commit_master wait when
1171 last llog_commit_thread exited, but this never success because was
1172 called from llog_commit_thread.
1175 Frequency : only if OST index is skipped
1177 Description: NULL lov_tgts causing MDS oops
1178 Details : more safe checks for NULL lov_tgts for avoid oops.
1180 Severity : enhancement
1182 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1184 Severity : enhancement
1186 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1191 Description: make mgs_setparam() handle fsname containing dash
1192 Details : fsname containing a dash does not work with lctl conf_param
1194 Severity : enhancement
1196 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
1199 Frequency : rare, in recovery and (or) destroy lovobjid file.
1201 Description: rewrite lov objid code.
1202 Details : Cleanup for lov objid code, remove scability problems and wrong
1203 locking. Fix sending last_id into ost.
1205 Severity : enhancement
1207 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1209 Severity : enhancement
1211 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1212 Details : Use d_move_locked instead of __d_move.
1215 Frequency : rare, at shutdown
1217 Description: access already free / zero obd_namespace.
1218 Details : if client_disconnect_export was called without force flag set,
1219 and exist connect request in flight, this can produce access to
1220 NULL pointer (or already free pointer) when connect_interpret
1221 store ocd flags in obd_namespace.
1224 Frequency : only at startup
1226 Description: not alloc memory with spinlock held.
1227 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1228 if any spinlock held.
1233 Description: lfs find does not continue on file error
1234 Details : Continue other files processing when a file/dir is absent.
1238 Description: Inconsistent usage of lustre_pack_reply()
1239 Details : Standardize the usage of lustre_pack_reply() such that it
1240 always generate a CERROR on failure.
1243 Frequency : very rare
1245 Description: Fix replay if there is an un-replied request and open
1246 Details : In some cases, older replay request will revert the
1247 mcd->mcd_last_xid on MDS which is used to record the client's
1248 latest sent request.
1250 Severity : enhancement
1252 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1254 Severity : enhancement
1256 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1258 Severity : enhancement
1260 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1262 Severity : enhancement
1264 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
1266 Severity : enhancement
1268 Description: Build SLES10 patchless client fails
1269 Details : The configure was broken by run ./configure with
1270 --with-linux-obj=.... argument for patchless client. When the
1271 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1272 can't search header adequately. Use absolute path such as
1273 -I($LINUX)/include instead.
1275 Severity : enhancement
1277 Description: Lustre Page Accounting
1278 Details : New macros for page alloc and free which enable accounting
1279 of page allocation of Lustre. Use percpu counters to store memory
1280 and page statistics.
1283 Frequency : only if debugging is disabled
1285 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1286 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1287 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1288 these macros and replace them with their definitions instead.
1293 Description: interrupt oig_wait produce painc on resend.
1294 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1295 private set, and this produce situation when rq_ptlrpcd_data not
1296 copyed to new allocated request and triggered LBUG on assert
1297 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1298 ptlrpcd set if request is changed and can be safety changed to use
1301 Severity : enhancement
1303 Description: organize the server-side client stats on per-nid basis
1304 Details : Change the structure of stats under obdfilter and mds to
1312 The "uuid"s file would list the uuids of _active_ exports.
1313 And the clear entry is to clear all stats and stale nids.
1318 Description: Processes looping in ll_readdir() on Lustre clients finally causing
1319 a full node pseudo-hang
1320 Details : Concurrent access to the same directory from multiple clients with
1321 intensive file creation/removal can cause a client node to spin in
1322 ll_readdir(). i_version must be increased every time the lock is
1323 cancelled to ensure a revalidate is done.
1328 Description: touch file failed when fs is not full
1329 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1330 otherwise we can get ENOSP while fs is not full.
1333 Frequency : only for Cray XT3
1334 Bugzilla : 12829/13455
1335 Description: Changing primary group doesn't change the group lustre assigns to
1337 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1338 overridden with the primary group provided by the group upcall,
1339 whereas the supplied fsgid can be trusted if it is in the list of
1340 supplementary groups returned by the group upcall.
1342 Severity : enhancement
1344 Description: Root Squash Functionality
1345 Details : Implementation of NFS-like root squash capability. Specifically,
1346 don't allow someone with root access on a client node to be able
1347 to manipulate files owned by root on a server node.
1349 Severity : enhancement
1351 Description: Slow trucate/writes to huge files at high offsets.
1352 Details : Directly associate cached pages to lock that protect those pages,
1353 this allows us to quickly find what pages to write and remove
1354 once lock callback is received.
1359 Description: Too many locks accumulating on client during NFS usage
1360 Details : mds_open improperly used accmode to find out access mode to a
1361 file. Also mdc_intent_lock logic to find out if we already have
1362 lock similar to just received was flawed since introduction of
1363 skiplists - locks are now added to the front of the granted
1368 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
1370 Details : In obd setup failure handler, obd_stopping will not necessarily to
1371 be 1, and obd_set_up should also be checked to make sure whether
1372 obd is completely setup.
1374 Severity : enhancement
1376 Description: Allow masking D_WARNING, D_ERROR messages from console
1377 Details : Console messages can now be disabled via lnet.printk.
1382 Description: User code with malformed file open parameter crashes client node
1383 Details : Before packing join_file req, all the related reference should be
1384 checked carefully in case some malformed flags cause fake join_file
1390 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
1391 grain of quota doing harm to quota's accuracy
1392 Details : qunit size will be changed when quota limitation is too low/high;
1393 record the pending quota write in order to get more accureate
1394 quota; delete the patch for bug12588, which is unnecessary when
1395 this patch is landed. This bug also contains fixes for bug 14526,
1396 14299, 14601 and 13794.
1400 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1401 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1402 mode is equal to the granted mode. However, it should also check
1403 that LDLM_FL_CP_REQD is not set.
1407 Description: Heavy nfs access might result in deadlocks
1408 Details : After ELC code landed, it is now improper to enqueue any mds
1409 locks under och_sem, because enqueue might want to decide to
1410 cancel open locks for same inode we are holding och_sem for.
1414 Description: 35% write performance drop with ldiskfs2 when quotas are on
1415 Details : Enable ext3 journalled quota by default to improve performance
1416 when quotas are turned on.
1420 Description: Client eviction while running blogbench
1421 Details : A lot of unlink operations with concurrent I/O can lead to a
1422 deadlock causing evictions. To address the problem, the number of
1423 oustanding OST_DESTROY requests is now throttled to
1424 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1425 callbacks are processed in priority.
1428 Frequency : RHEL4 only
1430 Description: mkfs is very slow on IA64/RHEL4
1431 Details : A performance regression has been discovered in the MPT Fusion
1432 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1433 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1434 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1437 Frequency : PPC/PPC64 only
1439 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1440 Details : fix duplicated definitions between asm-ppc64/types.h and
1441 lustre_types.h on PPC.
1444 Frequency : PPC/PPC64 only
1446 Description: asm-ppc/segment.h does not exist
1447 Details : fix compile issue on PPC.
1451 Description: data checksumming impacts single node performance
1452 Details : add support for several checksum algorithms. Currently, CRC32 and
1453 Adler-32 are supported. The checksum type can be changed on the fly
1454 through /proc/fs/lustre/osc/*/checksum_type.
1458 Description: use adler32 for page checksums
1459 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1464 Description: better handle error messages in extents code
1466 Severity : enhancement
1468 Description: SNMP support enhancement
1469 Details : Adding total number of sampled request for an MDS node in snmp
1472 Severity : enhancement
1474 Description: Optimize ldlm waiting list processing for PR extent locks
1475 Details : When processing waiting list for read extent lock and meeting read
1476 lock that is same or wider to it that is not contended, skip
1477 processing rest of the list and immediatelly return current
1478 status of conflictness, since we are guaranteed there are no
1479 conflicting locks in the rest of the list.
1483 Description: Time out and refuse to reconnect
1484 Details : When the failover node is the primary node, it is possible
1485 to have two identical connections in imp_conn_list. We must
1486 compare not conn's pointers but NIDs, otherwise we
1487 can defeat connection throttling.
1491 Description: Client not clear own cache if answer to reconnect is lost.
1492 Details : Client gets evicted from server. Now client also thinks it is
1493 disconnected (or gets ENOTCONN on its operation) and decides to
1494 reconnect. Server receives reconnect message, but cannot find
1495 export. New export is created that is fully valid (new cookie!),
1496 but reply is lost and not reported to client. Client reconnects
1497 again and gets back a just-created connection, but it is not new
1498 so client thinks it was not evicted and does not do recovery.
1502 Description: Detect stride IO mode in read-ahead
1503 Details : When a client does stride read, read-ahead should detect that and
1504 read-ahead pages according to the detected stride pattern.
1508 Description: build for x2 fails
1509 Details : fix compile issue on Cray systems.
1511 Severity : enhancement
1513 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
1514 Details : This greatly improves speed of NFS writes on 2.6 kernels.
1517 Frequency : only on PPC/SLES10
1519 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1520 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1521 be defined. Add a hack in mkfs_lustre.c to work around this compile
1526 Description: LASSERT on MDS when client holding flock lock dies
1527 Details : ldlm pool logic depends on number of granted locks equal to
1528 number of released locks which is not true for flock locks, so
1529 just exclude such locks from consideration.
1533 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1534 Details : Use fsfilt_sync() for both the whole filesystem sync and
1535 individual file sync to eliminate dangerous inode locking
1536 with I_LOCK that can lead to a deadlock.
1540 Description: Update an obsolete wirecheck.c generator
1541 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
1545 Description: Client can panic on open sometimes
1546 Details : It is possible that we try to free already freed request in
1547 ll_file_open in some error cases when we send request from
1552 Description: performance in 1.6.3
1553 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1554 sure that 1MB requests are not fragmented by the block layer.
1558 Description: LDLM soft lockups - improvement
1559 Details : It is be possible to send the lock handle along with each read
1560 or write request because the client is already doing a lock match
1561 itself so there isn't any reason the OST should have to re-do that
1567 Description: lfs quota fails with deactivated OSTS
1568 Details : With this patch, three improvements are included:
1569 1. detete the softlimit in mds and osts when use "lfs quota".
1570 2. display the inaccurate data in the output of "lfs quota".
1571 3. try to get quota info when "lfs quota" is executed.
1576 Description: Extent locks not granted with no conflicts sometimes.
1577 Details : When race occurs in glimpse handler and nothing is returned,
1578 we do not reprocess the queue after lock cancel, and that leads
1579 to a stall until next activity on a resource
1582 Frequency : failover with quotaon
1584 Description: during mds failovers with quota on, OSTs got into deadlock state
1585 and causing dumpstack.
1586 Details : for every quota slave, at any time, there is only one quota req
1587 is sent to quota master for every uid/gid. Before that quota req
1588 returns, all the thread relative to the same uid/gid will wait.
1589 So if the quota req is lost because mds failovers or any other
1590 reasons, this bug will be hit. Now, dqacq_interpret() will handle
1591 quota reqs who time out.
1593 Severity : enhancement
1596 Description: when quota slave checks if quota is enough, there is an unnecessary
1598 Details : place this wait on necessary place instead of always waiting.
1600 --------------------------------------------------------------------------------
1602 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
1604 * Support for kernels:
1605 2.6.5-7.286 (SLES 9),
1606 2.6.9-55.0.9.EL (RHEL 4),
1607 2.6.16.53-0.8 (SLES 10),
1608 2.6.18-8.1.14.el5 (RHEL 5),
1609 2.6.18.8 vanilla (kernel.org)
1610 * Client support for unpatched kernels:
1611 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1612 2.6.16 - 2.6.22 vanilla (kernel.org)
1613 * Due to recently discovered recovery problems, we do not recommend
1614 using patchless RHEL 4 clients with this or any earlier release.
1615 * Recommended e2fsprogs version: 1.40.2-cfs1
1616 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1619 Frequency : occasional
1621 Description: MDS hang or stay in waiting lock
1622 Details : If client receive lock with CBPENDING flag ldlm need send lock
1623 cancel as separate rpc, to avoid situation when cancel request
1624 can't processed due all i/o threads stay in wait lock.
1627 Frequency : occasional
1629 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1630 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1631 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1632 oscc_has_objects and see OSC has no objests and call
1633 oscc_internal_create to resend create request.
1635 Severity : enhancement
1637 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1638 Details : Update which_patch & target file for SLES10 latest kernel.
1640 Severity : enhancement
1642 Description: add --type and --size parameters to lfs find
1643 Details : Enhance lfs find by adding filetype and filesize parameters. Also
1644 multiple OBDs can now be specified for the --obd option.
1646 Severity : enhancement
1648 Description: eliminate client locks in face of contention
1649 Details : file contention detection and lockless i/o implementation
1650 for contended files.
1652 Severity : enhancement
1654 Description: Remove client patches from SLES 10 kernel.
1655 Details : This causes SLES 10 clients to behave as patchless clients
1656 even on a Lustre-patched (server) kernel.
1658 Severity : enhancement
1660 Description: use i_size_read and i_size_write in 2.6 port
1661 Details : replace inode->i_size access with i_size_read/write()
1664 Frequency : when removing large files
1666 Description: scheduling issue during removal of large Lustre files
1667 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1668 It causes scheduling issues when removing large files (17TB in the
1674 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
1675 Details : Impossible (EPROTO is returned) to access a directory that has a
1676 non-default striping and ACLs.
1679 Frequency : only on ppc
1681 Description: /proc/fs/lustre/devices broken on ppc
1682 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1683 We should make sure the type of 'index' is loff_t and then cast
1684 explicitly as needed below. Do not assign an explicitly cast
1688 Frequency : only for rhel5
1690 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1691 Details : Modify the target file & which_kernel.
1694 Frequency : if the uninit_groups feature is enabled on ldiskfs
1696 Description: e2fsck reports "invalid unused inodes count"
1697 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
1698 feature and only a single inode is created in a group then the
1699 "bg_unused_inodes" count is incorrectly updated. Creating a
1700 second inode in that group would update it correctly.
1705 Description: buffer overruns could theoretically occur
1706 Details : llapi_semantic_traverse() modifies the "path" argument by
1707 appending values to the end of the origin string, and a buffer
1708 overrun may occur. Adding buffer overrun check in liblustreapi.
1710 Severity : enhancement
1712 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
1715 Frequency : only if filesystem is inconsistent
1717 Description: handle "serious error: objid * already exists" more gracefully
1718 Details : If LAST_ID value on disk is smaller than the objects existing in
1719 the O/0/d* directories, it indicates disk corruption and causes an
1720 LBUG(). If the object is 0-length, then we should use the existing
1721 object. This will help to avoid a full fsck in most cases.
1723 Severity : enhancement
1725 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1726 Details : Modify vm-tunables-rhel4.patch.
1728 Severity : enhancement
1730 Description: Kernel config for 2.6.18-vanilla.
1731 Details : Modify targets/2.6-vanilla.target.in.
1732 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1733 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1734 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1735 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1738 Frequency : occasional
1740 Description: improve handling recoverable errors
1741 Details : If request processed with error which can be recoverable on server
1742 request should be resend, otherwise page released from cache and
1748 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1749 Details : Modify target file & which_patch.
1750 A flaw was found in the IA32 system call emulation provided
1751 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1752 value could be stored in the %RAX register, which could trigger an
1753 out-of-bounds system call table access. An untrusted local user
1754 could exploit this flaw to run code in the kernel
1755 (ie a root privilege escalation). (CVE-2007-4573).
1759 Description: change order of libsysio includes
1760 Details : '#include sysio.h' should always come before '#include xtio.h'
1762 Severity : enhancement
1764 Description: adapt the lustre_config script to support the upgrade case
1765 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1766 server targets to 1.6 in parallel.
1771 Description: To avoid grant space > avaible space when the disk is almost
1772 full. Without this patch you might see the error "grant XXXX >
1773 available" or some LBUG about grant, when the disk is almost
1775 Details : In filter_check_grant, for non_grant cache write, we should
1776 check the left space by if (*left > ungranted + bytes), instead
1777 of (*left > ungranted), because only we are sure the left space
1778 is enough for another "bytes", then the ungrant space should be
1779 increase. In client, we should update cl_avail_grant only there is
1780 OBD_MD_FLGRANT in the reply.
1785 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1786 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1787 described in CVE-2007-4573. This problem could allow untrusted
1788 local users to gain root access.
1791 Frequency : when using O_DIRECT and quotas
1793 Description: Incorrect file ownership on O_DIRECT output files
1794 Details : block usage reported by 'lfs quota' does not take into account
1795 files that have been written with O_DIRECT.
1799 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
1800 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
1801 window to 0. If it does hit cache miss, and read-pattern is still
1802 stride-io mode, does not reset the stride window, but also does not
1803 increase the stride window length in this case.
1805 --------------------------------------------------------------------------------
1807 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
1809 * Support for kernels:
1810 2.6.5-7.286 (SLES 9),
1811 2.6.9-55.0.2.EL (RHEL 4),
1812 2.6.16.46-0.14 (SLES 10),
1813 2.6.18-8.1.8.el5 (RHEL 5),
1814 2.6.18.8 vanilla (kernel.org)
1815 * Client support for unpatched kernels:
1816 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1817 2.6.16 - 2.6.21 vanilla (kernel.org)
1818 * Due to recently discovered recovery problems, we do not recommend
1819 using patchless RHEL 4 clients with this or any earlier release.
1820 * Recommended e2fsprogs version: 1.40.2-cfs1
1821 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1825 Description: Fix errors in lfs documentation
1826 Details : Fixes man pages
1828 Severity : enhancement
1830 Description: Adaptive timeouts
1831 Details : RPC timeouts adapt to changing server load and network
1832 conditions to reduce resend attempts and improve recovery time.
1834 Severity : enhancement
1836 Description: llapi_file_create() does not allow some changes
1837 Details : add llapi_file_open() that allows specifying the file creation
1838 mode and open flags, and also returns an open file handle.
1840 Severity : enhancement
1842 Description: df doesn't work properly if diskfs blocksize != 4K
1843 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1846 Frequency : always on directories with default striping set
1848 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1849 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1851 Severity : enhancement
1853 Description: merge and cleanup kernel patches.
1854 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1856 Severity : enhancement
1858 Description: RedHat Update kernel for RHEL5
1859 Details : Kernel config file for RHEL5.
1861 Severity : enhancement
1863 Description: OSS needs mutliple precreate threads
1864 Details : Add ability to start more than one create thread per OSS.
1867 Frequency : only with quota on the root user
1869 Description: mds_obd_create error creating tmp object
1870 Details : When the user sets quota on root, llog will be affected and can't
1871 create files and write files.
1874 Frequency : Always on ia64 patchless client, and possibly others.
1876 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1877 Details : This allows the patchless client to be loaded on architectures
1878 without this export.
1883 Description: disorder of journal start and llog_add cause deadlock.
1884 Details : in llog_origin_connect, journal start should happen before
1885 llog_add keep the same order as other functions to avoid
1888 Severity : enhancement
1890 Description: RedHat Update kernel for RHEL5
1891 Details : Modify the kernel config file more closer RHEL5.
1893 Severity : enhancement
1895 Description: Build failure against Centos5 (RHEL5)
1896 Details : Define PAGE_SIZE when it isn't present.
1899 Frequency : occasionally when using NFS
1901 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1902 Details : with NFS, the anon dentry's parent was set to itself in
1903 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1904 corresponding dentry other than use rec->ur_name.
1906 Severity : enhancement
1908 Description: enable data checksumming by default
1909 Details : enable checksum by default, allow --disable-checksum
1910 configure option and "-o nochecksum" mount option. Checksums
1911 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1914 Frequency : Occasionally with failover
1916 Description: Client eviction due to failover config
1917 Details : after a connection loss, the lustre client should attempt to
1918 reconnect to the last active server first before trying the
1919 other potential connections.
1921 Severity : enhancement
1923 Description: client-side metadata stat-ahead during readdir(directory readahead)
1924 Details : perform client-side metadata stat-ahead when the client detects
1925 readdir and sequential stat of dir entries therein
1930 Description: when mds and osts use different quota unit(32bit and 64bit),
1931 quota will be released repeatly.
1932 Details : void sending multiple quota reqs to mds, which will keep the
1933 status between the reqs.
1936 Frequency : only with liblustre clients on XT3
1938 Description: evictions taking too long
1939 Details : allow llrd to evict clients directly on OSTs
1943 Description: osts not allocated evenly to files
1944 Details : change the condition to increase offset_idx
1946 Severity : enhancement
1948 Description: self-adjustable client's lru lists
1949 Details : use adaptive algorithm for managing client cached locks lru
1950 lists according to current server load, other client's work
1951 pattern, memory activities, etc. Both, server and client
1952 side namespaces provide number of proc tunables for controlling
1957 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
1958 Details : rewrite inherited code (uuid parsing code from ext2 utils and
1959 readdir code from ext3) from scratch preserving functionality.
1963 Description: Only those disconnect error should be returned by rq_status.
1964 Details : In open/enqueue processs, Some errors, which will cause client
1965 disconnected, should be returned by rq_status, while other
1966 errors should still be returned by intent, then mdc or llite will
1969 Severity : enhancement
1971 Description: Tune the kernel for good SCSI performance.
1972 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1973 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1977 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1978 Bugzilla : 13547, 13627
1979 Description: Data corruption for OSTs that are formatted larger than 2TB
1981 Details : When generating the bio request for lustre file writes the
1982 sector number would overflow a temporary variable before being
1983 used for the IO. The data reads correctly from Lustre (which
1984 will overflow in a similar manner) but other file data or
1985 filesystem metadata may be corrupted in some cases.
1989 Description: "lfs find -obd UUID" prints directories
1990 Details : "lfs find -obd UUID" will return all directory names instead
1991 of just file names. It is incorrect because the directories
1992 do not reside on the OSTs.
1995 Bugzilla : 13596, 13614
1996 Description: MDS hang after unclean shutdown of lots of clients
1997 Details : Never resend AST requests.
2001 Frequency : Always, for kernels after 2.6.16
2002 Description: Fix warning idr_remove called for id=.. which is not allocated.
2003 Details : Last kernels save old s_dev before kill super and not allow
2004 to restore from callback - restore it before call kill_anon_super.
2008 Description: Fix error on 'ls .' at the top of the Lustre mount.
2009 Details : Don't revalidate dentry if it is a root dentry.
2011 --------------------------------------------------------------------------------
2013 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
2015 * Support for kernels:
2016 2.6.5-7.286 (SLES 9),
2017 2.6.9-55.0.2.EL (RHEL 4),
2018 2.6.16.46-0.14 (SLES 10),
2019 2.6.18-8.1.8.el5 (RHEL 5),
2020 2.6.18.8 vanilla (kernel.org)
2021 * Client support for unpatched kernels:
2022 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2023 2.6.16 - 2.6.22 vanilla (kernel.org)
2024 * Due to recently discovered recovery problems, we do not recommend
2025 using patchless RHEL 4 clients with this or any earlier release.
2026 * Recommended e2fsprogs version: 1.39.cfs8
2027 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2032 Description: block reactivating mgc import until all deactivates complete
2033 Details : Fix race when failing back MDT/MGS to itself (testing)
2035 Severity : enhancement
2037 Description: lfs setstripe enhancement
2038 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2041 Frequency : mds/oss recovery
2043 Description: llog ctxt is refrenced after it has been freed.
2044 Details : llog ctxt refcount was added to avoide the race between ctxt free
2045 and llog recovery process. Each llog user must hold ctxt refcount
2046 before it access the llog. And the llog ctxt can only be freed
2047 when its refcount is zero.
2049 Severity : enhancement
2051 Description: randomly memory allocation failure util
2052 Details : Make lustre randomly failed allocating memory for testing purpose.
2055 Frequency : only for SLES10
2057 Description: Update kernel patch for SLES10 SP1
2058 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
2060 Severity : enhancement
2062 Description: omit set fsid for export NFS
2063 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
2064 as FS_REQUIRES_DEV for avoid problems with generate fsid.
2067 Frequency : after network failures
2069 Description: Add sync option to mount_lustre.c
2070 Details : Client loses data written to lustre after a network interruption.
2072 Severity : enhancement
2074 Description: Error message improvement.
2075 Details : Merging of two LCONSOLE_ERROR_MSG into one.
2077 Severity : enhancement
2079 Description: don't use GFP_* in generic Lustre code.
2080 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
2082 Severity : enhancement
2084 Description: obdclass is limited by single OBD_ALLOC(idarray)
2085 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
2087 Severity : enhancement
2089 Description: updated patchess for new RHEL4 kernel
2090 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
2092 Severity : enhancement
2094 Description: warnings with build patchless client with vanila 2.6.19 and up
2095 Details : change old ctl_table style and replace ctl_table/ctl_table_header
2096 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
2098 Severity : enhancement
2100 Description: O_DIRECT bypasses client statistics.
2101 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2102 read_ahead_stats were updated. Copied stats section from
2103 osc_send_oap_rpc() into async_internal().
2106 Frequency : only for Cray XT3
2108 Description: peer credits not enough on many OST per OSS systems.
2109 Details : Use new lnet way to add credits as we need those for pings and ASTs
2112 Frequency : only with liblustre
2114 Description: Liblustre is not releasing flock locks on file close.
2115 Details : Release flock locks on file close.
2118 Frequency : only for RHEL4
2120 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
2121 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
2122 Update target file and kernel config.
2126 Description: lustre support for RHEL5
2127 Details : Add support for RHEL5.
2132 Description: ASSERTION(export != NULL) failed in target_handle_connect
2133 Details : Assetion hit is result of rare race between disconnect and connet
2134 to same nid. target_handle_connect found old connect cockie and
2135 tried to reconnect, but can't find export for this cockie.
2140 Description: umount blocks forever on error
2141 Details : In result of wrong using obd_no_recov and obd_force flags client
2142 can hand if cancel or some other requests is lost.
2145 Frequency : Only for SLES
2147 Description: sanity_quota fail test_1
2148 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
2149 file, which makes TSTID[2] inunique.
2151 Severity : enhancement
2153 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2154 Details : Update target/ChangeLog/which_patch .
2156 Severity : enhancement
2158 Description: jbd statistics
2159 Details : Port older jbd statistics patch for sles10
2164 Description: lvbo_init failed for resource with missing objects.
2165 Details : Fix returning error if we do stat for file with missing/corrupted
2166 objects and i_size set to all sum of size of all avaible objects.
2167 if we to truncate/write to missing object - it is recreated.
2172 Description: Oops in read and write path when failing to allocate lock.
2173 Details : Check if lock allocation failed and return error back.
2176 Frequency : When flocks are used.
2178 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2179 Details : Flock locks might destroy just granted lock if it could be merged
2180 with another existing flock, this is done in completion handler,
2181 so teach ldlm_cli_enquque_fini that this is a valid case for
2187 Description: reply_lock_interpret crash due to race with it and lock cancel.
2188 Details : Do not replay locks that are being cancelled. Do not reference
2189 locks by their address during replay, just by their handle.
2192 Frequency : only with deactivated OSTs
2194 Description: lstripe command fails for valid OST index
2195 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
2196 instead of lov->desc.ld_active_tgt_count.
2198 Severity : enhancement
2200 Description: Build failure against Centos5 (RHEL5)
2201 Details : Use getpagesize() instead of PAGE_SIZE.
2203 --------------------------------------------------------------------------------
2205 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
2207 * Support for kernels:
2208 2.6.5-7.283 (SLES 9),
2209 2.6.9-55.EL (RHEL 4),
2210 2.6.16.46-0.14 (SLES 10),
2211 2.6.18.8 vanilla (kernel.org)
2212 * Client support for unpatched kernels:
2213 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2214 2.6.16 - 2.6.22 vanilla (kernel.org)
2215 * Due to recently discovered recovery problems, we do not recommend
2216 using patchless RHEL 4 clients with this or any earlier release.
2217 * Recommended e2fsprogs version: 1.39.cfs8
2218 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2219 * Starting with this release, the ldiskfs backing filesystem required
2220 by Lustre is now in its own package, lustre-ldiskfs. This package
2221 should be installed. It is versioned separately from Lustre and
2222 may be released separately in future.
2227 Description: server LBUG when shutting down
2228 Details : Block umount forever until the mount refcount is zero rather
2229 than giving up after an arbitrary timeout.
2231 Severity : enhancement
2233 Description: add optional extra BUILD_VERSION info
2234 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2235 to override the lustre version.
2238 Frequency : 2.6.18 servers only
2240 Description: ll_kern_mount() doesn't release the module reference
2241 Details : The ldiskfs module reference count never drops down to 0
2242 because ll_kern_mount() doesn't release the module reference.
2247 Description: server LBUG when using old ost_num_threads parameter
2248 Details : Accept the old ost_num_threads parameter but warn that it
2249 is deprecated, and fix an off-by-one error that caused an LBUG.
2254 Description: Transient SCSI error results in persistent IO issue
2255 Details : iobuf->dr_error is not reinitialized to 0 between two
2259 Frequency : sometimes when underlying device returns I/O errors
2261 Description: OSTs not going read-only during write failures
2262 Details : OSTs are not remounted read-only when the journal commit threads
2263 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2264 instead of the ext3 wrappers.
2267 Frequency : only with 10000 clients or more
2269 Description: poor connect scaling with increasing client count
2270 Details : Don't run filter_grant_sanity_check for more than 100 exports
2271 to improve scaling for large numbers of clients.
2274 Frequency : SLES10 only
2276 Description: sanity-quota.sh quotacheck failed: rc = -22
2277 Details : Quotas cannot be enabled on SLES10.
2280 Frequency : liblustre clients only
2282 Description: getdirentries does not give error when run on compute nodes
2283 Details : getdirentries does not fail when the size specified as an argument
2284 is too small to contain at least one entry
2286 Severity : enhancement
2288 Description: Add LNET router traceability for debug purposes
2289 Details : If a checksum failure occurs with a router as part of the
2290 IO path, the NID of the last router that forwarded the bulk data
2291 is printed so it can be identified.
2296 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2297 Details : Due to a race condition, liblustre clients were occasionally
2298 evicted incorrectly.
2300 Severity : enhancement
2302 Description: lfs setstripe use optional parameters instead of postional
2305 Severity : enhancement
2307 Description: Nanosecond timestamp support for ldiskfs
2308 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2309 resolution timestamps. There is not yet support for this at
2310 the Lustre filesystem level.
2313 Frequency : during server recovery
2315 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2316 Details : request with rq_no_resend flag not awake l_wait_event if they get a
2320 Frequency : nfs export on patchless client
2322 Description: connectathon hang when test nfs export over patchless client
2323 Details : Disconnected dentry cannot be found with lookup, so we do not need
2324 to unhash it or make it invalid
2327 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2328 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2329 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2334 Description: open req refcounting wrong on reconnect
2335 Details : If reconnect happened between getting open reply from server and
2336 call to mdc_set_replay_data in ll_file_open, we will schedule
2337 replay for unreferenced request that we are about to free.
2338 Subsequent close will crash in variety of ways.
2339 Check that request is still eligible for replay in
2340 mdc_set_replay_data().
2345 Description: disable writes to filesystem when reading health_check file
2346 Details : the default for reading the health_check proc file has changed
2347 to NOT do a journal transaction and write to disk, because this
2348 can cause reads of the /proc file to hang and block HA state
2349 checking on a healthy but otherwise heavily loaded system. It
2350 is possible to return to the previous behaviour during configure
2351 with --enable-health-write.
2353 Severity : enhancement
2355 Description: 64-bit inode version
2356 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2357 to the inode. This will be required for version-based recovery.
2362 Description: MDS fails to start if a duplicate client export is detected
2363 Details : in some rare cases it was possible for a client to connect to
2364 an MDS multiple times. Upon recovery the MDS would detect this
2365 and fail during startup. Handle this more gracefully.
2367 Severity : enhancement
2369 Description: Add -o localflock option to simulate old noflock behaviour.
2370 Details : This will achieve local-only flock/fcntl locks coherentness.
2375 Description: log_commit_thread vs filter_destroy race leads to crash
2376 Details : Take import reference before releasing llog record semaphore
2381 Description: Wrong request locking in request set processing
2382 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2383 imp_delayed_list, in this place should be used imp_lock.
2386 Frequency : when reconnecting
2388 Description: Grant leak when OSC reconnect to OST
2389 Details : When osc reconnect ost, OST(filter) should check whether it
2390 should grant more space to client by comparing fed_grant and
2391 cl_avail_grant, and return the granted space to client instead
2392 of "new granted" space, because client will call osc_init_grant
2393 to update the client grant space info.
2396 Frequency : when client reconnects to OST
2398 Description: Grant leak when OSC does a resend and replays bulk write
2399 Details : When osc reconnect to OST, OST(filter) should clear grant info of
2400 bulk write request, because the grant info while be sync between
2401 OSC and OST when reconnect, and we should ignore the grant info
2402 these of resend/replay write req.
2407 Description: Grant space more than avaiable space sometimes.
2408 Details : When then OST is about to be full, if two bulk writing from
2409 different clients came to OST. Accord the avaliable space of the
2410 OST, the first req should be permitted, and the second one
2411 should be denied by ENOSPC. But if the seconde arrived before
2412 the first one is commited. The OST might wrongly permit second
2413 writing, which will cause grant space > avaiable space.
2416 Frequency : when client is evicted
2418 Description: Grant might be wrongly erased when osc is evicted by OST
2419 Details : when the import is evicted by server, it will fork another
2420 thread ptlrpc_invalidate_import_thread to invalidate the
2421 import, where the grant will be set to 0. While the original
2422 thread will update the grant it got when connecting. So if
2423 the former happened latter, the grant will be wrongly errased
2424 because of this race.
2429 Description: Checking Stale with correct fid
2430 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2431 to check whether it is stale, because sometimes, we want the
2432 enqueue happened anyway, and op_data.fid2 will not be initialized.
2434 Severity : enhancement
2436 Description: update patchless client
2437 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2440 Frequency : only with 2.4 kernel
2442 Description: random memory corruption
2443 Details : size of struct ll_inode_info is to big for union inode.u and this
2444 can be cause of random memory corruption.
2449 Description: Memory leak in recovery
2450 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2451 It should also check obd_fail before fsfilt_start, otherwise if
2452 fsfilt_start return -EROFS,(failover mds during mds recovery).
2453 then the req will return with repmsg->transno = 0 and rc = EROFS.
2454 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2455 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2456 be freed no matter whether fsfilt_commit success or not.
2459 Frequency : only with large numbers of cores on a single node
2461 Description: Prevents from taking the superblock lock in llap_from_page for
2463 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2464 for llap_from_page call in ll_removepage() prevents from taking
2465 the superblock lock for a soon died page.
2470 Description: Not check open intent error before release open handle
2471 Details : in some rare cases, the open intent error is not checked before
2472 release open handle, which may cause
2473 ASSERTION(open_req->rq_transno != 0), because it tries to release
2474 the failed open handle.
2479 Description: Set cat log bitmap only after create log success.
2480 Details : in some rare cases, the cat log bitmap is set too early. and it
2481 should be set only after create log success.
2485 Description: Accessing a block bevice can re-enable I/O when Lustre is
2486 tearing down a device.
2487 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2492 Description: service threads may hog cpus when there are a lot of requests
2493 Details : Insert cond_resched to give other threads a chance to use some CPU
2498 Description: the cat log was not initialized in recovery
2499 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2500 unlink log on mds will not be initialized until mds post
2501 recovery. And also in mds post recovery, the unlink log will
2502 initialization will be done asynchronausly, so there will be race
2503 between add unlink log and unlink log initialization.
2507 Description: brw_stats were being printed incorrectly
2508 Details : brw_stats were being printed as log2 but all of them were not
2509 recorded as log2. Also remove some code duplication arising from
2510 filter_tally_{read,write}.
2514 Frequency : rare, only in recovery.
2515 Description: ASSERTION(req->rq_type != LI_POISON) failed
2516 Details : imp_lock should be held while iterating over imp_sending_list for
2517 prevent destroy request after get timeout in ptlrpc_queue_wait.
2521 Description: replay-single.sh test 52 fails
2522 Details : A lock's skiplist need to be cleanup when it being unlinked
2523 from its resource list.
2528 Description: Short directio read returns full requested size rather than
2530 Details : Direct I/O operations should return actual amount of bytes
2531 transferred rather than requested size.
2533 Severity : enhancement
2535 Description: metadata RPC reduction (e.g. for rm performance)
2536 Details : decrease the amount of synchronous RPC between clients and servers
2537 by canceling conflicing lock before the operation on the client
2538 and packing thier handles into the main operation RPC to server.
2540 Severity : enhancement
2542 Description: add #ifdef HAVE_KERNEL_CONFIG_H
2543 Details : kernels from 2.6.19 not need include linux/config.h, but add
2544 include linux/autoconf.h in commpiler command line.
2546 Severity : enhancement
2548 Description: patchless client support for 2.6.22 kernel
2549 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
2550 removed. In this case we need drop using os depended interface to
2551 kmem_cache and use cfs_mem_cache API.
2556 Description: fix mal-formatted messages
2557 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
2561 Frequency : always in liblustre
2562 Description: wrong IS_ERR implementation in liblustre.h
2563 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
2568 Description: Correct condition for output debug message.
2569 Details : inode i_nlink equal zero is not enough for output message about
2570 disk corruption, i_ctime and i_mode should be also checked.
2574 Frequency : always in patchless client
2575 Description: add configure check for truncate_complete_page
2576 Details : improve checks for exported symbols. This allow run check without
2577 sources, but with Module.symvers shipped with kernel distribution.
2578 add check for truncate_complete_page used by patchless client.
2582 Description: sanity.sh test_77h fails with "test_77h file compare failed"
2583 Details : test_77h uses a file which was messed by other test case.
2587 Description: Not Check whether lov_tgts is NULL in some lov functions
2588 Details : Checking whether lov_tgts is NULL in some functions.
2592 Description: replace obdo_alloc() with OBDO_ALLOC macro
2593 Details : nothing special is done in obdo_alloc() function, and for
2594 debugging purpose, it needs to be replaced with macros.
2598 Description: bad return value and errno from fcntl call
2599 Details : In liblustre API, errno should be a negative value if error
2604 Description: ptlrpc_check_set() LBUG
2605 Details : In case of positive reply from server and failed client bulk
2606 callback after bulk transfer shouldn't LBUG, but process this
2607 request as erroneous.
2609 Severity : enhancement
2611 Description: VFS operations stats tool.
2612 Details : Tool which collects stats by tracking value written in pid,
2613 ppid, gid and uses llstat to generate output to plot graph using
2615 Updated lustre/utils/Makefile.am
2616 Added lustre/utils/ltrack_stats.c
2618 Severity : enhancement
2620 Description: 2.6.18 server support (lustre 1.6.1)
2621 Details : Support for 2.6.18 kernels on the server side.
2626 Description: ASSERTION(imp->imp_conn_current) failed
2627 Details : an assertion failure is hit if a client node boots and attempts to
2628 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
2631 Frequency : only for i686
2633 Description: 1.4.11 RC1 build fails for RHEL 4, i686
2634 Details : Fixed config variable for build.
2639 Description: Updated patchess for new RHEL4 kernel
2640 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
2641 Updated patch jbd-stats-2.6.9.patch
2642 Updated patch qsnet-rhel4-2.6.patch
2643 Updated patch quota-deadlock-on-pagelock-core.patch
2644 Updated patch vfs_intent-2.6-rhel4.patch
2645 Updated patch vfs_races-2.6-rhel4.patch
2646 Updated series file 2.6-rhel4-titech.series
2647 Updated series file 2.6-rhel4.series
2648 Updated kernel config files
2653 Description: lquota slave complains LBUG when reconnecting with mds
2655 Details : quota slave depends qctxt->lqc_import to send its quota request.
2656 This pointer will be invalid if mds did failover or broke its
2657 connect to osts, which leads to LBUG.
2660 Frequency : when qunit size is too small(less than 20M)
2662 Description: write is stopped by improper -EDQUOT
2663 Details : If the master is busy and qunit size is small enough(let's say 1M),
2664 the slave can not get quota from master on time, which will lead
2665 slave to trigger a -EQUOTA to client.
2670 Description: Deadlock during metadata tests
2671 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
2672 called with the per-dentry lock held.
2675 Frequency : SLES9 only
2677 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
2678 Details : set CONFIG_NR_CPUS to 128 instead of 8.
2680 Severity : enhancement
2682 Description: remove fs_prep_san_write operation and related patches
2683 Details : remove the ext3-san-jdike patches which are no longer useful.
2688 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
2689 Details : a disk corruption can cause the mballoc code to assert on a
2690 double free or other extent corruptions. Handle these with
2691 ext3_error() instead of assertions.
2694 Frequency : only with mballoc3 code and deep extent trees
2696 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
2697 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
2698 can cause the filesystem to be remounted read-only.
2703 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
2704 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
2705 few external users of this ioctl number is lfsck's call to
2706 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
2707 Add the old ioctl number to the handler so both old and new
2712 Description: parallel lock callbacks
2713 Details : Instead of sending blocking and completion callbacks as separated
2714 requests, adding them to a set and sending in parallel.
2718 Description: Disable most debugging by default
2719 Details : To improve performance, disable most logging (for debug purposes)
2720 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
2721 default, and HA includes fewer messages.
2725 Frequency : only run on patchless client.
2726 Description: use do_facet on sanity.sh for test handling recoverables errors
2727 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
2731 Description: hash tables for lists of nids, connections and uuids
2732 Details : Hash tables noticeably help when a lot of clients connect to a
2733 server, to faster identify duplicate connections or reconnects,
2734 also to faster find export to evict in manual eviction case.
2738 Description: Sometimes, when the server evict a client, and the client will
2739 not be evicted as soon as possible.
2740 Details : In enqueue req, the error was returned by intent, instead of
2741 rq_status which make ptlrpc layer not detect this error, and
2742 does not evict the client. So enqueue error should be returned
2746 Frequency : only at startup
2748 Description: Delay client connections to MDT until fist MDT->OST connect
2749 Details : If a client tried to create a new file before the MDT had
2750 connected to any OSTs, the create would return EIO. Now
2751 the client will simply block until the MDT connects to the
2752 first OST and the create can succeed.
2755 Frequency : only for SLES9
2757 Description: Routinely utilize latest Quadrics drivers in CFS releases
2758 Details : Update patch qsnet-suse-2.6.patch.
2761 Frequency : only for sles10
2763 Description: Update patches for SLES 10 SP1 kernel.
2764 Details : Update the patch vfs_intent-2.6-sles10.patch.
2769 Description: Routinely utilize latest Quadrics drivers in CFS releases
2770 Details : Update patch qsnet-rhel4-2.6.patch.
2773 Frequency : at statup only
2775 Description: mds_lov_synchronize race leads to various problems
2776 Details : simultaneous MDT->OST connections at startup can cause the
2777 sync to abort, leaving the OSC in a bad state.
2782 Description: Using wrong pointer in osc_brw_prep_request
2783 Details : Access to array[-1] can produce panic if kernel compiled with
2784 CONFIG_PAGE_ALLOC enabled
2786 Severity : enhancement
2788 Description: Async OSC create to avoid the blocking unnecessarily.
2789 Details : If a OST has no remain object, system will block on the creating
2790 when need to create a new object on this OST. Now, ways use
2791 pre-created objects when available, instead of blocking on an
2792 empty osc while others are not empty. If we must block, we block
2793 for the shortest possible period of time.
2797 Frequency : only in recovery
2798 Description: Mark OST as early accessible if his start SYNC.
2799 Details : osc_precreate return flag early accessible if oscc marked as
2800 OSCC_FLAG_SYNC_IN_PROGRESS.
2805 Description: Sometimes precreate code can triger create object on wrong ost
2806 Details : Wrong protected or not not restored variables aftre precreate loop
2807 can produce creation object on wrong ost.
2810 Frequency : oss recovery
2812 Description: llog_commit_thread cleanup should sync with llog_commit_thread
2814 Details : llog_commit_thread_count should be synced between llog_commit
2815 start and cleanup, so new llog_commit thread should not be started
2816 when llog_commit threads being stopped to avoid accessing some
2819 Severity : enhancement
2821 Description: Add printing inode info into message about error in writepage.
2823 --------------------------------------------------------------------------------
2825 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2830 Frequency : on some architectures
2832 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2833 Details : Uninitialized flags sometimes cause configuration commands to
2837 Frequency : patchless clients only
2839 Description: missing __iget() symbol export
2840 Details : The __iget() symbol export is missing. To avoid the need for
2841 this on patchless clients the deathrow inode reaper is turned
2842 off, and we depend on the VM to clean up old inodes. This
2843 dependency was during via the fix for bug 12181.
2848 Description: sanity.sh fail: test_52b
2849 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
2850 incorrect inode's flags to client.
2852 --------------------------------------------------------------------------------
2854 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2856 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2857 INTEROPERATE with older versions automatically. In many cases a
2858 special upgrade step is needed. Please read the
2859 user documentation before upgrading any part of a 1.4.x system.
2860 * WARNING: Lustre configuration and startup changes are required with
2861 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2863 * Support for kernels:
2864 2.4.21-47.0.1.EL (RHEL 3),
2865 2.6.5-7.283 (SLES 9),
2866 2.6.9-42.0.10.EL (RHEL 4),
2867 2.6.12.6 vanilla (kernel.org),
2868 2.6.16.27-0.9 (SLES10)
2869 * Client support for unpatched kernels:
2870 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
2871 2.6.16 - 2.6.19 vanilla (kernel.org),
2872 2.6.9-42.0.8EL (RHEL 4)
2873 * Recommended e2fsprogs version: 1.39.cfs6
2874 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2877 Severity : enhancement
2879 Description: Async OSC create to avoid the blocking unnecessarily.
2880 Details : If a OST has no remain object, system will block on the creating
2881 when need to create a new object on this OST. Now, ways use
2882 pre-created objects when available, instead of blocking on an
2883 empty osc while others are not empty. If we must block, we block
2884 for the shortest possible period of time.
2886 Severity : enhancement
2888 Description: MountConf
2889 Details : Lustre configuration is now managed via mkfs and mount
2890 commands instead of lmc and lconf. New obd types (MGS, MGC)
2891 are added for dynamic configuration management. See
2892 https://mail.clusterfs.com/wikis/lustre/MountConf for
2895 Severity : enhancement
2897 Description: dynamic OST addition
2898 Details : OSTs can now be added to a live filesystem
2900 Severity : enhancement
2902 Description: startup order invariance
2903 Details : MDTs and OSTs can be started in any order. Clients only
2904 require the MDT to complete startup.
2906 Severity : enhancement
2908 Description: parallel, asynchronous orphan cleanup
2909 Details : orphan cleanup is now performed in separate threads for each
2910 OST, allowing parallel non-blocking operation.
2912 Severity : enhancement
2914 Description: optimized stripe assignment
2915 Details : stripe assignments are now made based on ost space available,
2916 ost previous usage, and OSS previous usage, in order to try
2917 to optimize storage space and networking resources.
2919 Severity : enhancement
2921 Description: Permanently set tunables
2922 Details : All writable /proc/fs/lustre tunables can now be permanently
2923 set on a per-server basis, at mkfs time or on a live
2926 Severity : enhancement
2928 Description: Lustre message v2
2929 Details : Add lustre message format v2.
2931 Severity : enhancement
2933 Description: client OST exclusion list
2934 Details : Clients can be started with a list of OSTs that should be
2935 declared "inactive" for known non-responsive OSTs.
2939 Description: ENOENT returned for valid filehandle during dbench.
2940 Details : Check if a directory has children when invalidating dentries
2941 associated with an inode during lock cancellation. This fixes
2942 an incorrect ENOENT sometimes seen for valid filehandles during
2943 testing with dbench.
2946 Frequency : SFS test only (otherwise harmless)
2948 Description: SPEC SFS validation failure on NFS v2 over lustre.
2949 Details : Changes the blocksize for regular files to be 2x RPC size,
2950 and not depend on stripe size.
2952 Severity : enhancement
2954 Description: fine-grained SMP locking inside DLM
2955 Details : Improve DLM performance on SMP systems by removing the single
2956 per-namespace lock and replace it with per-resource locks.
2958 Severity : enhancement
2960 Description: don't hold multiple extent locks at one time
2961 Details : To avoid client eviction during large writes, locks are not
2962 held on multiple stripes at one time or for very large writes.
2963 Otherwise, clients can block waiting for a lock on a failed OST
2964 while holding locks on other OSTs and be evicted.
2966 Severity : enhancement
2968 Description: Multiple MD RPCs in flight.
2969 Details : Further unserialise some read-only MDT RPCs - learn about intents.
2970 To avoid overly-overloading MDT, introduce a limit on number of
2971 MDT RPCs in flight for a single client and add /proc controls
2972 to adjust this limit.
2974 Severity : enhancement
2976 Description: client read/write statistics
2977 Details : Add client read/write call usage stats for performance
2978 analysis of user processes.
2979 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2980 file access. extents_stats shows chunk size distribution.
2981 extents_stats_per_process show chunk size distribution per
2984 Severity : enhancement
2986 Description: per-client statistics on server
2987 Details : Add ldlm and operations statistics for each client in
2988 /proc/fs/lustre/mds|obdfilter/*/exports/
2990 Severity : enhancement
2992 Description: improved MDT statistics
2993 Details : Add detailed MDT operations statistics in
2994 /proc/fs/lustre/mds/*/stats
2996 Severity : enhancement
2998 Description: VFS operations stats
2999 Details : Add client VFS call stats, trackable by pid, ppid, or gid
3000 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
3005 Description: Fix client-side osc byte counters
3006 Details : The osc read/write byte counters in
3007 /proc/fs/lustre/osc/*/stats are now working
3010 Frequency : always as root on SLES
3012 Description: Failure of copying files with lustre special EAs.
3013 Details : Client side always return success for setxattr call for lustre
3014 special xattr (currently only "trusted.lov").
3019 Description: Refcount LNET uuids
3020 Details : The global LNET uuid list grew linearly with every startup;
3021 refcount repeated list entries instead of always adding to
3024 Severity : enhancement
3026 Description: Dynamic service threads
3027 Details : Within a small range, start extra service threads
3028 automatically when the request queue builds up.
3031 Frequency : mixed-endian client/server environments
3033 Description: mixed-endian crashes
3034 Details : The new msg_v2 system had some failures in mixed-endian
3037 Severity : enhancement
3039 Description: Easy OST removal
3040 Details : OSTs can be permanently deactivated with e.g. 'lctl
3041 conf_param lustre-OST0001.osc.active=0'
3043 Severity : enhancement
3045 Description: MGS proc entries
3046 Details : Added basic proc entries for the MGS showing what filesystems
3049 Severity : enhancement
3051 Description: provide MGS failover
3052 Details : Added config lock reacquisition after MGS server failover.
3054 Severity : enhancement
3056 Description: add Linux 2.4 support
3057 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3061 Description: a large application tries to do I/O to the same resource and dies
3062 in the middle of it.
3063 Details : Check the req->rq_arrival time after the call to
3064 ost_brw_lock_get(), but before we do anything about
3065 processing it & sending the BULK transfer request. This
3066 should help move old stale pending locks off the queue as
3067 quickly as obd_timeout.
3070 Frequency : when an incorrect nid is specified during startup
3072 Description: ptlrpc connect to non-existant node causes kernel crash
3073 Details : LNET can't be re-entered from an event callback, which
3074 happened when we expire a message after the export has been
3075 cleaned up. Instead, hand the zombie cleanup off to another
3078 Severity : enhancement
3080 Description: plain/inodebits lock performance improvement
3081 Details : Grouping plain/inodebits in granted list by their request modes
3082 and bits policy, thus improving the performance of search through
3086 Frequency : only if OST filesystem is corrupted
3088 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3089 Details : for a short time RPCs with bulk IO are in the replay list,
3090 but replay of bulk IOs is unimplemented. If the OST filesystem
3091 is corrupted due to disk cache incoherency and then replay is
3092 started it is possible to trip an assertion. Avoid putting
3093 committed RPCs into the replay list at all to avoid this issue.
3096 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3098 Description: System hang on startup
3099 Details : This bug allowed the liblustre (e.g. catamount) client to
3100 return to the app before handling all startup RPCs. This
3101 could leave the node unresponsive to lustre network traffic
3102 and manifested as a server ptllnd timeout.
3104 Severity : enhancement
3106 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3107 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3108 boolean to control whether to print peer debug info when a
3109 client's RPC times out.
3112 Frequency : only for kernels with patches from Lustre below 1.4.3
3114 Description: Remove old rdonly API
3115 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3118 Frequency : only for devices with external journals
3120 Description: Set external device read-only also
3121 Details : During a commanded failover stop, we set the disk device
3122 read-only while the server shuts down. We now also set any
3123 external journal device read-only at the same time.
3126 Frequency : when upgrading from 1.4 while trying to change parameters
3128 Description: The wrong (new) MDC name was used when setting parameters for
3129 upgraded MDT's. Also allows changing of OSC (and MDC)
3130 parameters if --writeconf is specified at tunefs upgrade time.
3133 Frequency : when setting specific ost indicies
3135 Description: QOS code breaks on skipped indicies
3136 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3137 created with --index need not be sequential.
3139 Severity : enhancement
3141 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3142 Details : The uninit_groups feature works in conjunction with the kernel
3143 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3144 up the pass1 processing of e2fsck. This is a read-only feature
3145 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3146 mount filesystems that have had this feature enabled.
3148 Severity : enhancement
3150 Description: Improve multi-block allocation algorithm to avoid fragmentation
3151 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3152 allocation locality and avoid filesystem fragmentation.
3154 ------------------------------------------------------------------------------
3156 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3158 * Support for kernels:
3159 2.4.21-47.0.1.EL (RHEL 3)
3160 2.6.5-7.283 (SLES 9)
3161 2.6.9-42.0.10.EL (RHEL 4)
3162 2.6.12.6 vanilla (kernel.org)
3163 2.6.16.27-0.9 (SLES 10)
3164 * Recommended e2fsprogs version: 1.39.cfs5
3166 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3170 Frequency : occasional, depends on client load and configuration
3171 Bugzilla : 12181, 12203
3172 Description: data loss for recently-modified files
3174 Details : In some cases it is possible that recently written or created
3175 files may not be written to disk in a timely manner (this should
3176 normally be within 30s unless client IO load is very high).
3177 The problem appears as zero-length files or files that are a
3178 multiple of 1MB in size after a client crash or client eviction
3179 that are missing data at the end of the file.
3181 This problem is more likely to be hit on clients where files are
3182 repeatedly created and unlinked in the same directory, clients
3183 have a large amount of RAM, have many CPUs, the filesystem has
3184 many OSTs, the clients are rebooted frequently, and/or the files
3185 are not accessed by other nodes after being written.
3187 The presence of the problem can be detected by looking at
3188 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3189 smaller than the second (nr_unused) then dirty files will not
3190 be flushed automatically to disk. "sync; sleep 10" should be
3191 run several times on the node before unmounting it to update
3192 Lustre (this is also safe to run on nodes without this problem).
3194 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3195 that can cause this same problem, so customers using that kernel
3196 also need to update the kernel in addition to Lustre. In order
3197 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3199 It is normal that files written just before a client crash (less
3200 than 30s) may not yet have been flushed to disk, even for local
3204 Frequency : frequent on thin XT3 nodes
3206 Description: UUID collision on thin XT3 Linux nodes
3207 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3208 randomly, since we relied on an insufficiently-seeded PRNG.
3213 Description: OSS hangs after "All ost request buffers busy"
3214 Details : A deadlock between quota and journal operations caused OSS
3215 hangs after printing "All ost request buffers busy."
3218 Frequency : always on liblustre builds
3220 Description: Cleanup compiler warnings on liblustre
3223 Frequency : always on liblustre builds on XT3
3225 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3229 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3230 > 32000 subdirectories
3231 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3232 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3233 subdirectory count crosses 32000. This will aid e2fsck to
3234 correctly handle more than 32000 subdirectories.
3237 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3239 Description: System hang on startup
3240 Details : This bug allowed the liblustre (e.g. catamount) client to
3241 return to the app before handling all startup RPCs. This
3242 could leave the node unresponsive to lustre network traffic
3243 and manifested as a server ptllnd timeout.
3245 Severity : enhancement
3247 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3248 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3249 boolean to control whether to print peer debug info when a
3250 client's RPC times out.
3255 Description: make O_SYNC working on 2.6 kernels
3256 Details : 2.6 kernels use different method for mark pages for write,
3257 so need add a code to lustre for O_SYNC work.
3262 Description: Failure to close file and release space on NFS
3263 Details : Put inode details into lock acquired in ll_intent_file_open.
3264 Use mdc_intent_lock in ll_intent_open to properly
3265 detect all kind of errors unhandled by mdc_enqueue.
3270 Description: proc file read during shutdown sometimes raced obd removal,
3272 Details : Add lock to prevent obd access after proc file removal.
3275 Frequency : Only for files larger than 4GB on 32-bit clients.
3277 Description: improperly doing page alignment of locks
3278 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3279 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3284 Description: RPCs being resent when they shouldn't be
3285 Details : Some RPCs that should not be resent are being resent. This
3286 can cause inconsistencies in the RPC state machine. Do not
3287 resend such requests.
3290 Frequency : rare, only with NFS export
3292 Description: Crash on NFS re-export node
3293 Details : under very unusual load conditions an assertion is hit in
3294 ll_intent_file_open()
3297 Frequency : only if OST filesystem is corrupted
3299 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3300 Details : for a short time RPCs with bulk IO are in the replay list,
3301 but replay of bulk IOs is unimplemented. If the OST filesystem
3302 is corrupted due to disk cache incoherency and then replay is
3303 started it is possible to trip an assertion. Avoid putting
3304 committed RPCs into the replay list at all to avoid this issue.
3309 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3310 Details : Large single O_DIRECT read and write calls can fail to allocate
3311 a sufficiently large buffer to process the request. In case of
3312 allocation failure the allocation is retried with a smaller
3313 buffer and broken into smaller requests.
3315 Severity : enhancement
3317 Description: Add -o localflock option to simulate old noflock behaviour.
3318 Details : This will achieve local-only flock/fcntl locks coherentness.
3323 Description: versioning check is incomplete
3324 Details : Checking the version difference of client vs. server, report
3325 error if the gap is too big.
3330 Description: add support PG_writeback bit
3331 Details : add support for PG_writeback bit for Lustre, for more carefull
3332 work with page cache in 2.6 kernel. This also fix some deadlocks
3333 and remove hack for work O_SYNC with 2.6 kernel.
3335 Severity : enhancement
3337 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3338 Details : The uninit_groups feature works in conjunction with the kernel
3339 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3340 up the pass1 processing of e2fsck. This is a read-only feature
3341 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3342 mount filesystems that have had this feature enabled.
3344 Severity : enhancement
3346 Description: Improve multi-block allocation algorithm to avoid fragmentation
3347 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3348 allocation locality and avoid filesystem fragmentation.
3350 ------------------------------------------------------------------------------
3352 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3354 * Support for kernels:
3355 2.6.9-42.0.3.EL (RHEL 4)
3356 2.6.5-7.276 (SLES 9)
3357 2.4.21-47.0.1.EL (RHEL 3)
3358 2.6.12.6 vanilla (kernel.org)
3359 2.6.16.21-0.8 (SLES10)
3360 * Recommended e2fsprogs version: 1.39.cfs2-0
3362 * The backwards-compatible /proc/sys/portals symlink has been removed
3363 in this release. Before upgrading, please ensure that you change
3364 any configuration scripts or /etc/sysctl.conf files that access
3365 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3366 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3367 in advance of the upgrade on any system running Lustre 1.4.6 or
3368 newer, since /proc/sys/lnet was added in that version.
3369 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3373 Frequency : only when quota is used
3375 Description: avoid scanning export list for quota master
3376 Details : Change the algorithms to avoid scanning export list in order
3377 to improve the efficiency.
3380 Frequency : MDS failover only, very rarely
3382 Description: "went back in time" messages on mds failover
3383 Details : The greatest transno may be lost when the current operation
3384 finishes with an error (transno==0) and the client's last_rcvd
3385 record is over-written. Save the greatest transno in the
3386 mds_last_transno for this case.
3389 Frequency : always for specific kernels and striping counts
3391 Description: client may get "Matching packet too big" without ACL support
3392 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3393 when trying to access files in certain configurations. The
3394 clients should in fact be denied when mounting because they do
3395 not understand ACLs.
3398 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3400 Description: many clients connecting with IO in progress causes connect timeouts
3401 Details : Avoid synchronous journal commits to avoid delays caused by many
3402 clients connecting/disconnecting when bulk IO is in progress.
3403 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3404 OST_IO_PORTAL to avoid delays behind potentially many pending
3408 Frequency : occasionally with multiple writers to a single file
3410 Description: shared writes to file may result in wrong size reported by stat()
3411 Details : Allow growing of kms when extent lock is cancelled
3414 Frequency : always with random mmap IO to multi-striped file
3416 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3417 Details : Only if the hole is at the end of OST object so that kms is too
3418 small. Fix is to increase kms accordingly in ll_nopage.
3421 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3423 Description: writes to a missing object would leak memory on the OST
3424 Details : If there is an inconsistency between the MDS and OST filesystems,
3425 such that the MDS references an object that doesn't exist, writes
3426 to that object will leak memory due to incorrect cleanup in the
3427 error handling path, eventually running out of memory on the OST.
3432 Description: Creating too long symlink causes lustre errors
3433 Details : Check symlink and name lengths before sending requests to MDS.
3436 Frequency : only if flock is enabled (not on by default)
3438 Description: posix locks not released on fd closure on 2.6.9+
3439 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3440 kernels, this caused such locks not to be released on fd close and
3441 then assertions on fs unmount about still used locks.
3444 Frequency : MDS failover only, very rarely
3446 Description: clients may get ASSERTION(granted_lock != NULL)
3447 Details : When request was taking a long time, and a client was resending
3448 a getattr by name lock request. The were multiple lock requests
3449 with the same client lock handle and
3450 mds_getattr_name->fixup_handle_for_resent_request found one of the
3451 lock handles but later failed with ASSERTION(granted_lock != NULL).
3456 Description: handle->h_buffer_credits > 0, assertion failure
3457 Details : h_buffer_credits is zero after truncate, causing assertion
3458 failure. This patch extends the transaction or creates a new
3462 Frequency : NFS re-export or patchless client
3463 Bugzilla : 11179, 10796
3464 Description: Crash on NFS re-export node (__d_move)
3465 Details : We do not want to hash the dentry if we don't have a lock.
3466 But if this dentry is later used in d_move, we'd hit uninitialised
3467 list head d_hash, so we just do this to init d_hash field but
3468 leave dentry unhashed.
3471 Frequency : NFS re-export or patchless client
3473 Description: NFS exports has problem with symbolic link
3474 Details : lustre client didn't properly install dentry when re-exported
3475 to NFS or running patchless client.
3478 Frequency : NFS re-export or patchless client
3480 Description: Various nfs/patchless fixes.
3481 Details : fixes reuse disconected alias for lookup process - this fixes
3482 warning "find_exported_dentry: npd != pd",
3483 fix permission error with open files at nfs.
3484 fix apply umask when do revalidate.
3487 Frequency : occasional
3489 Description: Crash on NFS re-export node
3490 Details : calling clear_page() on the wrong pointer triggered oops in
3491 generic_mapping_read().
3494 Frequency : rarely, using O_DIRECT IO
3496 Description: unaligned directio crashes client with LASSERT
3497 Details : check for unaligned buffers before trying any requests.
3500 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3502 Description: stale data returned from RAID cache
3503 Details : If only a small amount of IO is done to the RAID device before
3504 reading it again it is possible to get stale data from the RAID
3505 cache instead of reading it from disk.
3508 Frequency : always for sles10 kernel
3510 Description: sles10 support
3511 Details : ll_follow_link: compile fixes and using of nd_set_link
3512 under newer kernels.
3515 Frequency : depends on arch, kernel and compiler version, always on sles10
3518 Description: recursive or deep enough symlinks cause stack overflow
3519 Details : getting rid of large stack-allocated variable in
3523 Frequency : depends on hardware
3525 Description: lustre write performance loss in the SLES10 kernel
3526 Details : the performance loss is caused by using of write barriers in the
3527 ext3 code. The SLES10 kernel turns barrier support on by
3528 default. The fix is to undo that change for ldiskfs.
3530 ------------------------------------------------------------------------------
3532 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3534 * Support for kernels:
3535 2.6.9-42.0.3EL (RHEL 4)
3536 2.6.5-7.276 (SLES 9)
3537 2.4.21-47.0.1.EL (RHEL 3)
3538 2.6.12.6 vanilla (kernel.org)
3542 Frequency : quota enabled and large files being deleted
3544 Description: releasing more than 4GB of quota at once hangs OST
3545 Details : If a user deletes more than 4GB of files on a single OST it
3546 will cause the OST to spin in an infinite loop. Release
3547 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3552 Description: statfs data retrieved from /proc may be stale or zero
3553 Details : When reading per-device statfs data from /proc, in the
3554 {kbytes,files}_{total,free,avail} files, it may appear
3555 as zero or be out of date.
3558 Frequency : systems with MD RAID1 external journal devices
3560 Description: lconf's call to blkid is confused by RAID1 journal devices
3561 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3562 of returning all block devices that match the journal UUID.
3565 Frequency : always, for aggregate stripe size over 4GB
3567 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3568 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3569 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3572 Frequency : always if "lfs find" used on a local file/directory
3574 Description: "lfs find" segfaults if used on a local file/directory
3575 Details : The case where a directory component was not specified wasn't
3576 handled correctly. Handle this properly.
3579 Frequency : always on ppc64
3581 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3582 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3583 perform endianness conversion.
3586 Frequency : rarely (truncate to non-zero file size after write under load)
3587 Bugzilla : 10730, 10687
3588 Description: Files padded with zeros to next 4K multiple
3589 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3590 it is possible that files that are truncated to a non-zero size
3591 immediately after being written are filled with zero bytes beyond
3592 the truncated size. No file data is lost.
3594 Severity : enhancement
3596 Description: Allow recovery/failover for liblustre clients.
3597 Details : liblustre clients were unaware of failover configurations until
3600 Severity : enhancement
3602 Description: user file locks should fail when not mounting with flock option
3603 Details : Set up an error-returning stub in ll_file_operations.lock field
3604 to prevent incorrect behaviour when client is mounted without
3605 flock option. Also, set up properly f_op->flock field for
3609 Frequency : always on ia64
3611 Description: "lfs df" loops on printing out MDS statfs information
3612 Details : The obd_ioctl_data was not initialized and in some systems
3613 this caused a failure during the ioctl that did not return
3614 an error. Initialize the struct and return an error on failure.
3617 Frequency : SLES 9 only
3619 Description: Error of copying files with lustre special EAs as root
3620 Details : Client side always return success for setxattr call for lustre
3621 special xattr (currently only "trusted.lov").
3624 Frequency : rarely on clusters with both ia64+i386 clients
3626 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3627 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3628 clients are doing concurrent writes to the same file it is
3629 possible that the ia64 clients may LASSERT because the OST
3630 extent locks are not PAGE_SIZE aligned. Ensure that grown
3631 locks are always aligned on the request boundary.
3634 Frequency : specific use, occasional
3636 Description: Overwriting in use executable truncates on-disk binary image
3637 Details : If one node attempts to overwrite an executable in use by
3638 another node, we now correctly return ETXTBSY instead of
3639 truncating the file.
3641 Severity : enhancement
3643 Description: Async OSC create to avoid the blocking unnecessarily.
3644 Details : If a OST has no remain object, system will block on the creating
3645 when need to create a new object on this OST. Now, ways use
3646 pre-created objects when available, instead of blocking on an
3647 empty osc while others are not empty. If we must block, we block
3648 for the shortest possible period of time.
3653 Description: chmod on Lustre root is propagated to other clients
3654 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3655 invalid by the follow_mount time.
3658 Frequency : liblustre clients only
3660 Description: Race in 'instant cancel' lock handling could lead to such locks
3661 never to be granted in case of SMP MDS
3662 Details : Do not destroy not yet granted but cbpending locks in
3666 Frequency : replay/resend of open
3668 Description: non null lock assetion failure in mds_intent_policy
3669 Details : Trying to replay/resend lockless open requests resulted in
3670 mds_open() returning 0 with no lock. Now it sets a flag if
3671 a lock is going to be returned.
3673 Severity : enhancement
3675 Description: Checksum enhancements
3676 Details : New checksum enhancements allow for resending RPCs that failed
3679 Severity : enhancement
3681 Description: Tunables on number of dirty pages in cacche
3682 Details : Allow to set limit on number of dirty pages cached.
3687 Description: client crash on unmount - lock still has references
3688 Details : In some error handling cases it was possible to leak a lock
3689 reference on a client while accessing a file. This was not
3690 harmful to the client during operation, but would cause the
3691 client to crash when the filesystem is unmounted.
3694 Frequency : specific case, rare
3696 Description: ETXTBSY on mds though file not in use by client
3697 Details : ETXTBSY is no longer incorrectly returned when attempting to
3698 chmod or chown a directory that the user previously tried to
3699 execute or a currently-executing binary.
3702 Frequency : extremely rare except on liblustre-based clients
3704 Description: Lustre space not freed when files are deleted
3705 Details : Clean up open-unlinked files after client eviction. Previously
3706 the unlink was skipped and the files remained as orphans.
3711 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3712 Details : In some cases it was possible to send a blocking callback to
3713 a client doing a glimpse, even though that client didn't get
3714 a lock granted. When the glimpse lock is cancelled on the OST
3715 the freed lock is left on the waiting list and corrupted the list.
3718 Frequency : all core dumps
3720 Description: Broke core dumps to lustre
3721 Details : Negative dentry may be unhashed if parent does not have UPDATE
3722 lock, but some callers, e.g. do_coredump, expect dentry to be
3723 hashed after successful create, hash it in ll_create_it.
3725 ------------------------------------------------------------------------------
3727 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3729 * Support for kernels:
3730 2.6.9-42.0.2.EL (RHEL 4)
3731 2.6.5-7.276 (SLES 9)
3732 2.4.21-40.EL (RHEL 3)
3733 2.6.12.6 vanilla (kernel.org)
3737 Frequency : always on RHEL 3
3739 Description: Number of open files grows over time
3740 Details : The number of open files grows over time, whether or not
3741 Lustre is started. This was due to a filp leak introduced
3742 by one of our kernel patches.
3744 ------------------------------------------------------------------------------
3746 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
3748 * Support for kernels:
3749 2.6.9-42.EL (RHEL 4)
3750 2.6.5-7.267 (SLES 9)
3751 2.4.21-40.EL (RHEL 3)
3752 2.6.12.6 vanilla (kernel.org)
3757 Bugzilla : 5719, 9635, 9792, 9684
3758 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3759 Details : If a server is under heavy load and cannot reply to new
3760 connection requests before the client resends the (re)connect,
3761 the connection handling code can behave badly if two service
3762 threads are concurrently handing separate (re)connections from
3763 the same client. Add better locking to the connection handling
3764 code, and ensure that only a single connection will be processed
3765 for a given client UUID, even if the lock is dropped.
3767 Severity : enhancement
3769 Description: add TCP zero-copy support to kernel
3770 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3771 sends if the hardware supports scatter-gather and checksumming.
3772 This allows socklnd to do client-write and server-read more
3773 efficiently and reduce CPU utilization from skbuf copying.
3776 Frequency : only if NFS exporting from client
3778 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3779 Details : If setting attributes on a file created under NFS that had
3780 never been opened it would be possible to oops the client
3781 if the file had no objects.
3785 Bugzilla : 9326, 10402, 10897
3786 Description: client crash in ptlrpcd_wake() thread when sending async RPC
3787 Details : It is possible that ptlrpcd_wake() dereferences a freed async
3788 RPC. In rare cases the ptlrpcd thread alread processed the RPC
3789 before ptlrpcd_wake() was called and the request was freed.
3792 Frequency : always for liblustre
3794 Description: liblustre client does MDS+OSTs setattr RPC for each write
3795 Details : When doing a write from a liblustre client, the client
3796 incorrectly issued an RPC to the MDS and each OST the file was
3797 striped over in order to update the timestamps. When writing
3798 with small chunks and many clients this could overwhelm the MDS
3799 with RPCs. In all cases it would slow down the write because
3800 these RPCs are unnecessary.
3802 Severity : enhancement
3804 Description: allow number of MDS service threads to be changed at module load
3805 Details : It is now possible to change the number of MDS service threads
3806 running. Adding "options mds mds_num_threads={N}" to the MDS's
3807 /etc/modprobe.conf will set the number of threads for the next
3808 time Lustre is restarted (assuming the "mds" module is also
3809 reloaded at that time). The default number of threads will
3810 stay the same, 32 for most systems.
3815 Description: OST crash if filesystem is unformatted or corrupt
3816 Details : If an OST is started on a device that has never been formatted
3817 or if the filesystem is corrupt and cannot even mount then the
3818 error handling cleanup routines would dereference a NULL pointer.
3823 Description: NULL pointer deref in llap_from_page.
3824 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3825 incorrect) mapping if the page was truncated/reclaimed while it was
3826 searched for. Check for this condition and skip such pages when
3827 doing readahead. Introduce extra check to llap_from_page() to
3828 verify page->mapping->host is non-NULL (so page is not anonymous).
3831 Frequency : Sometimes when using sys_sendfile
3833 Description: "page not covered by a lock" warnings from ll_readpage
3834 Details : sendfile called ll_readpage without right page locks present.
3835 Now we introduced ll_file_sendfile that does necessary locking
3836 around call to generic_file_sendfile() much like we do in
3840 Frequency : with certain MDS communication failures at client mount time
3842 Description: NULL pointer deref after failed client mount
3843 Details : a client connection request may delayed by the network layer
3844 and not be sent until after the PTLRPC layer has timed out the
3845 request. If the client fails the mount immediately it will try
3846 to clean up before the network times out the request. Add a
3847 reference from the request import to the obd device and delay
3848 the cleanup until the network drops the request.
3851 Frequency : occasionally during client (re)connect
3853 Description: assertion failure during client (re)connect
3854 Details : processing a client connection request may be delayed by the
3855 client or server longer than the client connect timeout. This
3856 causes the client to resend the connection request. If the
3857 original connection request is replied in this interval, the
3858 client may trip an assertion failure in ptlrpc_connect_interpret()
3859 which thought it would be the only running connect process.
3862 Frequency : only with obd_echo servers and clients that are rebooted
3864 Description: kernel BUG accessing uninitialized data structure
3865 Details : When running an obd_echo server it did not start the ping_evictor
3866 thread, and when a client was evicted an uninitialized data
3867 structure was accessed. Start the ping_evictor in the RPC
3868 service startup instead of the OBD startup.
3870 Severity : enhancement
3871 Bugzilla : 10193 (patchless)
3872 Description: Remove dependency on various unexported kernel interfaces.
3873 Details : No longer need reparent_to_init, exit_mm, exit_files,
3874 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3877 Frequency : rare (only users of deprecated and unsupported LDAP config)
3879 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3880 Details : LDAP apparently contains 'lustreName' attributes instead of
3881 'name'. A simple remapping of the name is sufficient.
3884 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3886 Description: waiting_locks_callback trips kernel BUG if client is evicted
3887 Details : Running with the dump_on_timeout debug flag turned on makes
3888 it possible that the waiting_locks_callback() can try to dump
3889 the Lustre kernel debug logs from an interrupt handler. Defer
3890 this log dumping to the expired_lock_main() thread.
3892 Severity : enhancement
3894 Description: Support NFS exporting on 2.6 kernels.
3895 Details : Implement non-rawops metadata methods for NFS server to use without
3896 changing NFS server code.
3899 Frequency : very rare (synthetic metadata workload only)
3901 Description: two racing renames might cause an MDS thread to deadlock
3902 Details : Running the "racer" program may cause one MDS thread to rename
3903 a file from being the source of a rename to being the target of
3904 a rename at exactly the same time that another thread is doing
3905 so, and the second thread has already enqueued these locks after
3906 doing a lookup of the target and is trying to relock them in
3907 order. Ensure that we don't try to re-lock the same resource.
3910 Frequency : only very large systems with liblustre clients
3912 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3913 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3914 all OSTs in parallel. This allows the eviction of stale liblustre
3915 clients to proceed much faster than if they were done in series,
3916 and also offers similar improvements for other set_info RPCs.
3921 Description: excessive CPU usage during initial read phase on client
3922 Details : During the initial read phase on a client, it would agressively
3923 retry readahead on the file, consuming too much CPU and impacting
3924 performance (since 1.4.5.8). Improve the readahead algorithm
3925 to avoid this, and also improve some other common cases (read
3926 of small files in particular, where "small" is files smaller than
3927 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3932 Description: MDS crash when receiving packet with unknown intent.
3933 Details : Do not LBUG in unknown intent case, just return -EFAULT
3935 Severity : enhancement
3936 Bugzilla : 9293, 9385
3937 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3938 Details : Do not serialize getattr (non-intent version) and statfs.
3941 Frequency : occasional, when OST network is overloaded/intermittent
3943 Description: client evicted by OST after bulk IO timeout
3944 Details : If a client sends a bulk IO request (read or write) the OST
3945 may evict the client if it is unresposive to its data GET/PUT
3946 request. This is incorrect if the network is overloaded (takes
3947 too long to transfer the RPC data) or dropped the OST GET/PUT
3948 request. There is no need to evict the client at all, since
3949 the pinger and/or lock callbacks will handle this, and the
3950 client can restart the bulk request.
3953 Frequency : Always when mmapping file with no objects
3955 Description: client crashes when mmapping file with no objects
3956 Details : Check that we actually have objects in a file before doing any
3957 operations on objects in ll_vm_open, ll_vm_close and
3963 Description: Request leak when working with deleted CWD
3964 Details : Introduce advanced request refcount tracking for requests
3965 referenced from lustre intent.
3967 Severity : Enhancement
3969 Description: Cache open file handles on client.
3970 Details : MDS now will return special lock along with openhandle, if
3971 requested and client is allowed to hold openhandle, even if unused,
3972 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3973 closing files for every read/write openration.
3975 Severity : Enhancement
3977 Description: Cache open negative dentries on client when possible.
3978 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3979 negative dentries on lock revocation.
3984 Description: Remounting a client read-only wasn't possible with a zconf mount
3985 Details : It wasn't possible to remount a client read-only with llmount.
3987 Severity : enhancement
3988 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3989 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3990 MPICH is included to add Lustre-specific ADIO interfaces.
3991 This is based closely on the UFS ADIO layer and only differs
3992 in file creation, in order to allow the OST striping to be set.
3993 This is user-contributed code and not supported by CFS.
3998 Description: extended inode attributes (immutable, append-only) work improperly
3999 when 2.4 and 2.6 kernels are used on client/server or vice versa
4000 Details : Introduce kernel-independent values for these flags.
4002 Severity : enhancement
4005 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4006 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4007 granularity. Now, Fractional MB granularity is be supported,
4008 this is very useful for low memory system.
4010 Severity : enhancement
4012 Description: Getattr by fid
4013 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4014 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4015 have locks and avoids getattr rpc requests for it on every stat.
4018 Frequency : Always, for filesystems larger than 2TB
4020 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4021 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4022 filesystems larger than 2TB to be mounted with mballoc enabled.
4025 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4027 Description: filesystem corruption for non-standard kernels and very large OSTs
4028 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4029 filesystems larger than 2TB is mounted then the kernel will
4030 silently corrupt the start of the filesystem. CONFIG_LBD is
4031 enabled for all CFS-supported kernels, but the possibility of
4032 this happening with a modified kernel config exists.
4034 Severity : enhancement
4036 Description: add client O_DIRECT support for 2.6 kernels
4037 Details : It is now possible to do O_DIRECT reads and writes to files
4038 in the Lustre client mountpoint on 2.6 kernel clients.
4040 Severity : enhancement
4042 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4043 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4044 parallel, not waiting for response from every OST before sending
4045 a rpc to the next OST.
4050 Description: setattr vs write race when updating file timestamps
4051 Details : Client processes that update a file timestamp into the past
4052 right after writing to the file (e.g. tar) it is possible that
4053 the updated file modification time can be reset to the current
4054 time due to a race between processing the setattr and write RPC.
4056 Severity : enhancement
4058 Description: Bring 'lfs find' closer in line with regular Linux find.
4059 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4060 -print0 options and obtains all the needed info through the lustre
4063 Severity : enhancement
4065 Description: support up to 1024 configured devices on one node
4066 Details : change obd_dev array from statically allocated to dynamically
4067 allocated structs as they are first used to reduce memory usage
4072 Description: Flush dirty partially truncated pages during truncate
4073 Details : Immediatelly flush partially truncated pages in filter_setattr,
4074 this way we completely avoid having any pages in page cache on OST
4075 and can retire ugly workarounds during writes to flush such pages.
4080 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4081 Details : protect inode from truncation within vfs_unlink() context
4082 just take a reference before calling vfs_unlink() and release it
4083 when parent's i_sem is free.
4086 Frequency : always, if extents are used on OSTs
4088 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4089 Details : OSTs using the extents format would not zero the high 16 bits of
4090 the index physical block number. This is not a problem for any
4091 OST filesystems smaller than 16TB, and no kernels support ext3
4092 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4093 new/modified files) and can be fixed for existing filesystems
4094 with e2fsprogs-1.39-cfs1.
4099 Description: import connection selection may be incorrect if timer wraps
4100 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4101 import connections to be ignored if the 32-bit jiffies counter
4102 wraps. Use a 64-bit jiffies counter.
4105 Frequency : during server recovery
4107 Description: crash after server is denying duplicate export
4108 Details : If clients are resending connect requests to the server, the
4109 server refuses to allow a client to connect multiple times.
4110 Fixed a bug in the handling of this case.
4113 Frequency : very large clusters immediately after boot
4115 Description: LNET request buffers exhausted under heavy short-term load
4116 Details : If a large number of client requests are generated on a service
4117 that has previously never seen so many requests it is possible
4118 that the request buffer growth cannot keep up with the spike in
4119 demand. Instead of dropping incoming requests, they are held in
4120 the LND until the RPC service can accept more requests.
4123 Frequency : Sometimes during replay
4125 Description: Assertion failure in ll_local_open after replay.
4126 Details : If replay happened on an open request reply before we were able
4127 to set replay handler, reply will become not swabbed tripping the
4128 assertion in ll_local_open. Now we set the handler right after
4129 recognising of open request
4132 Frequency : very rare
4134 Description: kernel reports "badness in vsnprintf"
4135 Details : Reading from the "recovery_status" /proc file in small chunks
4136 may cause a negative length in lprocfs_obd_rd_recovery_status()
4137 call to vsnprintf() (which is otherwise harmless). Exit early
4138 if there is no more space in the output buffer.
4140 Severity : enhancement
4142 Description: clear OBD RPC statistics by writing to them
4143 Details : It is now possible to clear the OBD RPC statistics by writing
4144 to the "stats" file.
4149 Description: Client mtime is not the same on different clients after utimes
4150 Details : In some cases, the client was using the utimes() syscall on
4151 a file cached on another node. The clients now validate the
4152 ctime from the MDS + OSTs to determine which one is right.
4157 Description: Inability to activate failout mode
4158 Details : lconf script incorrectly assumed that in python string's numeric
4159 value is used in comparisons.
4162 Frequency : always with multiple stripes per file
4164 Description: Inefficient object allocation for mutli-stripe files
4165 Details : When selecting which OSTs to stripe files over, for files with
4166 a stripe count that divides evenly into the number of OSTs,
4167 the MDS is always picking the same starting OST for each file.
4168 Return the OST selection heuristic to the original design.
4173 Description: mount failures may take full timeout to return an error
4174 Details : Under some heavy load conditions it is possible that a
4175 failed mount can wait for the full obd_timeout interval,
4176 possibly several minutes, before reporting an error.
4177 Instead return an error as soon as the status is known.
4179 ------------------------------------------------------------------------------
4181 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
4183 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4184 INTEROPERATE with older versions automatically. Please read the
4185 user documentation before upgrading any part of a live system.
4186 * WARNING: Lustre networking configuration changes are required with
4187 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4190 * Support for kernels:
4191 2.6.9-22.0.2.EL (RHEL 4)
4192 2.6.5-7.244 (SLES 9)
4193 2.6.12.6 vanilla (kernel.org)
4196 Severity : enhancement
4197 Bugzilla : 7981/8208
4198 Description: Introduced Lustre Networking (LNET)
4199 Details : LNET is new networking infrastructure for Lustre, it includes
4200 a reorganized network configuration mode (see the user
4201 documentation for full details) as well as support for routing
4202 between different network fabrics. Lustre Networking Devices
4203 (LNDs) for the supported network fabrics have also been
4204 created for this new infrastructure.
4206 Severity : enhancement
4207 Description: Introduced Access control lists
4208 Details : clients can set ACLs on files and directories in order to have
4209 more fine-grained permissions than the standard Unix UGO+RWX.
4210 The MDS must be started with the "-o acl" mount option.
4212 Severity : enhancement
4213 Description: Introduced filesystem quotas
4214 Details : Administrators may now establish per-user quotas on the
4217 Severity : enhancement
4219 Description: Configuration change for the XT3
4220 The PTLLND is now used to run Lustre over Portals on the XT3
4221 The configure option(s) --with-cray-portals are no longer used.
4222 Rather --with-portals=<path-to-portals-includes> is used to
4223 enable building on the XT3. In addition to enable XT3 specific
4224 features the option --enable-cray-xt3 must be used.
4229 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4230 Details : A race between allocating a new llcd and re-getting the llcd_lock
4231 allowed another thread to grab newly-allocated llcd.
4233 Severity : enhancement
4235 Description: 2.6 OST async journal commit and locking fix to improve performance
4236 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4237 2.6 kernels are now async as they already were in 2.4 kernels so
4238 that they can commit concurrently with the network bulk transfer.
4239 For block-allocated files the filter allocation semaphore is held
4240 to avoid filesystem fragmentation during allocation. BKL lock
4241 removed for 2.6 xattr operations where it is no longer needed.
4246 Description: lconf incorrectly determined whether two IP networks could talk
4247 Details : In some more complicated routing and multiple-network
4248 configurations, lconf will avoid trying to make a network
4249 connection to a disjoint part of the IP space. It was doing the
4250 math incorrectly for one set of cases.
4255 Description: Fix for potential infinite loop processing records in an llog.
4256 Details : If an llog record is corrupted/zeroed, it is possible to loop
4257 forever in llog_process(). Validate the llog record length
4258 and skip the remainder of the block on error.
4261 Frequency : occasional (liblustre only)
4263 Description: liblustre could not open files whose last component is a symlink
4264 Details : sysio_path_walk() would incorrectly pass the open intent to
4265 intermediate path components.
4268 Frequency : rare (liblustre only with non-standard tuning)
4269 Bugzilla : 7201 (7350)
4270 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4271 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4272 after it has been released, only do this for referenced locks
4274 Severity : enhancement
4276 Description: specifying an (invalid) directory default stripe_size of -1
4277 would reset the directory default striping
4278 Details : stripe_size -1 was used internally to signal directory stripe
4279 removal, now use "all default" to signal dir stripe removal
4280 as a directory striping of "all default" is not useful
4283 Frequency : common for large clusters running liblustre clients
4285 Description: doing an ls when liblustre clients are running is slow
4286 Details : sending a glimpse AST to a liblustre client waits for every AST
4287 to time out, as liblustre clients will not respond. Since they
4288 cannot cache data we refresh the OST lock LVB from disk instead.
4290 Severity : enhancement
4292 Description: doing an ls at the same time as file IO can be slow
4293 Details : enqueue and other "small" requests can be blocked behind many
4294 large IO requests. Create a new OST IO portal for non-IO
4295 requests so they can be processed faster.
4298 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4300 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4301 created file with mode 000
4302 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4303 allow an ftruncate() as a non-root user to a file with mode 000.
4304 Lustre now respects this flag to disable mode checks when
4305 truncating a file owned by the user
4308 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4311 Description: Revoking locks from clients that went dead or catatonic might take
4313 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4314 cancellation of such locks instant on servers without waiting for
4315 any reply from clients. Clients drops these locks when cancel
4316 notification from server is received without replying.
4319 Frequency : liblustre-only, when liblustre client dies or becomes busy
4321 Description: Doing ls on Linux clients can take a long time with active
4323 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4324 granting such locks to it in the first place if possible. Locks
4325 are taken by proxy on the OST during the read or write and
4326 dropped immediately afterward. Add connect flags handling, do
4327 not grant locks to liblustre clients for glimpse ASTs.
4329 Severity : enhancement
4331 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4332 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4333 sequential reads and behaves badly with applications doing random
4334 reads. Improve it by reading ahead at least read region, and
4335 avoiding excessive large RPC for small reads.
4337 Severity : enhancement
4339 Description: Creating more than 1000 files for a single job may cause a load
4340 imbalance on the OSTs if there are also a large number of OSTs.
4341 Details : qos_prep_create() uses an OST index reseed value that is an
4342 even multiple of the number of available OSTs so that if the
4343 reseed happens in the middle of the object allocation it will
4344 still utilize the OSTs as uniformly as possible.
4349 Description: OST or MDS may oops in ping_evictor_main()
4350 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4351 but doesn't restart at beginning of obd_exports_timed list
4354 Severity : enhancement
4356 Description: improve by-nid export eviction on the MDS and OST
4357 Details : allow multiple exports with the same NID to be evicted at one
4358 time without re-searching the exports list.
4361 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4363 Description: MDS may oops in groups_free()
4364 Details : in rare race conditions a newly allocated group_info struct is
4365 freed again, and this can be NULL. The 2.4 compatibility code
4366 for groups_free() checked for a NULL pointer, but 2.6 did not.
4369 Frequency : common for liblustre clients doing little filesystem IO
4370 Bugzilla : 9352, 7313
4371 Description: server may evict liblustre clients accessing contended locks
4372 Details : if a client is granted a lock or receives a completion AST
4373 with a blocking AST already set it would not reply to the AST
4374 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4377 Frequency : lfs setstripe, only systems with more than 160 OSTs
4379 Description: unable to set striping with a starting offset beyond OST 160
4380 Details : llapi_create_file() incorrectly limited the starting stripe
4381 index to the maximum single-file stripe count.
4384 Frequency : LDAP users only
4386 Description: lconf did not handle in-kernel recovery with LDAP properly
4387 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4389 Severity : enhancement
4391 Description: bind OST threads to NUMA nodes to improve performance
4392 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4393 node and do their allocations there to localize memory access
4395 Severity : enhancement
4397 Description: llmount can determine client NID directly from Myrinet (GM)
4398 Details : the client NID code from gmnalnid was moved directly into
4399 llmount, removing the need to use this or specifying the
4400 client NID explicitly when mounting GM clients with zeroconf
4403 Frequency : if client is started with down MDS
4405 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4406 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4407 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4408 ptlrpc_import_delayed_req() because we didn't check if the
4409 request was interrupted, and we also didn't break out of the
4410 event loop if there was no timeout
4415 Description: data loss during non-page-aligned writes to a single file from
4416 both multiple nodes and multiple threads on one node at same time
4417 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4418 inconsistency led to false short-reads, that were cached and later
4419 used by ->prepare_write() to fill in partially written page,
4420 leading to data loss.
4423 Frequency : always, if lconf --abort_recovery used
4425 Description: lconf --abort_recovery fails with 'Operation not supported'
4426 Details : lconf was attempting to abort recovery on the MDT device and not
4429 Severity : enhancement
4431 Description: remove cleanup logs
4432 Details : replace lconf-generated cleanup logs with lustre internal
4433 cleanup routines. Eliminates the need for client-cleanup and
4436 Severity : enhancement
4438 Description: add support for EAs (user and system) on lustre filesystems
4439 Details : it is now possible to store extended attributes in the Lustre
4440 client filesystem, and with the user_xattr mount option it
4441 is possible to allow users to store EAs on their files also
4443 Severity : enhancement
4445 Description: Add possibility (config option) to show minimal available OST free
4447 Details : When compiled with --enable-mindf configure option, statfs(2)
4448 (and so, df) will return least minimal free space available from
4449 all OSTs as amount of free space on FS, instead of summary of
4450 free spaces of all OSTs.
4452 Severity : enhancement
4454 Description: do not expand extent locks acquired on OST-side
4455 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4456 by server: they are not cached anyway.
4459 Frequency : when mmap is used/binaries executed from Lustre
4461 Description: Unmmap pages before throwing them away from read cache.
4462 Details : llap_shrink cache now attempts to unmap pages before discarding
4463 them (if unmapping failed - do not discard). SLES9 kernel has
4464 extra checks that trigger if this unmapping is not done first.
4469 Description: lconf didn't resolve symlinks before checking to see whether a
4470 given mountpoint was already in use
4473 Frequency : when migrating failover services
4474 Bugzilla : 6395, 9514
4475 Description: When migrating a subset of services from a node (e.g. failback
4476 from a failover service node) the remaining services would
4477 time out and evict clients.
4478 Details : lconf --force (implied by --failover) sets the global obd_timeout
4479 to 5 seconds in order to quickly disconnect, but this caused
4480 other RPCs to time out too quickly. Do not change the global
4481 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4483 Severity : enhancement
4484 Frequency : if MDS is started with down OST
4485 Bugzilla : 9439,5706
4486 Description: Allow startup/shutdown of an MDS without depending on the
4487 availability of the OSTs.
4488 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4489 Add appropriate locking and lov-osc refcounts for safe
4490 cleaning. Add osc abort_inflight calls in case the
4491 synchronize never started.
4494 Frequency : occasional (Cray XT3 only)
4496 Description: root not authorized to access files in CRAY_PORTALS environment
4497 Details : The client process capabilities were not honoured on the MDS in
4498 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4499 been accessed by an authorized user then root was able to access
4500 the file on the local client also. The root user capabilities
4501 are now allowed on the MDS, as this environment has secure UID.
4504 Frequency : occasional
4506 Description: ldiskfs "too long searching" message happens too often
4507 Details : A debugging message (otherwise harmless) prints too often on
4508 the OST console. This has been reduced to only happen when
4509 there are fragmentation problems on the filesystem.
4514 Description: Division by zero in statfs when all OSCs are inactive
4515 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4516 lov_statfs divides by value returned by lov_get_stripecnt().
4520 Bugzilla : 9489, 3273
4521 Description: First write from each client to each OST was only 4kB in size,
4522 to initialize client writeback cache, which caused sub-optimal
4523 RPCs and poor layout on disk for the first writen file.
4524 Details : Clients now request an initial cache grant at (re)connect time
4525 and so that they can start streaming writes to the cache right
4526 away and always do full-sized RPCs if there is enough data.
4527 If the OST is rebooted the client also re-establishes its grant
4528 so that client cached writes will be honoured under the grant.
4533 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4534 Details : Now I/O RPCs go to different portal number and (presumably) fast
4535 lock requests (and glimses) and other RPCs get their own service
4536 threads pool that should be able to service those RPCs
4539 Severity : enhancement
4541 Description: Ability to exchange lustre version between client and servers and
4542 issue warnings at client side if client is too old. Also for
4543 liblustre clients there is ability to refuse connection of too old
4545 Details : New 'version' field is added to connect data structure that is
4546 filled with version info. That info is later checked by server and
4550 Frequency : rare, liblustre only.
4551 Bugzilla : 9296, 9581
4552 Description: Two simultaneous writes from liblustre at offset within same page
4553 might proceed at the same time overwriting eachother with stale
4555 Details : I/O lock withing llu_file_prwv was released too early, before data
4556 actually was hitting the wire. Extended lock-holding time until
4557 server acknowledges receiving data.
4560 Frequency : extremely rare. Never observed in practice.
4562 Description: avoid generating lustre_handle cookie of 0.
4563 Details : class_handle_hash() generates handle cookies by incrementing
4564 global counter, and can hit 0 occasionaly (this is unlikely, but
4565 not impossible, because initial value of cookie counter is
4566 selected randonly). Value of 0 is used as a sentinel meaning
4567 "unassigned handle" --- avoid it. Also coalesce two critical
4568 sections in this function into one.
4570 Severity : enhancement
4572 Description: allow liblustre clients to delegate truncate locking to OST
4573 Details : To avoid overhead of locking, liblustre client instructs OST to
4574 take extent lock in ost_punch() on client's behalf. New connection
4575 flag is added to handle backward compatibility.
4577 Severity : enhancement
4578 Bugzilla : 4928, 7341, 9758
4579 Description: allow number of OST service threads to be specified
4580 Details : a module parameter allows the number of OST service threads
4581 to be specified via "options ost ost_num_threads={N}" in the
4582 OSS's /etc/modules.conf or /etc/modprobe.conf.
4586 Bugzilla : 6146, 9635, 9895
4587 Description: servers crash with bad pointer in target_handle_connect()
4588 Details : In rare cases when a client is reconnecting it was possible that
4589 the connection request was the last reference for that export.
4590 We would temporarily drop the export reference and get a new
4591 one, but this may have been the last reference and the export
4592 was just destroyed. Get new reference before dropping old one.
4594 Severity : enhancement
4595 Frequency : if client is started with failover MDS
4597 Description: Allow multiple MDS hostnames in the mount command
4598 Details : Try to read the configuration from all specified MDS
4599 hostnames during a client mount in case the "primary"
4602 Severity : enhancement
4604 Description: Stop sending data to evicted clients as soon as possible.
4605 Details : Check if the client we are about to send or are sending data to
4606 was evicted already. (Check is done every second of waiting,
4607 for which l_wait_event interface was extended to allow checking
4608 of exit condition at specified intervals).
4611 Frequency : rare, normally only when NFS exporting is done from client
4613 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4615 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4618 Frequency : always, if srand() is called before liblustre initialization
4620 Description: Liblustre uses system PRNG disturbing its usage by user application
4621 Details : Introduce internal to lustre fast and high-quality PRNG for
4622 lustre usage and make liblustre and some other places in generic
4623 lustre code to use it.
4625 Severity : enhancement
4626 Bugzilla : 9477, 9557, 9870
4627 Description: Verify that the MDS configuration logs are updated when xml is
4628 Details : Check if the .xml configuration logs are newer than the config
4629 logs stored on the MDS and report an error if this is the case.
4630 Request --write-conf, or allow starting with --old_conf.
4632 Severity : enhancement
4634 Description: Handle symlinks in the path when checking if Lustre is mounted.
4635 Details : Resolve intermediate symlinks when checking if a client has
4636 mounted a filesystem to avoid duplicate client mounts.
4641 Description: lconf can hit an error exception but still return success.
4642 Details : The lconf command catches the Command error exception at the top
4643 level script context and will exit with the associated exit
4644 status, but doesn't ensure that this exit status is non-zero.
4649 Description: failure of ptlrpc thread startup can cause oops
4650 Details : Starting a ptlrpc service thread can fail if there are a large
4651 number of threads or the server memory is very fragmented.
4652 Handle this without oopsing.
4655 Frequency : always, only if liblustre and non-default acceptor port was used
4657 Description: liblustre cannot connect to servers with non-default acceptor port
4658 Details : tcpnal_set_default_params() was not called and was therefore
4659 ignoring the environment varaible TCPNAL_PORT, as well as other
4660 TCPNAL_ environment variables
4665 Description: two objects could be created on the same OST for a single file
4666 Details : If an OST is down, in some cases it was possible to create two
4667 objects on a single OST for a single file. No problems other
4668 than potential performance impact and spurious error messages.
4672 Bugzilla : 5681, 9562
4673 Description: Client may oops in ll_unhash_aliases
4674 Details : Client dcache may become inconsistent in race condition.
4675 In some cases "getcwd" can fail if the current directory is
4681 Description: Inode refcounting problems in NFS export code
4682 Details : link_raw functions used to call d_instantiate without obtaining
4683 extra inode reference first.
4687 Bugzilla : 9942, 9903
4688 Description: Referencing freed requests leading to crash, memleaks with NFS.
4689 Details : We used to require that call to ll_revalidate_it was always
4690 followed by ll_lookup_it. Also with revalidate_special() it is
4691 possible to call ll_revalidate_it() twice for the same dentry
4692 even if first occurence returned success. This fix changes semantic
4693 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4694 reference on a request referred from the intent.
4695 ll_intent_release() then releases such a request.
4698 Frequency : rare, normally benchmark loads only
4700 Description: unlinked inodes were kept in memory on the client
4701 Details : If a client is repeatedly creating and unlinking files it
4702 can accumulate a lot of stale inodes in the inode slab cache.
4703 If there is no other client load running this can cause the
4704 client node to run out of memory. Instead flush old inodes
4705 from client cache that have the same inode number as a new inode.
4708 Frequency : SLES9 2.6.5 kernel and long filenames only
4709 Bugzilla : 9969, 10379
4710 Description: utime reports stale NFS file handle
4711 Details : SLES9 uses out-of-dentry names in some cases, which confused
4712 the lustre dentry revalidation. Change it to always use the
4716 Frequency : rare, unless heavy write-truncate concurrency is continuous
4717 Bugzilla : 4180, 6984, 7171, 9963, 9331
4718 Description: OST becomes very slow and/or deadlocked during object unlink
4719 Details : filter_destroy() was holding onto the parent directory lock
4720 while truncating+unlinking objects. For very large objects this
4721 may block other threads for a long time and slow overall OST
4722 responsiveness. It may also be possible to get a lock ordering
4723 deadlock in this case, or run out of journal credits because of
4724 the combined truncate+unlink. Solution is to do object truncate
4725 first in one transaction without parent lock, and then do the
4726 final unlink in a new transaction with the parent lock. This
4727 reduces the lock hold time dramatically.
4730 Frequency : rare, 2.4 kernels only
4732 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4733 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4734 before unmounting their filesystem in order to avoid deadlock.
4735 The kernel_locked() function in 2.4 kernels only checks whether
4736 the kernel lock is held, not whether it is this process that is
4737 holding it as 2.6 kernels do.
4742 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4743 Details : The client ptlrpc code may be trying to reconnect to a down
4744 server before a previous connection attempt has timed out.
4745 Increase the reconnect interval to be longer than the connection
4746 timeout interval to avoid sending duplicate connections to
4750 Frequency : echo_client brw_test command
4752 Description: fix echo_client to work with OST preallocated code
4753 Details : OST preallocation code (5137) didn't take echo_client IO path
4754 into account: echo_client calls filter methods outside of any
4755 OST thread and, hence, there is no per-thread preallocated
4756 pages and buffers to use. Solution: hijack pga pages for IO. As
4757 a byproduct, this avoids unnecessary data copying.
4761 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4762 Description: Client can oops in mdc_commit_close() after open replay
4763 Details : It was possible for the MDS to return an open request with no
4764 transaction number in mds_finish_transno() if the client was
4765 evicted, but without actually returning an error. Clients
4766 would later try to replay that open and may trip an assertion
4767 Simplify the client close codepath, and always return an error
4768 from the MDS in case the open is not successful.
4771 Frequency : rare, 2.6 OSTs only
4773 Description: OST may deadlock under high load on fragmented files
4774 Details : If there was a heavy load and highly-fragmented OST filesystems
4775 it was possible to have all the OST threads deadlock waiting on
4776 allocation of biovecs, because the biovecs were not released
4777 until the entire RPC IO was completed. Instead, release biovecs
4778 as soon as they are complete to ensure forward IO progress.
4780 Severity : enhancement
4782 Description: Support for specifying external journal device at mount
4783 Details : If an OST or MDS device is formatted with an external journal
4784 device, this device major/minor is stored in the ext3 superblock
4785 and may not be valid for failover. Allow detecting and
4786 specifying the external journal at mount time.
4791 Description: Mounting an MDS with pending unlinked files may cause oops
4792 Details : target_finish_recovery() calls mds_postrecov() which returned
4793 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4794 considers this an error and immediately begins cleaning up the
4795 lov, just after starting the mds_lov process
4797 Severity : enhancement
4799 Description: Implement 'lfs df' to report actual free space on per-OST basis
4800 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4801 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4802 usage of objects; '-h' to report in human readable format.
4804 ------------------------------------------------------------------------------
4806 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
4813 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4814 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4815 at startup during orphan unlinking. Instead, skip buddy update
4816 before setup, as it will be regenerated after recovery is complete.
4821 Description: If an OST is inactive, its locks might reference stale inodes.
4822 Details : lov_change_cbdata() must iterate over all namespaces, even if
4823 they are inactive to clear inode references from the lock.
4825 Severity : enhancement
4826 Frequency : occasional, if non-standard max_dirty_mb used
4828 Description: Client will block write RPCs if not enough grant
4829 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4830 then the client will block writes while waiting for another RPC
4831 to complete instead of consuming its dirty limit. With change
4832 we get improved performance when max_dirty_mb is small.
4834 Severity : enhancement
4835 Bugzilla : 3389, 6253
4836 Description: Add support for supplementary groups on the MDS.
4837 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4838 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4839 lookups for user supplementary groups into a cache.
4843 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4844 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4845 for liblustre O_EXCL is enforced by MDS.
4849 Description: readdir never returns NULL in liblustre.
4850 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4851 dentry in struct dirent.
4855 Frequency : liblustre only, depends on application IO pattern
4856 Description: liblustre clients evicted if not contacting servers
4857 Details : Don't put liblustre clients into the ping_evictor list, so
4858 they will not be evicted by the pinger ever.
4860 Severity : enhancement
4862 Description: Add ability to evict clients by NID from MDS.
4863 Details : By echoing "nid:$NID" string into
4864 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4865 $NID would be instantly evicted from this MDS and from all active
4866 OSTs connected to it.
4870 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4871 Details : lookup_it_finish() used to query file size from OSTs that was not
4876 Description: service threads change working directory to that of init
4877 Details : Starting lustre service threads may pin the working directory
4878 of the parent thread, making that filesystem busy. Threads
4879 now change to the working directory of init to avoid this.
4883 Frequency : during shutdown only
4884 Description: shutdown with a failed MDS or OST can cause unmount to hang
4885 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4890 Frequency : 2.6 only
4891 Description: chmod/chown may include an extra supplementary group
4892 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4893 supplementary group and if none is specified this is used.
4896 Bugzilla : 5479 (6816)
4898 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4899 Details : If lookup is in progress on a file that is unlinked we might try
4900 to revalidate the inode and fail in revalidate after lookup is
4901 complete and ll_file_open() enqueues the open again but
4902 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4905 Frequency : always, if lconf --abort_recovery used
4907 Description: lconf --abort_recovery fails with 'Operation not supported'
4908 Details : lconf was attempting to abort recovery on the MDT device and not
4911 ------------------------------------------------------------------------------
4913 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4918 Frequency : rare (only unsupported configurations with a node running as an
4920 Bugzilla : 6514, 5137
4921 Description: Mounting a Lustre file system on a node running as an OST could
4923 Details : OSTs now preallocates memory needed to write out data at
4924 startup, instead of when needed, to avoid having to
4925 allocate memory in possibly low memory situations.
4926 Specifically, if the file system is mounted on on OST,
4927 memory pressure could force it to try to write out data,
4928 which it needed to allocate memory to do. Due to the low
4929 memory, it would be unable to do so and the node would
4930 become unresponsive.
4932 Severity : enhancement
4934 Description: Addition of lconf --service command line option
4935 Details : lconf now accepts a '--service <arg>' option, which is
4936 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4938 Severity : enhancement
4940 Description: Failover mode is now the default for OSTs.
4941 Details : By default, OSTs will now run in failover mode. To return to
4942 the old behaviour, add '--failout' to the lmc line for OSTs.
4944 Severity : enhancement
4946 Description: Health checks are now provided for MDS and OSTs
4947 Details : Additional detailed health check information on MSD and OSTs
4948 is now provided through the procfs health_check value.
4951 Frequency : occasional, depends on IO load
4953 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4954 after numerous create/delete cycles
4955 Details : The ext3 inode allocation policy would not allocate new inodes
4956 very well on the OSTs because there are no new directories
4957 being created. Instead we look for groups with free space if
4958 the parent directories are nearly full.
4963 Description: Network or server problems during mount may cause partially
4964 mounted clients instead of returning an error.
4965 Details : The config llog parsing code may overwrite the error return
4966 code during mount error handling, returning success instead
4972 Description: MDS can fail to allocate large reply buffers
4973 Details : After long uptimes the MDS can fail to allocate large reply
4974 buffers (e.g. zconf client mount config records) due to memory
4975 fragmentation or consumption by the buffer cache. Preallocate
4976 some large reply buffers so that these replies can be sent even
4977 under memory pressure.
4981 Frequency : rare (liblustre)
4982 Description: fsx running with liblustre complained that using truncate() to
4983 extend the file doesn't work. This patch corrects that issue.
4984 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4985 ATTR_SIZE and lsm use in llu_setattr_raw.
4989 Frequency : rare, only 2.6 kernels
4990 Description: Unusual file access patterns on the MDS may result in inode
4991 data being lost in very rare circumstances.
4992 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4993 code in the 2.6 kernel caused the inode and/or EA data not to
4994 be read from disk, causing single-file corruption.
4998 Frequency : rare, only 2.6 filesystems using extents
4999 Description: Heavy concurrent write and delete load may cause data corruption.
5000 Details : It was possible under high-load situations to have an extent
5001 metadata block in the block device cache from a just-unlinked
5002 file overwrite a newly-allocated data block. We now unmap any
5003 metadata buffers that alias just-allocated data blocks.
5007 Frequency : filesystems with default stripe_count larger than 77
5008 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5009 Details : lconf specifies an inode size of 4096 bytes when the default
5010 stripe_count is larger than 77. This conflicts with the default
5011 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5012 this case to avoid pinning too much memory for large EAs.
5014 ------------------------------------------------------------------------------
5016 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5021 Frequency : rare (extremely heavy IO load with hundreds of clients)
5023 Description: Client is evicted, gets IO error writing to file
5024 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5025 caused clients to be evicted for AST timeouts. The fixes in
5026 bug 5192 mean we no longer need to have such short AST timeouts
5027 so ldlm_timeout has been increased.
5030 Frequency : occasional during --force or --failover shutdown under load
5031 Bugzilla : 5949, 4834
5032 Description: Server oops/LBUG if stopped with --force or --failover under load
5033 Details : a collection of import/export refcount and cleanup ordering
5034 issues fixed for safer force cleanup
5037 Frequency : only filesystems larger than 120 OSTs
5038 Bugzilla : 5990, 6223
5039 Description: lfs getstripe would oops on a very large filesystem
5040 Details : lov_getconfig used kfree on vmalloc'd memory
5043 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5044 Bugzilla : 6242, 6243
5045 Description: reading from files that had been truncated to a non-zero size
5046 but never opened returned no data
5047 Details : ll_file_read() reads zeros from no-object files to EOF
5052 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5053 Details : There's little harm in aborting MDS/OSS recovery and letting it
5054 try again, so I removed the LASSERT and return an error instead.
5056 Severity : enhancement
5058 Description: New debugging infrastructure for tracking down data corruption
5059 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5060 (b) cover more of the client-side code path, and (c) try to narrow
5061 down where problems occurred
5065 Bugzilla : 3819, 4364, 4397, 6313
5066 Description: Racing close and eviction MDS could cause assertion in mds_close
5067 Details : It was possible to get multiple mfd references during close and
5068 client eviction, leading to one thread referencing a freed mfd.
5070 Severity: : enhancement
5071 Bugzilla : 3262, 6359
5072 Description: Attempts to reconnect to servers are now more aggressive.
5073 Details : This builds on the enhanced upcall-less recovery that was added
5074 in 1.4.2. When trying to reconnect to servers, clients will
5075 now try each server in the failover group every 10 seconds. By
5076 default, clients would previously try one server every 25 seconds.
5081 Description: After recovery, certain operations trigger a failed
5082 assertion on a client.
5083 Details : Failing over an mds, using lconf -d --failover, while a
5084 client was doing a readdir() call would cause the client to
5085 LBUG after recovery completed and the readdir() was resent.
5087 Severity : enhancement
5089 Description: Default groups are now added by lconf
5090 Details : You can now run lconf --group <servicename> without having to
5091 manually add groups with lmc.
5094 Frequency : occasional
5096 Description: Nodes with an elan id of 0 trigger a failed assertion
5099 Frequency : always when accessing e.g. tty/console device nodes
5101 Description: tty and some other devices nodes cannot be used on lustre
5102 Details : file's private_data field is used by device data and lustre
5103 values in there got lost. New field was added to struct file to
5104 store fs-specific private data.
5107 Frequency : when exporting Lustre via NFS
5109 Description: NFSD failed occasionally when looking up a path component
5110 Details : NFSD is looking up ".." which was broken in ext3 directories
5111 that had grown large enough to become hashed.
5114 Frequency : Clusters with multiple interfaces not on the same subnet
5116 Description: Nodes will repeatedly try to reconnect to an interface which it
5117 cannot reach and report an error to the log.
5118 Details : Extra peer list entries will be created by lconf with some peers
5119 unreachable. lconf now validates the peer before adding it.
5122 Frequency : Only if a default stripe is set on the filesystem root.
5124 Description: Setting a default stripe on the filesystem root prevented the
5125 filesystem from being remounted.
5126 Details : The client was sending extra request flags in the root getattr
5127 request and did not allocate a reply buffer for the dir EA.
5130 Frequency : occasional, higher if lots of files are accessed by one client
5131 Bugzilla : 6159, 6097
5132 Description: Client trips assertion regarding lsm mismatch/magic
5133 Details : While revalidating inodes the VFS looks up inodes with ifind()
5134 and in rare cases can find an inode that is being freed.
5135 The ll_test_inode() code will free the lsm during ifind()
5136 when it finds an existing inode and then the VFS later attaches
5137 this free lsm to a new inode.
5141 Bugzilla : 6422, 7030
5142 Description: MDS deadlock between mkdir and client eviction
5143 Details : Creating a new file via mkdir or mknod (starting a transaction
5144 and getting the ns lock) can deadlock with client eviction
5145 (gets ns lock and trying to finish a synchronous transaction).
5148 Frequency : occasional
5149 Description: While starting a server, the fsfilt_ext3 module could not be
5151 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5152 kernels. Previously, lconf would still use the ext3 name
5153 when trying to load modules. Now, it will correctly use
5154 ext3 on 2.4 and ldiskfs on 2.6.
5156 Severity : enhancement
5157 Description: The default stripe count has been changed to 1
5158 Details : The interpretation of the default stripe count (0, to lfs
5159 or lmc) has been changed to mean striping across a single
5160 OST, rather than all available. For general usage we have
5161 found a stripe count of 1 or 2 works best.
5163 Severity : enhancement
5164 Description: Add support for compiling against Cray portals.
5165 Details : Conditional compiling for some areas that are different
5169 Frequency : occasional
5170 Bugzilla : 6409, 6834
5171 Description: Creating files with an explicit stripe count may lead to
5172 a failed assertion on the MDS
5173 Details : If some OSTs are full or unavailable, creating files may
5174 trigger a failed assertion on the MDS. Now, Lustre will
5175 try to use other servers or return an error to the
5179 Frequency : occasional
5181 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5182 caused lock timeout during readahead (since 1.4.2).
5183 Details : Processes doing readahead might match a lock that hasn't been
5184 granted yet if there are overlapping and conflicting lock
5185 requests. The readahead process waits on ungranted lock
5186 (original lock is CBPENDING), while OST waits for that process
5187 to cancel CBPENDING read lock and eventually evicts client.
5189 Severity : enhancement
5191 Description: Initial enabling of flock support for clients
5192 Details : Implements fcntl advisory locking and file status functions.
5193 This feature is provided as an optional mount flag (default
5194 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5195 locking are implemented yet, and those that are are not guaranteed
5196 to be completely correct in production environments.
5197 mount -t lustre -o [flock|noflock] ...
5200 Frequency : occasional
5202 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5203 assertion in the ext3 JBD (journaling) layer.
5204 Details : The b_committed_data struct is protected by the big kernel lock
5205 in 2.4 kernels, serializing journal_commit_transaction() and
5206 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5207 access to this struct. In 2.6 kernels there is finer grained
5208 locking to improve SMP performance of the JBD layer.
5212 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5214 -----------------------------------------------------------------------------
5216 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5218 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5219 versions of Lustre. Please update all servers and clients to
5220 version 1.4.2 or later at the same time. You must also run
5221 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5222 to update the configuration logs.
5224 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5225 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5226 - swab llog records of type '0' so we get proper header size/idx (5861)
5227 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5228 - fix rename of one directory over another leaking an inode (5953)
5229 - avoid SetPageDirty on 2.6 (5981)
5230 - don't re-add just-being-destroyed locks to the waiting list (5653)
5231 - when creating new directories, inherit the parent's custom
5232 striping settings if present parent (3048)
5233 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5234 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5235 - don't hold client locks on temporary worklist from l_lru (5666)
5236 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5237 - automatically evict dead clients (5921)
5238 - Update file size properly in create+truncate+fstat case (6196)
5239 - Do not unhash mountpoint dentries, do not allow removal of
5241 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5242 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5243 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5244 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5245 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5246 - Correctly handle reads of files with no objects (6243)
5247 - lctl recover will also mark a device active if deactivate used (5933)
5249 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5250 - allow --write-conf on an MDS with different nettype than client (5619)
5251 - don't write config llogs to MDS for mounts not from that MDS (5617)
5252 - lconf should create multiple TCP connections from a client (5201)
5253 - init scripts are now turned off by default; run chkconfig --on
5254 lustre and chkconfig --on lustrefs to use them
5255 - upcalls are no longer needed for clients to recover to failover
5257 - add --abort-recovery option to lconf to abort recovery on device
5259 - add support for an arbitrary number of OSTs (3026)
5260 - Quota support protocol changes.
5261 - forward compatibility changes to wire structs (6007)
5262 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5263 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5264 - improved "lustre status" script
5265 - initialize blocksize for non-regular files (6062)
5266 - added --disable-server and --disable-client configure options (5782)
5267 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5268 - Vanilla 2.4.29 support
5269 - increase maximum number of obd devices to 520 (6242)
5270 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5271 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5272 SLES 9 2.6.5 kernels
5273 - limit stripes per file to 160 (the maximum EA size) (6093)
5275 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5278 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5279 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5280 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5281 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5282 - small ext3 extents cleanups and fixes (5733)
5283 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5284 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5285 - avoid needless client->OST connect, fix handle mismatch (5317)
5286 - fix DLM error path that led to out-of-sync client, long delays (5779)
5287 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5288 - fix several locking issues related to i_size (5492,5624,5654,5672)
5289 - don't move pending lock onto export if it is already evicted (5683)
5290 - fix kernel oops when creating .foo in unlinked directory (5548)
5291 - fix deadlock in obdfilter statistics vs. object create (5811)
5292 - use time_{before,after} to avoid timer jiffies wrap (5882)
5293 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5294 - Do not leak request if server was not able to process it (5154)
5295 - If mds_open unable to find parent dir, make that negative lookup(5154)
5296 - don't create new directories with extent-mapping (5909, 5936)
5298 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5299 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5300 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5301 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5302 - support for RHEL4 kernel on i686 (5773)
5303 - provide error messages when incompatible logs are encountered (5898)
5305 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5306 * version 1.4.0.10 (1.4.1 release candidate 1)
5308 - don't keep a lock reference when lock is not granted (4238)
5309 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5310 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5311 - drop import inflight refcount on signal_completed_replay error (5255)
5312 - unlock page after async write error during send (3677)
5313 - handle missing objects in filter_preprw_read properly (5265)
5314 - no transno return for symlink open, don't save no-trasno open (3440)
5315 - don't try to complete elan receive that already failed (4012)
5316 - free RPC server reply state on error (5406)
5317 - clean up thread from ptlrpc_start_thread() on error (5160)
5318 - readahead could read extra page into cache that wasn't ejected (5388)
5319 - prevent races in class_attach/setup/cleanup/detach (5260)
5320 - don't dereference de->d_inode after l_dput of de (5458)
5321 - use "int" for stripe value returned from lock_to_stripe (5544)
5322 - mballoc allocation and error-checking fixes in 2.6 (5504)
5323 - block device patches to fix I/O request sizes in 2.6 (5482)
5324 - look up hostnames for IB nals (5602)
5325 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5326 - don't start multiple acceptors for the same port (5277)
5327 - fix incorrect LASSERT in mds_getattr_name (5635)
5328 - export a proc file for general "ping" checking (5628)
5329 - fix "lfs check" to not block when the MDS is down (5628)
5331 - service request history (4965)
5332 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5333 - create an "evict_client" /proc entry on OSTs, like the MDS has
5334 - fix mount usage message, return errors per mount(8) (5168)
5335 - change grep [] to grep "[]" in tests so they work in more UMLs
5336 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5337 - remove ancient LOV_MAGIC_V0 EA support (5047)
5338 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5339 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5340 - allow readahead allocations to fail when low on memory (5383)
5341 - mmap locking landed again, after considerable improvement (2828)
5342 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5344 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5347 - send OST transaction number in read/write reply to free req (4966)
5348 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5349 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5350 - the watchdog thread now runs as interruptible (5246)
5351 - flock/lockf fixes (but it's still disabled, pending 5135)
5352 - don't use EXT3 constants in llite code (5094)
5353 - memory shortage at startup could cause assertion (5176)
5355 - reorganization of lov code
5356 - single portals codebase
5358 - add extents/mballoc support (5025)
5359 - direct I/O reads in the obdfilter (4048)
5360 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5362 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5365 - send OST transaction number in read/write reply to free req (4966)
5366 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5367 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5368 - the watchdog thread now runs as interruptible (5246)
5369 - handle missing objects in filter_preprw_read properly (5265)
5370 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5371 - drop import inflight refcount on signal_completed_replay error (5255)
5372 - unlock page after async write error during send (3677)
5373 - return original error code on reconstructed replies (3761)
5374 - no transno return for symlink open, don't save no-trasno open (3440)
5376 - add pid to ldlm debugging output (4922)
5377 - bump the watchdog timeouts -- we can't handle 30sec yet
5378 - extra debugging for orphan dentry/inode bug (5259)
5380 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5383 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5384 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5385 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5386 - fix NULL dereference in /proc/sys/portals/routes (4827)
5387 - allow failed mdc_close() operations to be interrupted (4561)
5388 - stop precreate on OST before MDS would time out on it (4778)
5389 - don't send partial-page writes before EOF from client (4410)
5390 - discard client grant for sub-page writes on large-page clients (4520)
5391 - don't free dentries not owned by NFS code, check generation (4806)
5392 - fix lsm leak if mds_create_objects() fails (4801)
5393 - limit debug_daemon file size, always print CERROR messages (4789)
5394 - use transno after validating reply (3892)
5395 - process timed out requests if import state changes (3754)
5396 - update mtime on OST during writes, return in glimpse (4829)
5397 - add mkfsoptions to LDAP (4679)
5398 - use ->max_readahead method instead of zapping global ra (5039)
5399 - don't interrupt __l_wait_event() during strace
5401 - add software watchdogs to catch hung threads quickly (4941)
5402 - make lustrefs init script start after nfs is mounted
5403 - fix CWARN/ERROR duplication (4930)
5404 - return async write errors to application if possible (2248)
5405 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5406 - print NAL number in %x format (4645)
5407 - update barely-supported suse-2.4.21-171 series (4842)
5408 - support for sles 9 %post scripts
5409 - support for building 2.6 kernel-source packages
5410 - support for sles km_* packages
5412 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5415 - ignore -ENOENT errors in osc_destroy (3639)
5416 - notify osc create thread that OSC is being cleaned up (4600)
5417 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5418 - reconstruct ost_handle() like mds_handle() (4657)
5419 - create a new thread to do import eviction to avoid deadlock (3969)
5420 - let lconf resolve symlinked-to devices (4629)
5421 - don't unlink "objects" from directory with default EA (4554)
5422 - hold socknal file ref over connect in case target is down (4394)
5423 - allow more than 32000 subdirectories in a single directory (3244)
5424 - fix blocks count for O_DIRECT writes (3751)
5425 - OST returns ENOSPC from object create when no space left (4539)
5426 - don't send truncate RPC if file size isn't changing (4410)
5427 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5428 - bind to privileged port in socknal and tcpnal (3689)
5430 - rate limit CERROR/CWARN console message to avoid overload (4519)
5431 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5432 - basic mmap support (3918)
5433 - kernel patch series update from b1_4 (4711)
5435 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5438 - avoid crash during MDS cleanup with OST shut down (2775)
5439 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5440 - don't use bad inodes on the MDS (3744)
5441 - dynamic object preallocation to improve recovery speed (4236)
5442 - don't hold spinlock over lock dumping or change debug flags (4401)
5443 - don't zero obd_dev when it is force cleaned (3651)
5444 - print grants to console if they go negative (4431)
5445 - "lctl deactivate" will stop automatic recovery attempts (3406)
5446 - look for existing locks in ldlm_handle_enqueue() (3764)
5447 - don't resolve lock handle twice in recovery avoiding race (4401)
5448 - revalidate should check working dir is a directory (4134)
5450 - don't always mark "slow" obdfilter messages as errors (4418)
5452 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5455 - don't close LustreDB during write_conf until it is done (3860)
5456 - fix typo in lconf for_each_profile (3821)
5457 - allow dumping logs from multiple threads at one time (3820)
5458 - don't allow multiple threads in OSC recovery (3812)
5459 - fix debug_size parameters (3864)
5460 - fix mds_postrecov to initialize import for llog ctxt (3121)
5461 - replace config semaphore with spinlock (3306)
5462 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5463 - don't allow enqueue to complete on a destroyed export (3822)
5464 - down write_lock before checking llog header bitmap (3825)
5465 - recover from lock replay timeout (3764)
5466 - up llog sem before sending rpc (3652)
5467 - reduce ns lock hold times when setting kms (3267)
5468 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5469 - fix NULL deref and obd_dev leak on setup error (3312)
5470 - replace some LBUG about llog ops with error handling (3841)
5471 - don't match INVALID dentries from d_lookup and spin (3784)
5472 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5473 - fix invalid assertion in ptlrpc_set_wait (3880)
5475 - add libwrap support for the TCP acceptor (3996)
5476 - add /proc/sys/portals/routes for non-root route listing (3994)
5477 - allow setting MDS UUID in .xml (2580)
5478 - print the stack of a process that LBUGs (4228)
5480 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5483 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5484 - make sure to unset replay flag from failed open requests (3440)
5485 - if default stripe count is 0, use OST count for inode size (3636)
5486 - update parent mtime/ctime on client for create/unlink (2611)
5487 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5488 - free recovery state on server during a forced cleanup (3571)
5489 - unregister_reply for resent reqs (3063)
5490 - loop back devices mounting and status check on 2.6 (3563)
5491 - fix resource-creation race that can provoke i_size == 0 (3513)
5492 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5493 - more debugging for page-accounting assertion (3746)
5494 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5495 - avoid deadlock after precreation failure (3758)
5496 - fix race and lock order deadlock in orphan handling (3450, 3750)
5497 - add validity checks when grabbing inodes from l_ast_data (3599)
5499 - add /proc/.../recovery_status to obdfilter (3428)
5500 - lightweight CDEBUG infrastructure, debug daemon (3668)
5501 - change default OSC RPC parameters to be better on small clusters
5502 - turn off OST read cache for files smaller than 32MB
5503 - install man pages and include them in rpms (3100)
5504 - add new init script for (un)mounting lustre filesystems (2593)
5505 - run chkconfig in %post for init scripts (3701)
5506 - drop scimac NAL (unmaintained)
5508 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5511 - clean kiobufs before and after use (3485)
5512 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5513 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5514 - change init script to fail more clearly if not run as root (1528)
5515 - allow clients to reconnect during replay (1742)
5516 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5517 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5518 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5519 - fix cond_resched (3554)
5520 - only evict unfinished clients after recovery (3515)
5521 - allow bulk resend, prevent data loss (3570)
5522 - dynamic ptlrpc request buffer allocation (2102)
5523 - don't allow unlinking open directory if it isn't empty (2904)
5524 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5525 - remove extraneous obd dereference causing LASSERT failure (3334)
5526 - don't use get_cycles() when creating temp. files on the mds (3156)
5527 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5528 - handle EEXIST for set-stripe, set proper directory name (3336)
5530 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5531 - fix ksocknal_fmb_callback() error messages (2918)
5533 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5536 - don't copy lvb into (possibly NULL) reply on error (2983)
5537 - don't deref dentry after dput, don't free lvb on error (2922)
5538 - use the kms to determine writeback rpc length (2947)
5539 - increment oti_logcookies when osc is inactive (2948)
5540 - update client's i_blocks count via lvb messages (2543)
5541 - handle intent open/close of special files properly (1557)
5542 - mount MDS with errors=remount-ro, like obdfilter (2009)
5543 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5544 - don't use cancelling-locks' kms values (2947)
5545 - use highest lock extent for kms, not last one (2925)
5546 - don't dereference ERR_PTR() dentry in error handling path (3107)
5547 - fix thread race in portals_debug_dumplog() (3122)
5548 - create lprocfs device entries at setup instead of at attach (1519)
5549 - common AST error handler, don't evict client on completion race (3145)
5550 - zero nameidata in detach_mnt in 2.6 (3118)
5551 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5552 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5553 - initialize RPC timeout timer earlier for 2.6 (3219)
5554 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5555 - print nal/nid for unknown nid (3258)
5556 - additional checks for oscc recovery before doing precreate (3284)
5557 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5558 - don't crash in mdc_close for bad permissions on open (3285)
5559 - zero i_rdev for non-device files (3147)
5560 - clear page->private before handing to FS, better assertion (3119)
5561 - tune the read pipeline (3236)
5562 - fix incorrect decref of invalidated dentry (2350)
5563 - provide read-ahead stats and refine rpc in flight stats (3328)
5564 - don't hold journal transaction open across create RPC (3313)
5565 - update atime on MDS at close time (3265)
5566 - close LDAP connection when recovering to avoid server load (3315)
5567 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5568 - don't leak open file on MDS after open resend (3325)
5569 - serialize filter_precreate and filter_destroy_precreated (3329)
5570 - loop device shouldn't call sync_dev() for nul device (3092)
5571 - clear page cache after eviction (2766)
5572 - resynchronize MDS->OST in background (2824)
5573 - refuse to mount the same filesystem twice on same mountpoint (3394)
5574 - allow llmount to create routes for mounting behind routers (3320)
5575 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5576 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5577 - fix rare problem with rename on htree directories (3417)
5579 - allow default OST striping configuration per directory (1414)
5580 - fix compilation for qswnal for 2.6 kernels (3125)
5581 - increase maximum number of MDS request buffers for large systems
5582 - change liblustreapi to be useful for external progs like lfsck (3098)
5583 - increase local configuration timeout for slow disks (3353)
5584 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5586 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5589 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5590 - don't overwrite extent policy data in reply if lock was blocked (2901)
5591 - drop filter export grants atomically with removal from device (2663)
5592 - del obd_self_export from work_list in class_disconnect_exports (2908)
5593 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5594 - swab reply message in mdc_close, other PPC fixes (2464)
5595 - fix destroying of named logs (2325)
5596 - overwrite old logs when running lconf --write_conf (2264)
5597 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5598 - fix race in target_handle_connect (2898)
5599 - mds_reint_create() should take same inode create lock (2926)
5600 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5601 - don't close files for self_export to avoid uninitialized obd (2936)
5602 - allow MDS with the same name as client node (2939)
5603 - hold dentry reference for closed log files for unlink (2325)
5604 - reserve space for all logs during transactions (2059)
5605 - don't evict page beyond end of stripe extent (2925)
5606 - don't oops on a deleted current working directory (2399)
5607 - handle hard links to targets without a parent properly (2517)
5608 - don't dereference NULL lock when racing during eviction (2867)
5609 - don't grow lock extents when lots of conflicting locks (2919)
5611 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5614 - account for cache space usage on clients to avoid data loss (974)
5615 - lfsck support in lustre kernel code (2349)
5616 - reduce journal credits needed for BRW writes (2370)
5617 - orphan handling to avoid losing space on client/server crashes
5618 - ptlrpcd can be blocked, stopping ALL progress (2477)
5619 - use lock value blocks to assist in proper KMS, faster stat (1021)
5620 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5621 - recovery for initial connections (2355)
5622 - fixes for mds_cleanup_orphans (1934)
5623 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5624 - block all file creations until orphan recovery completes (1901)
5625 - client remove rq_connection from request struct (2423)
5626 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5627 - recovery timer race (2670)
5628 - mdc_close recovey bug (2532)
5629 - ptlrpc cleanup bug (2710)
5630 - mds timeout on local locks (2588)
5631 - namespace lock held during RPCs (2431)
5632 - handle interrupted sync write properly (2503)
5633 - don't try to handle a message that hasn't been replied to (2699)
5634 - client assert failure during cleanup after abort recovery (2701)
5635 - leak mdc device after failed mount (2712)
5636 - ptlrpc_check_set allows timedout requests to complete (2714)
5637 - wait for inflight reqs when ptlrpcd finishes (2710)
5638 - make sure unregistered services are removed from the srv_list
5639 - reset bulk XID's when resending them (caught by 1138 test)
5640 - unregister_bulk after timeout
5641 - fix lconf error (2694)
5642 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5643 - readahead locks pages, leaves pending causing memory pressure (2673)
5644 - increase OST request buffers to 4096 on large machines (2729)
5645 - fix up permission of existing directories in simple_mkdir (2661)
5646 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5647 - don't assign transno to errored transactions (2742)
5648 - don't delete objects on OST if given a bogus objid from MDS (2751)
5649 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5650 - if rq_no_resend, then timeout request after recovery (2432)
5651 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5652 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5653 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5654 - MDS recovery shouldn't delete objects that it has given out (2730)
5655 - if enqueue arrives after completion, don't clobber LVB (2819)
5656 - don't unlock pages twice when trigger_group_io returns error (2814)
5657 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5658 - don't write pages to disk if there was an error (1450)
5659 - don't ping imports that have recovery disabled (2676)
5660 - take buffered bytes into account when balancing socknal conn (2817)
5661 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5662 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5663 - remove little-endian swabbing of llog records (1987)
5664 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5665 - retry reposting request buffers if they fail (1191)
5666 - grow extent at grant time to avoid granting a revoked lock (2809)
5667 - lock revoke doesn't evict page if covered by a second lock (2765)
5668 - disable VM readahead to avoid reading outside lock extents (2805)
5670 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5671 - updated kernel patches for hp-2.4.20 kernel (2681)
5673 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5676 - fix truncated write corruption (2366)
5677 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5679 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5680 - ensure len(uuid) < 37 in lmc (1171)
5681 - fix ia64 OOPS in llog_test (2255)
5682 - zero end of page at obdfilter for partial page writes (2648)
5683 - don't leave stale dentries around after renames (bug 2428)
5684 - fix timeouts when evicting a client with a single lock held (2642)
5685 - set deadline for the initial HELLO message to drain (2634)
5686 - print out dotted-quad IP addresses in the socknal (2302)
5688 - additional debugging for MDS client eviction problem (2443)
5689 - fix mkfsoptions support for osts (2603, 2604)
5691 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5694 - add series for the vanilla 2.6.0 kernel
5695 - add series for the vanilla 2.4.24 kernel
5696 - add series for a cray x86/64 UL kernel drop
5697 - fix xattr patches for the vanilla 2.4.19 series
5699 - generate true UUIDs in lmc (1171)
5700 - have portals stack dumping break in UML (2466)
5701 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5702 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5703 - wake callback waiting threads on client eviction (2460)
5704 - Add --ptldebug and --subsystem to lmc (1719)
5705 - update assertion to allow safe interrupt allocation
5706 - set rq_no_resend for cancel requests (2432)
5707 - recalculate ptlrpcd timeout after resend (2494)
5708 - call vfs_rmdir when removing pending directories (2368)
5709 - fix renaming a file to itself (2429)
5710 - lmc creates a default one-stripe lov (2454)
5711 - expand procfs space to handle large clusters (2326)
5712 - increase UML stack to avoid overflow
5713 - update lconf's list of debug and subsystem masks
5714 - fix lfs find --obd (2510)
5715 - /proc tunable for disabling filter read caching (2591)
5716 - stop rpm packages from altering slapd.conf (2301)
5717 - disable nagle in the socknal under 0conf (2578)
5718 - choose mds inode size based on stripe count (2572)
5719 - fix kernel-source rpm problems (2516)
5721 - add --disable-doc to avoid pdf generation (2421)
5722 - update documentation, tests, type-os, comments
5723 - avoid format warnings on ia64
5724 - remove the TOE NAL
5725 - tiny code cleanups by removing unused fields
5727 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5730 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5731 - remove the most blatant lies from BUILDING file (2371)
5732 - change default debug level to reasonable production setting
5733 - reduce client side cache size to reduce cache flush time
5734 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5735 - make TCP zerocopy and pinger support enabled by default (2476)
5736 - sync writes completed after process exits caused crashes (2319)
5737 - maintain correct mount count on the MDS (2356)
5738 - backout 1557, because 2316 wasn't really fixed
5739 - better file I/O statistics gathering in /proc
5740 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5741 - another kernel patch to fix zero-copy TCP function export
5742 - don't take duplicate lock when processing re-sent getattr (2420)
5743 - lctl uses obd_self_export instead of creating new conn (2353)
5744 - MDS/OST recovery case which requires object creation asserted (2425)
5745 - move lfs from /usr/sbin to /usr/bin in packages
5746 - fix race between mds_client_add and mds_client_free (2417)
5747 - use kmalloc instead of slabs in portals (2430)
5748 - don't create duplicate records when a failover MDS is present (2442)
5749 - remove unnecessary mount age check (2332)
5750 - don't remove directory inodes from locks prematurely (2451)
5751 - don't break if MDS service name is the same as hostname (2103)
5752 - fix races in client write RPC generation when cache full (2482)
5754 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5757 - remove now-unused request->rq_obd (278)
5758 - if an allocation fails, print out how much memory we've used (1933)
5759 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5760 - add the "configurable stack size" patch to most series files (1256)
5761 - ability to write large log records, for 100+ OST configs (2306)
5762 - fix NULL deref when filter_prep fails (2314)
5763 - fix operator precedence error in filter_sync
5764 - dynamic allocation of socknal TX descriptors (2315)
5765 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5766 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5767 - ensure that CWARN messages in Portals always get to the syslog
5768 - __init/__exit are not for prototype decls (ldlm_init/exit)
5769 - x86-64 compile warning fixes
5770 - fix gateway LMC keyword conflict (2318)
5771 - fix MDS lock inversions in getattr/reint paths (1844)
5772 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5773 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5774 - DLM race condition prevented some lock evictions (2328)
5775 - ENOMEM detection and retry on socknal sends (2230)
5776 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5777 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5778 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5779 - better error messages when a client is rejected during recovery (1505)
5780 - avoid cancelling locks which were never granted, after failure (2330)
5781 - fix i_sem/journal inversion in mds_client_add (2333)
5782 - fix truncate/getattr lock cycle deadlock (2334)
5783 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5784 - fix two rare exit paths which could leak an l_lock() ref (2321)
5785 - fencepost error in MDS/OST orphan recovery (2226)
5786 - make log record alignment 8 bytes (1988)
5787 - lstripe now fails when requested offset > ost_count (2237)
5788 - ensure that all kernel series have a complete list.h (1607)
5789 - fix crashes in special-file operations (2316)
5790 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5792 - allow configurable automake binary, for testing new versions
5793 - small update to the lfs documentation
5795 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5797 * fix negative export reference count in fsfilt_sync (2312)
5799 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5800 * release candidate 0.9.1
5802 - orphans are moved into the PENDING directory for possible recovery
5803 - replayed opens now open by fid for orphan/rename safety (1042)
5804 - last close of an orphan inode generates a transno (683)
5805 - chdir() and mount() now pin the directory entry (1020)
5806 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5807 - discard very old requests without processing them (1502)
5808 - remove some common, well-understood CERRORs (1505)
5809 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5810 - clear "grant" flags in OST replies until OST grant code lands (1644)
5811 - fix read performance by not clobbering i_blksize on client (1598)
5812 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5813 - make LRU size a /proc tunable, clears locks when reduced (707)
5814 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5815 - server locks take references on exports now (1558)
5816 - build fixes for 2.4.20-rh trees (1663)
5817 - return an error from lov_create if all OSCs are inactive (1751)
5818 - fix import levels when a reconnect happens without a timeout (1597)
5819 - exit early from mds_open if we get a lookup error (1749)
5820 - partial page read at EOF wouldn't wait for disk before sending (1642)
5821 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5822 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5823 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5824 - write/truncate lock inversion (1639)
5825 - Don't auto-load obdclass, portals modules during cleanup (1495)
5826 - fix timestamps from jumping to "now" (1763)
5827 - extra journal assertions (1648)
5828 - add an extra multiunlink test (1771)
5829 - fix read_record/write_record API (1776)
5830 - fix leak of offset_extent, possible incorrect i_size later (1772)
5831 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5832 - Debugging for the kqswnal_get_idle_tx problems (1820)
5833 - Allow recovery to be attempted multiple times (1536)
5834 - Write out MDS last_rcvd file after it is first created (1600)
5835 - Fix tx_descriptor leak in failed transmit situations (1827)
5836 - ext3 journaling fixes for assertion failure after IO error (1871)
5837 - class_export_put() on freed export after completion AST error (1896)
5838 - Fix revalidate looping in VFS (1322)
5839 - Don't access a freed export during MDS_REINT timeout (1521)
5840 - Add open-unlink recovery support on the MDS (1673,1764)
5841 - Return an error if no MDS data was read from last_rcvd (1946)
5842 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5843 - Don't setup a disk device that doesn't match exported UUID (317)
5844 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5845 - avoid committing NULL handle in force close
5846 - local.sh is now a one-stripe LOV configuration
5847 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5848 - fix ext3 htree duplicate directory entry corruption (1516)
5849 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5850 - update to new LOV EA format (2097)
5851 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5853 2003-06-15 Phil Schwan <phil@clusterfs.com>
5856 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5857 - per-import/export recovery handling (958, 931, 959)
5858 - multiple last-rcvd slots, for serving multiple FSes (949)
5859 - connections are again shared between multiple imp/exports (963, 964)
5860 - "umount -f" would hang if any requests needed to be sent (393, 978)
5861 - avoid pinning large req buffer by copying for queued messages (989)
5862 - add "uuid" to "lctl device" command to help upcalls (991)
5863 - "open" RPCs with transnos would confuse recovery counters (1037)
5864 - do proper endian conversion of all wire messages (288, 340, 891)
5865 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5866 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5867 - fix ptlrpc_connection leak in target_handle_connect (1174)
5868 - fix import refcounting bug in OST and MDS cleanup (1134)
5869 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5870 - fix very unlikely obd_types race condition (501)
5871 - remove osc_open hack for echo_client (1187)
5872 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5873 - a failure in read_inode2 leads to deadlock (1139)
5874 - cancel ack-locks as soon as transaction is committed (1072)
5875 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5876 - make sure to commitrw after any preprw to avoid deadlock (1162)
5877 - failing to execute a file in a lustre FS would lock inode (1203)
5878 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5879 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5880 - obd_uuid2tgt would walk off the end of the list (1255)
5881 - on IA64 the osc would give portals incorrect bulk size (1258)
5882 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5883 - fix lock inversion caused by new llite matching code (1282)
5884 - limit the number of dirty pages on a client to 10MB (1286)
5885 - timed out locks were not being corrected cancelled (1289)
5886 - fix O_DIRECT above 4GB on IA-32 (1292)
5887 * major user-visible changes
5888 - fail out/fail over policy now controlled by the upcall (993)
5890 - add OBD_PING to check server availability and failure (954)
5891 - lustre messages are now sent in sending host order (288, 340, 891)
5892 - add eadatalen to MDS getattr reply (340)
5893 - OST read replies may contain second buffer, with per-page status (593)
5895 2003-03-11 Phil Schwan <phil@clusterfs.com>
5898 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5899 - failed open()s could cause deadlock; fixed (867, 869)
5900 - stop cancelling OST locks when files are closed (481)
5901 - overlapping XID spaces caused network corruption (851, 853)
5902 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5903 - setattr_raw added, to do single-RPC, server-side setattrs
5904 - lmc/lconf syntax change for OST UUIDs
5905 - fix crashy race condition between ptlrpc_free_req and osc_close
5906 - don't use request in mdc_enqueue if we hit a timeout (889)
5907 - don't set the inode i_size for regular files from the MDS (896)
5908 - handle out of order completion AST (842)
5909 - don't LBUG if a lock request times out after receiving AST (913)
5910 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5911 - if a bad lock AST arrives, send an error instead of dropping entirely
5912 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5913 - fix leak in bulk IO when only partially completed (899, 900, 926)
5914 - fix O_DIRECT for ia64 (55)
5915 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5916 - C-z after timeout could hang a process forever; fixed (977)
5918 - client-side I/O cache (678, 924, 929, 941, 970)
5920 - READPAGE and SETATTRs which don't take server-side locks get
5923 2003-02-11 Phil Schwan <phil@clusterfs.com>
5926 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5927 - Fix fsfilt_extN_readpage() to read a full page of directory
5928 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5929 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5930 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5931 - Most intent operations were removed, in favour of a new RPC mode
5932 that does a single RPC to the server and bypasses most of the VFS
5933 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5934 - Aggressively cancel local locks on DLM servers
5935 - mds_reint_unlink sends EA to the client if it's the last nlink.
5936 client uses that EA to unlink OST objects.
5937 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5938 - recursive symlinks were fixed (439)
5939 - fixed NULL deref in DEBUG_REQ
5940 - filter_update_lastobjid no longer calls sync, which annoyed extN
5941 - fixed multi-client small-writes to a single file problem (445)
5942 - fixed mtime updates during file writes (607)
5943 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5944 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5945 - fixed Linux OST size reporting problem (444, 656)
5946 - OST now updates object mtime with writes or setattr (607, 619)
5947 - client verifies file size before zeroing page past EOF (445)
5948 - OST now writes last allocated objid to disk with allocation (108)
5949 - LOV on echo now works (409)
5951 - mds_reint_unlink sends a new buffer, with the EA included. this
5952 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5953 set if a regular file was being unlinked, and it was the last link
5954 - use PtlGet from the target for bulk writes (315)
5955 - OST now updates object mtime with writes or setattr (607, 619)
5956 - LDLM now has a grant-time callback to revalidate locked items, if
5958 - Many MDS operations were reorganized to combat race conditions
5960 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5961 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5963 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5966 - Fully reactivate OST imports after reconnection (512, others)
5967 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5968 - More graceful error handling for truncating on dead OST (515)
5969 - Don't error out unless we're actually accessing dead stripes (474)
5970 - Fix garbage sizes when stripes are missing (410)
5971 - LRU counters were broken, causing constant lock purge (433, 432)
5972 - garbage on read from stripes with failed OSTs (441)
5973 - mark OSCs as active before reconnecting during recovery (438)
5974 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5975 - lfind did not preserve OST order in output (443)
5976 - symlinks cause hung clients, incorrect data (439)
5977 - stop dereferencing request after dropping refcount (457)
5978 - don't LASSERT(spin_is_locked) on non-SMP (455)
5979 - fixes for many rename() bugs
5980 - fstat didn't correctly synchronize attributes (399)
5981 - server must handle lock cancellation during blocking AST prep (487)
5982 - bulk descriptors were free()d too soon (511)
5983 - fix paths in lconf, which would load incorrect modules (451, 507)
5984 - fix confusing lconf 'host not found' error message (386)
5985 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5986 - fix race condition in mdc_blocking_ast() for inode access (526)
5987 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5988 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5989 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5990 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5991 - fix lstripe in several ways (kernel side) (527)
5992 - fix request leak in ldlm_cli_enqueue (262)
5993 - incorrect OSC was marked inactive after OST failure
5994 - call mds_fs_cleanup before unmounting filesystem (524)
5995 - fix races between taking ns_lock and ldlm_lock_change_resource
5996 - fix races updating LOV export open file list
5997 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5998 - fix recovery NULL deref in ldlm_cli_cancel_unused
5999 - fix some DLM races by using new hash table for lock handles (419)
6000 - permit the client to specify desired inodes, at replay
6001 - duplicate requests when we queue them for replay reintegration
6002 - fix last_rcvd offset calculation
6003 - sync after each recovered transaction, so we always make progress
6004 - never, not always, ERESTART requests without transnos
6005 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6007 - skip replay if the MDS says that the client is already connected
6008 - don't check for a recovery-enabled export to match lctl's UUID
6009 - don't INC_USE_COUNT for phantom exports
6010 - don't crash when cleaning up phantom exports (567)
6011 - don't double-finish or set replay data for errored mdc_open requests
6012 - abort requests when they time out, so we don't get old replies
6013 - send/receive replies for AST messages again
6014 - if the client says that it doesn't have the lock, cancel it on the
6016 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6017 mark it as destroyed, it will all work out when decref is called
6018 - fix module use counts (22, 581)
6020 - ASTs now expect a reply (server cancels lock on error reply)
6022 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6025 - fix many simultaneous client startup (392)
6026 - fix dentry->d_it clobbering
6027 - credentials weren't being shipped for readdir/getattr operations
6028 - remove invalid assertions triggered during some concurrent MD
6030 - proper Lustre versions added (336, 389)
6031 - fix memory leak for create error case (398)
6032 - fix LOV locking bug that would get cli/srv out of sync
6033 - fix echo client over LOV (409)
6034 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6035 - fix double-O_EXCL intent crash (424)
6036 - avoid sending multiple lock CANCELs (352)
6038 - MDS can do multi-client recovery (modulo bugs in new code)
6040 - many updates, edits, cleanups
6042 2002-11-18 Phil Schwan <phil@clusterfs.com>
6045 - fix null d_it dereference (346)
6046 - fix full OST/dbench hang (333)
6047 - fix permission problem with file removal (286)
6048 - fix removal of OSCs from LOV when they fail
6049 - fix NULL deref during bulk timeout (214)
6050 - fix problems related to multiple filesystems on one MDS (241)
6051 - fixed serious subtle metadata locking bugs
6052 - free locks on clients when inodes are removed due to memory
6054 - fix inode pointer in lock data (285)
6055 - partial support for multiple MDS on a single host (241)
6056 - data locks weren't cancelled at clear_inode time (290, 311)
6057 - intent locks could lead to unbounded lock growth (205)
6058 - added a maximum lock count, an LRU list, and a flusher
6059 - fix multiple rename (365)
6060 - properly abstracted the echo client
6061 - OSC locked 1 byte too many; fixed
6062 - rewrote brw callback code:
6063 - fixed recovery bugs related to LOVs (306)
6064 - fixed too-many-pages-in-one-write crash (191)
6065 - fixed (again) crash in sync_io_timeout (214)
6066 - probably fixed callback-related race (385)
6068 - Add capability to MDS protocol
6069 - LDLM cancellations and callbacks on different portals
6071 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6074 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6075 - defer open object destruction to close time (601981, 138)
6076 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6077 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6078 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6079 - don't call obd_setattr in ll_file_release for destroyed objects
6080 * protocol change to lustre_msg: move |version| and add |flags|
6081 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6082 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6083 * added replay of create, unlink, link and rename operations during
6084 MDS failover; recovery should be much more robust now
6085 * remove failed OSCs from LOVs (only lov_create uses this so far)
6086 * the lustre-HOWTO was brought (more) up to date (582544)
6088 2002-10-23 Phil Schwan <phil@clusterfs.com>
6091 - in-use dentries weren't being reused properly (617851)
6092 - prevent multiple LDLM setup (599178)
6093 - fix LOV size calculations for truncate (617853)
6094 - fix client handling of MDS intent errors (POSIX)
6095 - fix permission bug in lovstripe.c test (624321)
6096 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6097 - truncate past end of file could corrupt data
6098 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6099 - a race in recovery could return ETIMEDOUT to apps (623947)
6100 - building outside the source directory was fixed
6101 * the lustre-HOWTO was brought (more) up to date (582544)
6102 * major progress was made on recovery functionality
6104 2002-10-10 Phil Schwan <phil@clusterfs.com>
6107 - recovery deadlock fix
6108 - rm -rf causes LBUG fix (617817)
6109 - file open by multiple tasks fix (618962)
6110 - directory permissions bugs (602707 and 620007)
6111 - journal_stop fixed with locking (611313)
6112 - O_APPEND failures resolved (618273, perhaps 614459)
6113 - lconf PATH fix (619770)
6114 - IA64 build fix (621450)
6115 - RPC buffer sizes scale with amount of memory
6117 2002-10-01 Phil Schwan <phil@clusterfs.com>
6120 - locks would be cancelled without throwing away data pages,
6121 resulting in inconsistent data (605627)
6122 - inode attributes were not always being refreshed (605627, 612449)
6123 - lconf now continues to cleanup after lctl reports an error
6124 - MDS now enforces user permissions (602707)
6125 - lprocfs cleanup fixed, but not yet enabled (614157)
6126 - fixed infinite server hang, should a client not respond to an AST
6127 - avoid going into recovery if user calls readlink() with a buffer
6128 that's too small (613941)
6129 - AST RPCs no longer require replies (614867) -- this may be changed
6130 - don't crash server if client sends an IOV that's too big (611336)
6131 - fixed lock conversion deadlock (611892)
6132 - fixed the following of symlinks (614622)
6133 * recovery: the server can remove locks from a client that dies, other
6134 clients can make progress
6135 * more extN patch fixes
6136 * compile-time configurable ptlrpc buffer allocations
6138 - collaborative read cache document
6139 - Lustre Lite Performance CDR document-in-progress
6141 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6144 - fix typo in patch-2.4.18
6146 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6149 - clear ptlrpc request each time in handle_incoming_request()
6150 - unlink of files now destroys the object on the OST
6152 2002-09-19 Peter Braam <braam@clusterfs.com>
6154 * add hard link support
6155 * change obdfile creation method
6156 * kernel patch changed
6158 2002-09-19 Peter Braam <braam@clusterfs.com>
6161 - stack overflow bug in extN fixed
6163 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6165 * documentation updates
6166 - add man pages for config tools
6167 - update tests/README to describe testing with new config tools
6168 - finish metadata API descriptions
6169 * bug fixes and cleanups
6170 - statfs workaround for 16TB limit
6171 - LOV stripe allocation improved, can stripe on subset of OSTs
6172 - LOV file size/IO offset was wrong for files > 4GB in size
6173 - object EA data was being dropped, caused files to be unreadable
6174 - memory overflow with non-LOV OST caused memory corruption
6175 - fixed regression tests to work with new config tools, obdfilter
6176 - fixed bug when directory size became larger than 1 block
6177 - fixed bug (for single client case) when PWD was deleted
6178 - invalidate local directory pages when doing intent-based ops
6179 - avoid LDLM oops when lock callback contained bad data
6181 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6183 * documentation updates
6184 * bug fixes and cleanups
6185 - configuration tools
6188 - 64-bit compile warnings
6189 - 64-bit internal statfs data
6191 * test_brw on persistent OST devices
6193 * lprocfs (disabled)
6195 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6197 * documentation updates
6198 * bug fixes and cleanups
6199 * configuration tools
6201 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6204 * many small fixes to 0_5_4
6205 * io/network handling
6206 * thinkos in MDS operations
6208 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6211 * crucial basic fixes to 0.5.3
6212 * IOR, Iozone work over Elan
6215 2002-08-07 Phil Schwan <phil@clusterfs.com>
6216 * version 0_5_3, our first alpha
6217 * we use the new Portals iovs
6218 * documentation updates
6219 * bug fixes and cleanups
6220 * small changes in the DLM wire protocol
6222 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6223 * version 0_5_1 with some initial stability,
6224 * locking on MD and file I/O.
6225 * documentation updates
6226 * several bug fixes since 0.5.0
6227 * small changes in wire protocol
6229 2002-07-18 Phil Schwan <phil@clusterfs.com>
6231 * delivered as Lustre Light Alpha
6232 * fixed a crash after handling invalid MDS requests
6233 * fixed directory pages for architectures with non-4k pages sizes
6235 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6236 * release version v0_4_4
6237 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6239 * Fixes a number of interruption problems with OST operations.
6240 * Update documentation for portals header changes
6241 * Move all wire protocol structs/defines to lustre_idl.h
6242 * Fixes symlink length bug.
6243 * Add tcpdump to repository.
6245 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6246 * release version v0_4_3
6247 * Fixes statfs for inodes on extN.
6248 * Fixes bug in runtests which would delete /etc/hosts.
6249 * Use 64-bit object IDs wherever possible (not into VFS though)
6250 Remove ost_get_info, which is unused by lustre, and out of date.
6252 2002-07-03 Peter Braam <braam@clusterfs.com>
6253 * release version v0_4_2 Fixes a lookup error (type not passed)
6254 * move forward to head of Portals
6255 * move forward to latest Lustre kernel
6257 2002-06-25 Peter Braam <braam@clusterfs.com>
6258 * release version v0_4_1. Hopefully stable on single node use.