1 tbd Sun Microsystems, Inc.
5 2.6.9-67.0.15.EL (RHEL 4),
6 2.6.16.54-0.2.5 (SLES 10),
7 2.6.18-53.1.21.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: 1.40.7-sun3
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 will not 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.
27 Severity : enhancement
29 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
33 Description: do not process already freed flock
34 Details : flock can possibly be freed by another thread before it reaches
35 to ldlm_flock_completion_ast.
39 Description: LBUG during stress test
40 Details : Need properly lock accesses the flock deadlock detection list.
44 Description: oops in page fault handler
45 Details : kernel page fault handler can return two special 'pages' in error case, don't
46 try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
50 Description: timeout with invalidate import.
51 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which should be
52 handled by ptlrpcd. This produce long age waiting and -ETIMEOUT
53 ptlrpc_invalidate_import and as result LASSERT.
56 Frequency : only with broken builds/installations
58 Description: do not LBUG if lquota.ko and fsfilt_ldiskfs.ko are of different versions
59 Details : just return an error to a user, put a console error message
61 Severity : enhancement
63 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
65 Severity : enhancement
67 Description: Update to RHEL4 kernel-2.6.9-67.0.15.
71 Description: enable MGS and MDT services start separately
72 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
73 a co-located MGS without starting the MGS, which is a complement
74 to 'nosvc' mount option.
77 Frequency : always, on ppc.
79 Description: cleanup in ptlrpc code, related to ppc platform
80 Details : store magic in native order avoid panic's in recovery on ppc node
81 and forbid from this error in future. Also fix posibily of twice swab
82 data. Fix get lov striping to userpace.
86 Frequency : rare, replay get lost on server
87 Description: server incorrectly drop resent replays lead to recovery failure.
88 Details : do not drop replay according to msg flags, instead we check the
89 per-export recovery request queue for duplication of transno.
93 Frequency : after recovery
94 Description: precreate to many object's after del orphan.
95 Details : del orphan st in oscc last_id == next_id and this triger growing
96 count of precreated objects. Set flag LOW to skip increase count
97 of precreated objects.
101 Frequency : rare, on clear nid stats
102 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
103 Details : when clean nid stats sometimes try destroy live entry,
104 and this produce panic in free.
108 Description: Stack overflow during MDS log replay
109 Details : ease stack pressure by using a thread dealing llog_process.
112 Frequency : very rare
114 Description: MDT cannot be unmounted, reporting "Mount still busy"
115 Details : Mountpoint references were being leaked during open reply
116 reconstruction after an MDS restart. Drop mountpoint reference
117 in reconstruct_open() and free dentry reference also.
121 Description: wait until IO finished before start new when do lock cancel.
122 Details : VM protocol want old IO finished before start new, in this case
123 need wait until PG_writeback is cleared until check dirty flag and
124 call writepages in lock cancel callback.
128 Description: mds_mfd_close() ASSERTION(rc == 0)
129 Details : In mds_mfd_close(), we need protect inode's writecount change
130 within its orphan write semaphore to prevent possible races.
134 Frequency : rare, on shutdown ost
135 Description: don't hit live lock with umount ost.
136 Details : shrink_dcache_parent can be in long loop with destroy dentries,
137 use shrink_dcache_sb instead.
141 Description: don't panic with use echo client
142 Details : echo client pass NULL as client nid pointer and this produce null
147 Description: fix build on ppc32
148 Details : compile code with -m64 flag produce wrong object file for ppc32.
152 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
153 Details : In reconstruct_* functions, LASSERTs on both the data supplied by
154 a client, and the data on disk are dangerous and incorrect. Change
155 them with client eviction.
159 Description: skiplist implementation simplification
160 Details : skiplists are used to group compatible locks on granted list
161 that was implemented as tracking first and last lock of each lock group
162 the patch changes that to using doubly linked lists
166 Description: delete compatibility for 32bit qdata
167 Details : as planned, when lustre is beyond b1_8, lquota won't support for 32bit
168 qunit. That means servers of b1_4 and servers of b1_8 can't be used
169 together if users want to use quota.
172 Frequency : blocks per group is less than blocksize*8 and uninit_groups is enabled
174 Description: ldiskfs error: XXX blocks in bitmap, YYY in gd
175 Details : If blocks per group is less than blocksize*8, set rest of the
179 Frequency : Application do stride read on lustre
181 Description: The read performance will drop a lot if the application does
183 Details : Because the stride_start_offset are missing in stride read-ahead,
184 it will cause clients read a lot of unused pages in read-ahead,
185 then the read-performance drops.
187 -------------------------------------------------------------------------------
190 2008-05-26 Sun Microsystems, Inc.
192 * Support for kernels:
193 2.6.5-7.311 (SLES 9),
194 2.6.9-67.0.7.EL (RHEL 4),
195 2.6.16.54-0.2.5 (SLES 10),
196 2.6.22.14 vanilla (kernel.org)
197 * Client support for unpatched kernels:
198 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
199 2.6.16 - 2.6.22 vanilla (kernel.org)
200 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
201 we do not recommend using patchless RHEL4 clients with kernels
202 prior to 2.6.9-55EL (RHEL4U5).
203 * Recommended e2fsprogs version: 1.40.7-sun1
204 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
205 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
206 removed cwd "./" (refer to Bugzilla 14399).
207 * A new quota file format has been introduced in 1.6.5.
208 The format conversion from prior releases is handled transparently,
209 but releases older than 1.4.12/1.6.5 will not understand this new
210 format. The automatic format conversion can be avoided by running
211 the following command on the MDS before upgrading:
212 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
213 For more information, please refer to bugzilla 13904.
217 Description: quota performance fix
218 Details : quota data is written in journalled mode instead of ordered to
223 Description: lfs support for human-readable quota grace time strings
224 Details : lfs setquota -t and lfs quota -t represent quota grace times
225 in "XXwXXdXXhXXmXXs" format instead of large values in seconds
228 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
230 Description: MDS may lose file striping (and hence file data) in some cases
231 Details : If there are additional extended attributes stored on the MDS,
232 in particular ACLs, SELinux, or user attributes (if user_xattr
233 is specified for the client mount options) then there is a risk
234 of attribute loss. Additionally, the Lustre file striping
235 needs to be larger than default (e.g. striped over all OSTs),
236 and an additional attribute must be stored initially in the
237 inode and then increase in size enough to be moved to the
238 external attribute block (e.g. ACL growing in size) for file
241 Severity : enhancement
243 Description: add message levels for liblustreapi
248 Description: MDT cannot be unmounted, reporting "Mount still busy"
249 Details : Mountpoint references were being leaked during open reply
250 reconstruction after an MDS restart. Drop mountpoint reference
251 in reconstruct_open() and free dentry reference also.
256 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
257 Details : Move the 'good_osts' check before the 'total_bavail' check. This
258 will result in an -EAGAIN and in the exit call path we call
259 alloc_rr() which will with increasing aggressiveness attempt to
260 aquire precreated objects on the minimum number of required OSCs.
264 Description: Use old size assignment to avoid deadlock
265 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
266 the scheduling while holding a spinlock. We do not need locking
267 for size in ll_update_inode() because size is only updated from
268 the MDS for directories or files without objects, so there is no
269 other place to do the update, and concurrent access to such inodes
270 are protected by the inode lock.
274 Description: Use __u64 instead of int for valid bits
278 Description: resolve "_IOWR redefined" build error on SLES10
282 Description: dump the memory debugging after all modules are unloaded to
283 suppress false negative in conf_sanity test 39
287 Description: the recovery timer never expires
288 Details : for new client connect request, the recovery timer should not be
289 reset, otherwise recovery timer will never expired, if the old
290 client never come. Only old client connect and first connection
291 req should trigger recovery timer reset.
295 Description: the min numbers of lproc stats are wrong
296 Details : adding a new constant LC_MIN_INIT and use it for initialization
300 Frequency : always with interactive lfs
302 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
305 Frequency : with multiple concurrent readdir processes in same directory
306 Bugzilla : 15406, 15169, 15175
307 Description: misc fixes for directory readahead.
308 Details : prevent previous statahead async RPC callback from processing the
309 current "statahead_info", race condition between async RPC callback
310 add dentry into dentry hash table and "ls" thread revalidate such
311 dentry, statahead his/miss control for hidden items, and so on.
313 Severity : enhancement
315 Description: build kernel-ib packages for OFED 1.3 in our release cycle
319 Description: incore types cleaning in quota code (with respect to 64-bit limits)
320 Details : several u32 variables declarations are replaced with u64 declarations
325 Description: fix SLES kernel versioning
326 Details : the kernel version for our SLES 10 kernel did not include a "-"
327 before the "smp" at the end. while this was not a problem in
328 general, it did mean that software trying to use the kernel
329 version to try to detect a vendor specific kernel would fail.
330 this was most evident by the OFED build scripts.
335 Description: Don't update lov_desc members until making sure they are valid
336 Details : When updating lov_desc members via proc fs, need fix their
337 validities before doing the real update.
340 Frequency : very rare
342 Description: don't put request into delay list while invalidate in flight.
343 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
344 import in flight. this produce timeout for invalidate and sometimes
345 can cause stale data.
347 Severity : enhancement
349 Description: Update kernel to SLES9 2.6.5-7.311.
351 Severity : enhancement
353 Description: Update kernel to RHEL4 2.6.9-67.0.7.
358 Frequency : on PPC only
359 Description: not convert ost objects for directory because it's not exist.
360 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
362 Severity : enhancement
364 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
365 __u64 as usingied long long type.
368 Frequency : rare, on shutdown
370 Description: race process ast vs remove callback
371 Details : removing callback before disconnect import open race with
374 Severity : enhancement
376 Description: Update kernel to SLES9 2.6.5-7.311.
378 Severity : enhancement
380 Description: Files open for execute are not marked busy on SLES10
381 Details : Add FMODE_EXEC to SLES10 SP1 server kernel series.
383 Severity : enhancement
385 Description: Add server support for vanilla-2.6.22.14.
388 Frequency : occasional
390 Description: Avoid lov_create() getting stuck in obd_statfs_rqset()
391 Details : If an OST is down the MDS will hang indefinitely in
392 obd_statfs_rqset() waiting for the statfs data. While for
393 MDS QOS usage of statfs, it should not stuck in waiting.
395 Severity : enhancement
397 Description: Disable adaptive timeouts by default
400 Frequency : on network error
402 Description: panic with double free request if network error
403 Details : mdc_finish_enqueue is finish request if any network error occuring,
404 but it's true only for synchronus enqueue, for async enqueue
405 (via ptlrpcd) this incorrect and ptlrpcd want finish request
409 Frequency : rare, on recovery
411 Description: read procfs can produce deadlock in some situation
412 Details : Holding lprocfs lock with send rpc can produce block for destroy
413 obd objects and this also block reconnect with -EALREADY.
414 This isn't fix all lprocfs bugs - but make it rare.
416 Severity : enhancement
418 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
421 Frequency : frequent on X2 node
423 Description: mdc_set_open_replay_data LBUG
424 Details : Set replay data for requests that are eligible for replay.
429 Description: lustre_mgs: operation 101 on unconnected MGS
430 Details : When MGC is disconnected from MGS long enough, MGS will evict the
431 MGC, and late on MGC cannot successfully connect to MGS and a lot
432 of the error messages complaining that MGS is not connected.
435 Frequency : rare, depends on device drivers and load
437 Description: MDS or OSS nodes crash due to stack overflow
438 Details : Code changes in 1.6.4 increased the stack usage of some functions.
439 In some cases, in conjunction with device drivers that use a lot
440 of stack, the MDS (or possibly OSS) service threads could overflow
441 the stack. One change which was identified to consume additional
442 stack has been reworked to avoid the extra stack usage.
444 Severity : enhancement
446 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
448 Severity : enhancement
450 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
452 Severity : enhancement
454 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
456 Serverity : enhancement
458 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
461 Frequency : rare on shutdown OST
463 Description: Don't allow skipping OSTs if index has been specified.
464 Details : Don't allow skipping OSTs if index has been specified, make locking
465 in internal create lots better.
470 Description: ASSERTION(!PageDirty(page)) failed
471 Details : Wrong check could lead to an assertion failure under specific
477 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
478 Details : access to bitfield in structure is always rounded to long
479 and this produce problem with not atomic change any bit.
484 Description: Lustre make rpms failed.
485 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
486 builds Lustre rpms from tarball.
488 Severity : enhancement
490 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
493 Frequency : rare on shutdown OST
495 Description: If llog cancel was not send before clean_exports phase, this can
496 produce deadlock in llog code.
497 Details : If llog thread has last reference to obd and call class_import_put
498 this produce deadlock because llog_cleanup_commit_master wait when
499 last llog_commit_thread exited, but this never success because was
500 called from llog_commit_thread.
503 Frequency : only if OST index is skipped
505 Description: NULL lov_tgts causing MDS oops
506 Details : more safe checks for NULL lov_tgts for avoid oops.
508 Severity : enhancement
510 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
512 Severity : enhancement
514 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
519 Description: make mgs_setparam() handle fsname containing dash
520 Details : fsname containing a dash does not work with lctl conf_param
522 Severity : enhancement
524 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
527 Frequency : rare, in recovery and (or) destroy lovobjid file.
529 Description: rewrite lov objid code.
530 Details : Cleanup for lov objid code, remove scability problems and wrong
531 locking. Fix sending last_id into ost.
533 Severity : enhancement
535 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
537 Severity : enhancement
539 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
540 Details : Use d_move_locked instead of __d_move.
543 Frequency : rare, at shutdown
545 Description: access already free / zero obd_namespace.
546 Details : if client_disconnect_export was called without force flag set,
547 and exist connect request in flight, this can produce access to
548 NULL pointer (or already free pointer) when connect_interpret
549 store ocd flags in obd_namespace.
552 Frequency : only at startup
554 Description: not alloc memory with spinlock held.
555 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
556 if any spinlock held.
561 Description: lfs find does not continue on file error
562 Details : Continue other files processing when a file/dir is absent.
566 Description: Inconsistent usage of lustre_pack_reply()
567 Details : Standardize the usage of lustre_pack_reply() such that it
568 always generate a CERROR on failure.
571 Frequency : very rare
573 Description: Fix replay if there is an un-replied request and open
574 Details : In some cases, older replay request will revert the
575 mcd->mcd_last_xid on MDS which is used to record the client's
578 Severity : enhancement
580 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
582 Severity : enhancement
584 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
586 Severity : enhancement
588 Description: Update to SLES9 kernel-2.6.5-7.287.3.
590 Severity : enhancement
592 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
594 Severity : enhancement
596 Description: Build SLES10 patchless client fails
597 Details : The configure was broken by run ./configure with
598 --with-linux-obj=.... argument for patchless client. When the
599 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
600 can't search header adequately. Use absolute path such as
601 -I($LINUX)/include instead.
603 Severity : enhancement
605 Description: Lustre Page Accounting
606 Details : New macros for page alloc and free which enable accounting
607 of page allocation of Lustre. Use percpu counters to store memory
611 Frequency : only if debugging is disabled
613 Description: LASSERT_{REQ,REP}SWAB macros are buggy
614 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
615 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
616 these macros and replace them with their definitions instead.
621 Description: interrupt oig_wait produce painc on resend.
622 Details : brw_redo_request can be used for resend requests from ptlrpcd and
623 private set, and this produce situation when rq_ptlrpcd_data not
624 copyed to new allocated request and triggered LBUG on assert
625 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
626 ptlrpcd set if request is changed and can be safety changed to use
629 Severity : enhancement
631 Description: organize the server-side client stats on per-nid basis
632 Details : Change the structure of stats under obdfilter and mds to
640 The "uuid"s file would list the uuids of _active_ exports.
641 And the clear entry is to clear all stats and stale nids.
646 Description: Processes looping in ll_readdir() on Lustre clients finally causing
647 a full node pseudo-hang
648 Details : Concurrent access to the same directory from multiple clients with
649 intensive file creation/removal can cause a client node to spin in
650 ll_readdir(). i_version must be increased every time the lock is
651 cancelled to ensure a revalidate is done.
656 Description: touch file failed when fs is not full
657 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
658 otherwise we can get ENOSP while fs is not full.
661 Frequency : only for Cray XT3
662 Bugzilla : 12829/13455
663 Description: Changing primary group doesn't change the group lustre assigns to
665 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
666 overridden with the primary group provided by the group upcall,
667 whereas the supplied fsgid can be trusted if it is in the list of
668 supplementary groups returned by the group upcall.
670 Severity : enhancement
672 Description: Root Squash Functionality
673 Details : Implementation of NFS-like root squash capability. Specifically,
674 don't allow someone with root access on a client node to be able
675 to manipulate files owned by root on a server node.
677 Severity : enhancement
679 Description: Slow trucate/writes to huge files at high offsets.
680 Details : Directly associate cached pages to lock that protect those pages,
681 this allows us to quickly find what pages to write and remove
682 once lock callback is received.
687 Description: Too many locks accumulating on client during NFS usage
688 Details : mds_open improperly used accmode to find out access mode to a
689 file. Also mdc_intent_lock logic to find out if we already have
690 lock similar to just received was flawed since introduction of
691 skiplists - locks are now added to the front of the granted
696 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
698 Details : In obd setup failure handler, obd_stopping will not necessarily to
699 be 1, and obd_set_up should also be checked to make sure whether
700 obd is completely setup.
702 Severity : enhancement
704 Description: Allow masking D_WARNING, D_ERROR messages from console
705 Details : Console messages can now be disabled via lnet.printk.
710 Description: User code with malformed file open parameter crashes client node
711 Details : Before packing join_file req, all the related reference should be
712 checked carefully in case some malformed flags cause fake join_file
718 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
719 grain of quota doing harm to quota's accuracy
720 Details : qunit size will be changed when quota limitation is too low/high;
721 record the pending quota write in order to get more accureate
722 quota; delete the patch for bug12588, which is unnecessary when
723 this patch is landed. This bug also contains fixes for bug 14526,
724 14299, 14601 and 13794.
728 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
729 Details : ldlm_completion_ast() assumes that a lock is granted when the req
730 mode is equal to the granted mode. However, it should also check
731 that LDLM_FL_CP_REQD is not set.
735 Description: Heavy nfs access might result in deadlocks
736 Details : After ELC code landed, it is now improper to enqueue any mds
737 locks under och_sem, because enqueue might want to decide to
738 cancel open locks for same inode we are holding och_sem for.
742 Description: 35% write performance drop with ldiskfs2 when quotas are on
743 Details : Enable ext3 journalled quota by default to improve performance
744 when quotas are turned on.
748 Description: Client eviction while running blogbench
749 Details : A lot of unlink operations with concurrent I/O can lead to a
750 deadlock causing evictions. To address the problem, the number of
751 oustanding OST_DESTROY requests is now throttled to
752 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
753 callbacks are processed in priority.
756 Frequency : RHEL4 only
758 Description: mkfs is very slow on IA64/RHEL4
759 Details : A performance regression has been discovered in the MPT Fusion
760 driver between versions 3.02.73rh and 3.02.99.00rh. As a
761 consequence, we have downgraded the MPT Fusion driver in the RHEL4
762 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
765 Frequency : PPC/PPC64 only
767 Description: conflicts between asm-ppc64/types.h and lustre_types.h
768 Details : fix duplicated definitions between asm-ppc64/types.h and
769 lustre_types.h on PPC.
772 Frequency : PPC/PPC64 only
774 Description: asm-ppc/segment.h does not exist
775 Details : fix compile issue on PPC.
779 Description: data checksumming impacts single node performance
780 Details : add support for several checksum algorithms. Currently, CRC32 and
781 Adler-32 are supported. The checksum type can be changed on the fly
782 through /proc/fs/lustre/osc/*/checksum_type.
786 Description: use adler32 for page checksums
787 Details : when available, use the Adler-32 algorithm instead of CRC32 for
792 Description: better handle error messages in extents code
794 Severity : enhancement
796 Description: SNMP support enhancement
797 Details : Adding total number of sampled request for an MDS node in snmp
800 Severity : enhancement
802 Description: Optimize ldlm waiting list processing for PR extent locks
803 Details : When processing waiting list for read extent lock and meeting read
804 lock that is same or wider to it that is not contended, skip
805 processing rest of the list and immediatelly return current
806 status of conflictness, since we are guaranteed there are no
807 conflicting locks in the rest of the list.
811 Description: Time out and refuse to reconnect
812 Details : When the failover node is the primary node, it is possible
813 to have two identical connections in imp_conn_list. We must
814 compare not conn's pointers but NIDs, otherwise we
815 can defeat connection throttling.
819 Description: Client not clear own cache if answer to reconnect is lost.
820 Details : Client gets evicted from server. Now client also thinks it is
821 disconnected (or gets ENOTCONN on its operation) and decides to
822 reconnect. Server receives reconnect message, but cannot find
823 export. New export is created that is fully valid (new cookie!),
824 but reply is lost and not reported to client. Client reconnects
825 again and gets back a just-created connection, but it is not new
826 so client thinks it was not evicted and does not do recovery.
830 Description: Detect stride IO mode in read-ahead
831 Details : When a client does stride read, read-ahead should detect that and
832 read-ahead pages according to the detected stride pattern.
836 Description: build for x2 fails
837 Details : fix compile issue on Cray systems.
839 Severity : enhancement
841 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
842 Details : This greatly improves speed of NFS writes on 2.6 kernels.
845 Frequency : only on PPC/SLES10
847 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
848 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
849 be defined. Add a hack in mkfs_lustre.c to work around this compile
854 Description: LASSERT on MDS when client holding flock lock dies
855 Details : ldlm pool logic depends on number of granted locks equal to
856 number of released locks which is not true for flock locks, so
857 just exclude such locks from consideration.
861 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
862 Details : Use fsfilt_sync() for both the whole filesystem sync and
863 individual file sync to eliminate dangerous inode locking
864 with I_LOCK that can lead to a deadlock.
868 Description: Update an obsolete wirecheck.c generator
869 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
873 Description: Client can panic on open sometimes
874 Details : It is possible that we try to free already freed request in
875 ll_file_open in some error cases when we send request from
880 Description: performance in 1.6.3
881 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
882 sure that 1MB requests are not fragmented by the block layer.
886 Description: LDLM soft lockups - improvement
887 Details : It is be possible to send the lock handle along with each read
888 or write request because the client is already doing a lock match
889 itself so there isn't any reason the OST should have to re-do that
895 Description: lfs quota fails with deactivated OSTS
896 Details : With this patch, three improvements are included:
897 1. detete the softlimit in mds and osts when use "lfs quota".
898 2. display the inaccurate data in the output of "lfs quota".
899 3. try to get quota info when "lfs quota" is executed.
904 Description: Extent locks not granted with no conflicts sometimes.
905 Details : When race occurs in glimpse handler and nothing is returned,
906 we do not reprocess the queue after lock cancel, and that leads
907 to a stall until next activity on a resource
910 Frequency : failover with quotaon
912 Description: during mds failovers with quota on, OSTs got into deadlock state
913 and causing dumpstack.
914 Details : for every quota slave, at any time, there is only one quota req
915 is sent to quota master for every uid/gid. Before that quota req
916 returns, all the thread relative to the same uid/gid will wait.
917 So if the quota req is lost because mds failovers or any other
918 reasons, this bug will be hit. Now, dqacq_interpret() will handle
919 quota reqs who time out.
921 Severity : enhancement
924 Description: when quota slave checks if quota is enough, there is an unnecessary
926 Details : place this wait on necessary place instead of always waiting.
928 --------------------------------------------------------------------------------
930 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
932 * Support for kernels:
933 2.6.5-7.286 (SLES 9),
934 2.6.9-55.0.9.EL (RHEL 4),
935 2.6.16.53-0.8 (SLES 10),
936 2.6.18-8.1.14.el5 (RHEL 5),
937 2.6.18.8 vanilla (kernel.org)
938 * Client support for unpatched kernels:
939 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
940 2.6.16 - 2.6.22 vanilla (kernel.org)
941 * Due to recently discovered recovery problems, we do not recommend
942 using patchless RHEL 4 clients with this or any earlier release.
943 * Recommended e2fsprogs version: 1.40.2-cfs1
944 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
947 Frequency : occasional
949 Description: MDS hang or stay in waiting lock
950 Details : If client receive lock with CBPENDING flag ldlm need send lock
951 cancel as separate rpc, to avoid situation when cancel request
952 can't processed due all i/o threads stay in wait lock.
955 Frequency : occasional
957 Description: Do not fail import if osc_interpret_create gets -EAGAIN
958 Details : If osc_interpret_create got -EAGAIN it immediately exits and
959 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
960 oscc_has_objects and see OSC has no objests and call
961 oscc_internal_create to resend create request.
963 Severity : enhancement
965 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
966 Details : Update which_patch & target file for SLES10 latest kernel.
968 Severity : enhancement
970 Description: add --type and --size parameters to lfs find
971 Details : Enhance lfs find by adding filetype and filesize parameters. Also
972 multiple OBDs can now be specified for the --obd option.
974 Severity : enhancement
976 Description: eliminate client locks in face of contention
977 Details : file contention detection and lockless i/o implementation
980 Severity : enhancement
982 Description: Remove client patches from SLES 10 kernel.
983 Details : This causes SLES 10 clients to behave as patchless clients
984 even on a Lustre-patched (server) kernel.
986 Severity : enhancement
988 Description: use i_size_read and i_size_write in 2.6 port
989 Details : replace inode->i_size access with i_size_read/write()
992 Frequency : when removing large files
994 Description: scheduling issue during removal of large Lustre files
995 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
996 It causes scheduling issues when removing large files (17TB in the
1002 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
1003 Details : Impossible (EPROTO is returned) to access a directory that has a
1004 non-default striping and ACLs.
1007 Frequency : only on ppc
1009 Description: /proc/fs/lustre/devices broken on ppc
1010 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1011 We should make sure the type of 'index' is loff_t and then cast
1012 explicitly as needed below. Do not assign an explicitly cast
1016 Frequency : only for rhel5
1018 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1019 Details : Modify the target file & which_kernel.
1022 Frequency : if the uninit_groups feature is enabled on ldiskfs
1024 Description: e2fsck reports "invalid unused inodes count"
1025 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
1026 feature and only a single inode is created in a group then the
1027 "bg_unused_inodes" count is incorrectly updated. Creating a
1028 second inode in that group would update it correctly.
1033 Description: buffer overruns could theoretically occur
1034 Details : llapi_semantic_traverse() modifies the "path" argument by
1035 appending values to the end of the origin string, and a buffer
1036 overrun may occur. Adding buffer overrun check in liblustreapi.
1038 Severity : enhancement
1040 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
1043 Frequency : only if filesystem is inconsistent
1045 Description: handle "serious error: objid * already exists" more gracefully
1046 Details : If LAST_ID value on disk is smaller than the objects existing in
1047 the O/0/d* directories, it indicates disk corruption and causes an
1048 LBUG(). If the object is 0-length, then we should use the existing
1049 object. This will help to avoid a full fsck in most cases.
1051 Severity : enhancement
1053 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1054 Details : Modify vm-tunables-rhel4.patch.
1056 Severity : enhancement
1058 Description: Kernel config for 2.6.18-vanilla.
1059 Details : Modify targets/2.6-vanilla.target.in.
1060 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1061 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1062 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1063 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1066 Frequency : occasional
1068 Description: improve handling recoverable errors
1069 Details : If request processed with error which can be recoverable on server
1070 request should be resend, otherwise page released from cache and
1076 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1077 Details : Modify target file & which_patch.
1078 A flaw was found in the IA32 system call emulation provided
1079 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1080 value could be stored in the %RAX register, which could trigger an
1081 out-of-bounds system call table access. An untrusted local user
1082 could exploit this flaw to run code in the kernel
1083 (ie a root privilege escalation). (CVE-2007-4573).
1087 Description: change order of libsysio includes
1088 Details : '#include sysio.h' should always come before '#include xtio.h'
1090 Severity : enhancement
1092 Description: adapt the lustre_config script to support the upgrade case
1093 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1094 server targets to 1.6 in parallel.
1099 Description: To avoid grant space > avaible space when the disk is almost
1100 full. Without this patch you might see the error "grant XXXX >
1101 available" or some LBUG about grant, when the disk is almost
1103 Details : In filter_check_grant, for non_grant cache write, we should
1104 check the left space by if (*left > ungranted + bytes), instead
1105 of (*left > ungranted), because only we are sure the left space
1106 is enough for another "bytes", then the ungrant space should be
1107 increase. In client, we should update cl_avail_grant only there is
1108 OBD_MD_FLGRANT in the reply.
1113 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1114 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1115 described in CVE-2007-4573. This problem could allow untrusted
1116 local users to gain root access.
1119 Frequency : when using O_DIRECT and quotas
1121 Description: Incorrect file ownership on O_DIRECT output files
1122 Details : block usage reported by 'lfs quota' does not take into account
1123 files that have been written with O_DIRECT.
1125 --------------------------------------------------------------------------------
1127 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
1129 * Support for kernels:
1130 2.6.5-7.286 (SLES 9),
1131 2.6.9-55.0.2.EL (RHEL 4),
1132 2.6.16.46-0.14 (SLES 10),
1133 2.6.18-8.1.8.el5 (RHEL 5),
1134 2.6.18.8 vanilla (kernel.org)
1135 * Client support for unpatched kernels:
1136 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1137 2.6.16 - 2.6.21 vanilla (kernel.org)
1138 * Due to recently discovered recovery problems, we do not recommend
1139 using patchless RHEL 4 clients with this or any earlier release.
1140 * Recommended e2fsprogs version: 1.40.2-cfs1
1141 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1145 Description: Fix errors in lfs documentation
1146 Details : Fixes man pages
1148 Severity : enhancement
1150 Description: Adaptive timeouts
1151 Details : RPC timeouts adapt to changing server load and network
1152 conditions to reduce resend attempts and improve recovery time.
1154 Severity : enhancement
1156 Description: llapi_file_create() does not allow some changes
1157 Details : add llapi_file_open() that allows specifying the file creation
1158 mode and open flags, and also returns an open file handle.
1160 Severity : enhancement
1162 Description: df doesn't work properly if diskfs blocksize != 4K
1163 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1166 Frequency : always on directories with default striping set
1168 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1169 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1171 Severity : enhancement
1173 Description: merge and cleanup kernel patches.
1174 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1176 Severity : enhancement
1178 Description: RedHat Update kernel for RHEL5
1179 Details : Kernel config file for RHEL5.
1181 Severity : enhancement
1183 Description: OSS needs mutliple precreate threads
1184 Details : Add ability to start more than one create thread per OSS.
1187 Frequency : only with quota on the root user
1189 Description: mds_obd_create error creating tmp object
1190 Details : When the user sets quota on root, llog will be affected and can't
1191 create files and write files.
1194 Frequency : Always on ia64 patchless client, and possibly others.
1196 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1197 Details : This allows the patchless client to be loaded on architectures
1198 without this export.
1203 Description: disorder of journal start and llog_add cause deadlock.
1204 Details : in llog_origin_connect, journal start should happen before
1205 llog_add keep the same order as other functions to avoid
1208 Severity : enhancement
1210 Description: RedHat Update kernel for RHEL5
1211 Details : Modify the kernel config file more closer RHEL5.
1213 Severity : enhancement
1215 Description: Build failure against Centos5 (RHEL5)
1216 Details : Define PAGE_SIZE when it isn't present.
1219 Frequency : occasionally when using NFS
1221 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1222 Details : with NFS, the anon dentry's parent was set to itself in
1223 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1224 corresponding dentry other than use rec->ur_name.
1226 Severity : enhancement
1228 Description: enable data checksumming by default
1229 Details : enable checksum by default, allow --disable-checksum
1230 configure option and "-o nochecksum" mount option. Checksums
1231 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1234 Frequency : Occasionally with failover
1236 Description: Client eviction due to failover config
1237 Details : after a connection loss, the lustre client should attempt to
1238 reconnect to the last active server first before trying the
1239 other potential connections.
1241 Severity : enhancement
1243 Description: client-side metadata stat-ahead during readdir(directory readahead)
1244 Details : perform client-side metadata stat-ahead when the client detects
1245 readdir and sequential stat of dir entries therein
1250 Description: when mds and osts use different quota unit(32bit and 64bit),
1251 quota will be released repeatly.
1252 Details : void sending multiple quota reqs to mds, which will keep the
1253 status between the reqs.
1256 Frequency : only with liblustre clients on XT3
1258 Description: evictions taking too long
1259 Details : allow llrd to evict clients directly on OSTs
1263 Description: osts not allocated evenly to files
1264 Details : change the condition to increase offset_idx
1266 Severity : enhancement
1268 Description: self-adjustable client's lru lists
1269 Details : use adaptive algorithm for managing client cached locks lru
1270 lists according to current server load, other client's work
1271 pattern, memory activities, etc. Both, server and client
1272 side namespaces provide number of proc tunables for controlling
1277 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
1278 Details : rewrite inherited code (uuid parsing code from ext2 utils and
1279 readdir code from ext3) from scratch preserving functionality.
1283 Description: Only those disconnect error should be returned by rq_status.
1284 Details : In open/enqueue processs, Some errors, which will cause client
1285 disconnected, should be returned by rq_status, while other
1286 errors should still be returned by intent, then mdc or llite will
1289 Severity : enhancement
1291 Description: Tune the kernel for good SCSI performance.
1292 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1293 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1297 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1298 Bugzilla : 13547, 13627
1299 Description: Data corruption for OSTs that are formatted larger than 2TB
1301 Details : When generating the bio request for lustre file writes the
1302 sector number would overflow a temporary variable before being
1303 used for the IO. The data reads correctly from Lustre (which
1304 will overflow in a similar manner) but other file data or
1305 filesystem metadata may be corrupted in some cases.
1309 Description: "lfs find -obd UUID" prints directories
1310 Details : "lfs find -obd UUID" will return all directory names instead
1311 of just file names. It is incorrect because the directories
1312 do not reside on the OSTs.
1315 Bugzilla : 13596, 13614
1316 Description: MDS hang after unclean shutdown of lots of clients
1317 Details : Never resend AST requests.
1321 Frequency : Always, for kernels after 2.6.16
1322 Description: Fix warning idr_remove called for id=.. which is not allocated.
1323 Details : Last kernels save old s_dev before kill super and not allow
1324 to restore from callback - restore it before call kill_anon_super.
1328 Description: Fix error on 'ls .' at the top of the Lustre mount.
1329 Details : Don't revalidate dentry if it is a root dentry.
1331 --------------------------------------------------------------------------------
1333 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
1335 * Support for kernels:
1336 2.6.5-7.286 (SLES 9),
1337 2.6.9-55.0.2.EL (RHEL 4),
1338 2.6.16.46-0.14 (SLES 10),
1339 2.6.18-8.1.8.el5 (RHEL 5),
1340 2.6.18.8 vanilla (kernel.org)
1341 * Client support for unpatched kernels:
1342 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1343 2.6.16 - 2.6.22 vanilla (kernel.org)
1344 * Due to recently discovered recovery problems, we do not recommend
1345 using patchless RHEL 4 clients with this or any earlier release.
1346 * Recommended e2fsprogs version: 1.39.cfs8
1347 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1352 Description: block reactivating mgc import until all deactivates complete
1353 Details : Fix race when failing back MDT/MGS to itself (testing)
1355 Severity : enhancement
1357 Description: lfs setstripe enhancement
1358 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1361 Frequency : mds/oss recovery
1363 Description: llog ctxt is refrenced after it has been freed.
1364 Details : llog ctxt refcount was added to avoide the race between ctxt free
1365 and llog recovery process. Each llog user must hold ctxt refcount
1366 before it access the llog. And the llog ctxt can only be freed
1367 when its refcount is zero.
1369 Severity : enhancement
1371 Description: randomly memory allocation failure util
1372 Details : Make lustre randomly failed allocating memory for testing purpose.
1375 Frequency : only for SLES10
1377 Description: Update kernel patch for SLES10 SP1
1378 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
1380 Severity : enhancement
1382 Description: omit set fsid for export NFS
1383 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
1384 as FS_REQUIRES_DEV for avoid problems with generate fsid.
1387 Frequency : after network failures
1389 Description: Add sync option to mount_lustre.c
1390 Details : Client loses data written to lustre after a network interruption.
1392 Severity : enhancement
1394 Description: Error message improvement.
1395 Details : Merging of two LCONSOLE_ERROR_MSG into one.
1397 Severity : enhancement
1399 Description: don't use GFP_* in generic Lustre code.
1400 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1402 Severity : enhancement
1404 Description: obdclass is limited by single OBD_ALLOC(idarray)
1405 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1407 Severity : enhancement
1409 Description: updated patchess for new RHEL4 kernel
1410 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
1412 Severity : enhancement
1414 Description: warnings with build patchless client with vanila 2.6.19 and up
1415 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1416 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1418 Severity : enhancement
1420 Description: O_DIRECT bypasses client statistics.
1421 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1422 read_ahead_stats were updated. Copied stats section from
1423 osc_send_oap_rpc() into async_internal().
1426 Frequency : only for Cray XT3
1428 Description: peer credits not enough on many OST per OSS systems.
1429 Details : Use new lnet way to add credits as we need those for pings and ASTs
1432 Frequency : only with liblustre
1434 Description: Liblustre is not releasing flock locks on file close.
1435 Details : Release flock locks on file close.
1438 Frequency : only for RHEL4
1440 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
1441 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
1442 Update target file and kernel config.
1446 Description: lustre support for RHEL5
1447 Details : Add support for RHEL5.
1452 Description: ASSERTION(export != NULL) failed in target_handle_connect
1453 Details : Assetion hit is result of rare race between disconnect and connet
1454 to same nid. target_handle_connect found old connect cockie and
1455 tried to reconnect, but can't find export for this cockie.
1460 Description: umount blocks forever on error
1461 Details : In result of wrong using obd_no_recov and obd_force flags client
1462 can hand if cancel or some other requests is lost.
1465 Frequency : Only for SLES
1467 Description: sanity_quota fail test_1
1468 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1469 file, which makes TSTID[2] inunique.
1471 Severity : enhancement
1473 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1474 Details : Update target/ChangeLog/which_patch .
1476 Severity : enhancement
1478 Description: jbd statistics
1479 Details : Port older jbd statistics patch for sles10
1484 Description: lvbo_init failed for resource with missing objects.
1485 Details : Fix returning error if we do stat for file with missing/corrupted
1486 objects and i_size set to all sum of size of all avaible objects.
1487 if we to truncate/write to missing object - it is recreated.
1492 Description: Oops in read and write path when failing to allocate lock.
1493 Details : Check if lock allocation failed and return error back.
1496 Frequency : When flocks are used.
1498 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1499 Details : Flock locks might destroy just granted lock if it could be merged
1500 with another existing flock, this is done in completion handler,
1501 so teach ldlm_cli_enquque_fini that this is a valid case for
1507 Description: reply_lock_interpret crash due to race with it and lock cancel.
1508 Details : Do not replay locks that are being cancelled. Do not reference
1509 locks by their address during replay, just by their handle.
1512 Frequency : only with deactivated OSTs
1514 Description: lstripe command fails for valid OST index
1515 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
1516 instead of lov->desc.ld_active_tgt_count.
1518 Severity : enhancement
1520 Description: Build failure against Centos5 (RHEL5)
1521 Details : Use getpagesize() instead of PAGE_SIZE.
1523 --------------------------------------------------------------------------------
1525 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
1527 * Support for kernels:
1528 2.6.5-7.283 (SLES 9),
1529 2.6.9-55.EL (RHEL 4),
1530 2.6.16.46-0.14 (SLES 10),
1531 2.6.18.8 vanilla (kernel.org)
1532 * Client support for unpatched kernels:
1533 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1534 2.6.16 - 2.6.22 vanilla (kernel.org)
1535 * Due to recently discovered recovery problems, we do not recommend
1536 using patchless RHEL 4 clients with this or any earlier release.
1537 * Recommended e2fsprogs version: 1.39.cfs8
1538 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1539 * Starting with this release, the ldiskfs backing filesystem required
1540 by Lustre is now in its own package, lustre-ldiskfs. This package
1541 should be installed. It is versioned separately from Lustre and
1542 may be released separately in future.
1547 Description: server LBUG when shutting down
1548 Details : Block umount forever until the mount refcount is zero rather
1549 than giving up after an arbitrary timeout.
1551 Severity : enhancement
1553 Description: add optional extra BUILD_VERSION info
1554 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1555 to override the lustre version.
1558 Frequency : 2.6.18 servers only
1560 Description: ll_kern_mount() doesn't release the module reference
1561 Details : The ldiskfs module reference count never drops down to 0
1562 because ll_kern_mount() doesn't release the module reference.
1567 Description: server LBUG when using old ost_num_threads parameter
1568 Details : Accept the old ost_num_threads parameter but warn that it
1569 is deprecated, and fix an off-by-one error that caused an LBUG.
1574 Description: Transient SCSI error results in persistent IO issue
1575 Details : iobuf->dr_error is not reinitialized to 0 between two
1579 Frequency : sometimes when underlying device returns I/O errors
1581 Description: OSTs not going read-only during write failures
1582 Details : OSTs are not remounted read-only when the journal commit threads
1583 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1584 instead of the ext3 wrappers.
1587 Frequency : only with 10000 clients or more
1589 Description: poor connect scaling with increasing client count
1590 Details : Don't run filter_grant_sanity_check for more than 100 exports
1591 to improve scaling for large numbers of clients.
1594 Frequency : SLES10 only
1596 Description: sanity-quota.sh quotacheck failed: rc = -22
1597 Details : Quotas cannot be enabled on SLES10.
1600 Frequency : liblustre clients only
1602 Description: getdirentries does not give error when run on compute nodes
1603 Details : getdirentries does not fail when the size specified as an argument
1604 is too small to contain at least one entry
1606 Severity : enhancement
1608 Description: Add LNET router traceability for debug purposes
1609 Details : If a checksum failure occurs with a router as part of the
1610 IO path, the NID of the last router that forwarded the bulk data
1611 is printed so it can be identified.
1616 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1617 Details : Due to a race condition, liblustre clients were occasionally
1618 evicted incorrectly.
1620 Severity : enhancement
1622 Description: lfs setstripe use optional parameters instead of postional
1625 Severity : enhancement
1627 Description: Nanosecond timestamp support for ldiskfs
1628 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1629 resolution timestamps. There is not yet support for this at
1630 the Lustre filesystem level.
1633 Frequency : during server recovery
1635 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1636 Details : request with rq_no_resend flag not awake l_wait_event if they get a
1640 Frequency : nfs export on patchless client
1642 Description: connectathon hang when test nfs export over patchless client
1643 Details : Disconnected dentry cannot be found with lookup, so we do not need
1644 to unhash it or make it invalid
1647 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1648 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1649 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1654 Description: open req refcounting wrong on reconnect
1655 Details : If reconnect happened between getting open reply from server and
1656 call to mdc_set_replay_data in ll_file_open, we will schedule
1657 replay for unreferenced request that we are about to free.
1658 Subsequent close will crash in variety of ways.
1659 Check that request is still eligible for replay in
1660 mdc_set_replay_data().
1665 Description: disable writes to filesystem when reading health_check file
1666 Details : the default for reading the health_check proc file has changed
1667 to NOT do a journal transaction and write to disk, because this
1668 can cause reads of the /proc file to hang and block HA state
1669 checking on a healthy but otherwise heavily loaded system. It
1670 is possible to return to the previous behaviour during configure
1671 with --enable-health-write.
1673 Severity : enhancement
1675 Description: 64-bit inode version
1676 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1677 to the inode. This will be required for version-based recovery.
1682 Description: MDS fails to start if a duplicate client export is detected
1683 Details : in some rare cases it was possible for a client to connect to
1684 an MDS multiple times. Upon recovery the MDS would detect this
1685 and fail during startup. Handle this more gracefully.
1687 Severity : enhancement
1689 Description: Add -o localflock option to simulate old noflock behaviour.
1690 Details : This will achieve local-only flock/fcntl locks coherentness.
1695 Description: log_commit_thread vs filter_destroy race leads to crash
1696 Details : Take import reference before releasing llog record semaphore
1701 Description: Wrong request locking in request set processing
1702 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1703 imp_delayed_list, in this place should be used imp_lock.
1706 Frequency : when reconnecting
1708 Description: Grant leak when OSC reconnect to OST
1709 Details : When osc reconnect ost, OST(filter) should check whether it
1710 should grant more space to client by comparing fed_grant and
1711 cl_avail_grant, and return the granted space to client instead
1712 of "new granted" space, because client will call osc_init_grant
1713 to update the client grant space info.
1716 Frequency : when client reconnects to OST
1718 Description: Grant leak when OSC does a resend and replays bulk write
1719 Details : When osc reconnect to OST, OST(filter) should clear grant info of
1720 bulk write request, because the grant info while be sync between
1721 OSC and OST when reconnect, and we should ignore the grant info
1722 these of resend/replay write req.
1727 Description: Grant space more than avaiable space sometimes.
1728 Details : When then OST is about to be full, if two bulk writing from
1729 different clients came to OST. Accord the avaliable space of the
1730 OST, the first req should be permitted, and the second one
1731 should be denied by ENOSPC. But if the seconde arrived before
1732 the first one is commited. The OST might wrongly permit second
1733 writing, which will cause grant space > avaiable space.
1736 Frequency : when client is evicted
1738 Description: Grant might be wrongly erased when osc is evicted by OST
1739 Details : when the import is evicted by server, it will fork another
1740 thread ptlrpc_invalidate_import_thread to invalidate the
1741 import, where the grant will be set to 0. While the original
1742 thread will update the grant it got when connecting. So if
1743 the former happened latter, the grant will be wrongly errased
1744 because of this race.
1749 Description: Checking Stale with correct fid
1750 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1751 to check whether it is stale, because sometimes, we want the
1752 enqueue happened anyway, and op_data.fid2 will not be initialized.
1754 Severity : enhancement
1756 Description: update patchless client
1757 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1760 Frequency : only with 2.4 kernel
1762 Description: random memory corruption
1763 Details : size of struct ll_inode_info is to big for union inode.u and this
1764 can be cause of random memory corruption.
1769 Description: Memory leak in recovery
1770 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1771 It should also check obd_fail before fsfilt_start, otherwise if
1772 fsfilt_start return -EROFS,(failover mds during mds recovery).
1773 then the req will return with repmsg->transno = 0 and rc = EROFS.
1774 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1775 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1776 be freed no matter whether fsfilt_commit success or not.
1779 Frequency : only with huge numbers of clients
1781 Description: Prevents from taking the superblock lock in llap_from_page for
1783 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1784 for llap_from_page call in ll_removepage() prevents from taking
1785 the superblock lock for a soon died page.
1790 Description: Not check open intent error before release open handle
1791 Details : in some rare cases, the open intent error is not checked before
1792 release open handle, which may cause
1793 ASSERTION(open_req->rq_transno != 0), because it tries to release
1794 the failed open handle.
1799 Description: Set cat log bitmap only after create log success.
1800 Details : in some rare cases, the cat log bitmap is set too early. and it
1801 should be set only after create log success.
1805 Description: Accessing a block bevice can re-enable I/O when Lustre is
1806 tearing down a device.
1807 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1812 Description: service threads may hog cpus when there are a lot of requests
1813 Details : Insert cond_resched to give other threads a chance to use some CPU
1818 Description: the cat log was not initialized in recovery
1819 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1820 unlink log on mds will not be initialized until mds post
1821 recovery. And also in mds post recovery, the unlink log will
1822 initialization will be done asynchronausly, so there will be race
1823 between add unlink log and unlink log initialization.
1827 Description: brw_stats were being printed incorrectly
1828 Details : brw_stats were being printed as log2 but all of them were not
1829 recorded as log2. Also remove some code duplication arising from
1830 filter_tally_{read,write}.
1834 Frequency : rare, only in recovery.
1835 Description: ASSERTION(req->rq_type != LI_POISON) failed
1836 Details : imp_lock should be held while iterating over imp_sending_list for
1837 prevent destroy request after get timeout in ptlrpc_queue_wait.
1841 Description: replay-single.sh test 52 fails
1842 Details : A lock's skiplist need to be cleanup when it being unlinked
1843 from its resource list.
1848 Description: Short directio read returns full requested size rather than
1850 Details : Direct I/O operations should return actual amount of bytes
1851 transferred rather than requested size.
1853 Severity : enhancement
1855 Description: metadata RPC reduction (e.g. for rm performance)
1856 Details : decrease the amount of synchronous RPC between clients and servers
1857 by canceling conflicing lock before the operation on the client
1858 and packing thier handles into the main operation RPC to server.
1860 Severity : enhancement
1862 Description: add #ifdef HAVE_KERNEL_CONFIG_H
1863 Details : kernels from 2.6.19 not need include linux/config.h, but add
1864 include linux/autoconf.h in commpiler command line.
1866 Severity : enhancement
1868 Description: patchless client support for 2.6.22 kernel
1869 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
1870 removed. In this case we need drop using os depended interface to
1871 kmem_cache and use cfs_mem_cache API.
1876 Description: fix mal-formatted messages
1877 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
1881 Frequency : always in liblustre
1882 Description: wrong IS_ERR implementation in liblustre.h
1883 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
1888 Description: Correct condition for output debug message.
1889 Details : inode i_nlink equal zero is not enough for output message about
1890 disk corruption, i_ctime and i_mode should be also checked.
1894 Frequency : always in patchless client
1895 Description: add configure check for truncate_complete_page
1896 Details : improve checks for exported symbols. This allow run check without
1897 sources, but with Module.symvers shipped with kernel distribution.
1898 add check for truncate_complete_page used by patchless client.
1902 Description: sanity.sh test_77h fails with "test_77h file compare failed"
1903 Details : test_77h uses a file which was messed by other test case.
1907 Description: Not Check whether lov_tgts is NULL in some lov functions
1908 Details : Checking whether lov_tgts is NULL in some functions.
1912 Description: replace obdo_alloc() with OBDO_ALLOC macro
1913 Details : nothing special is done in obdo_alloc() function, and for
1914 debugging purpose, it needs to be replaced with macros.
1918 Description: bad return value and errno from fcntl call
1919 Details : In liblustre API, errno should be a negative value if error
1924 Description: ptlrpc_check_set() LBUG
1925 Details : In case of positive reply from server and failed client bulk
1926 callback after bulk transfer shouldn't LBUG, but process this
1927 request as erroneous.
1929 Severity : enhancement
1931 Description: VFS operations stats tool.
1932 Details : Tool which collects stats by tracking value written in pid,
1933 ppid, gid and uses llstat to generate output to plot graph using
1935 Updated lustre/utils/Makefile.am
1936 Added lustre/utils/ltrack_stats.c
1938 Severity : enhancement
1940 Description: 2.6.18 server support (lustre 1.6.1)
1941 Details : Support for 2.6.18 kernels on the server side.
1946 Description: ASSERTION(imp->imp_conn_current) failed
1947 Details : an assertion failure is hit if a client node boots and attempts to
1948 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
1951 Frequency : only for i686
1953 Description: 1.4.11 RC1 build fails for RHEL 4, i686
1954 Details : Fixed config variable for build.
1959 Description: Updated patchess for new RHEL4 kernel
1960 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
1961 Updated patch jbd-stats-2.6.9.patch
1962 Updated patch qsnet-rhel4-2.6.patch
1963 Updated patch quota-deadlock-on-pagelock-core.patch
1964 Updated patch vfs_intent-2.6-rhel4.patch
1965 Updated patch vfs_races-2.6-rhel4.patch
1966 Updated series file 2.6-rhel4-titech.series
1967 Updated series file 2.6-rhel4.series
1968 Updated kernel config files
1973 Description: lquota slave complains LBUG when reconnecting with mds
1975 Details : quota slave depends qctxt->lqc_import to send its quota request.
1976 This pointer will be invalid if mds did failover or broke its
1977 connect to osts, which leads to LBUG.
1980 Frequency : when qunit size is too small(less than 20M)
1982 Description: write is stopped by improper -EDQUOT
1983 Details : If the master is busy and qunit size is small enough(let's say 1M),
1984 the slave can not get quota from master on time, which will lead
1985 slave to trigger a -EQUOTA to client.
1990 Description: Deadlock during metadata tests
1991 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
1992 called with the per-dentry lock held.
1995 Frequency : SLES9 only
1997 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
1998 Details : set CONFIG_NR_CPUS to 128 instead of 8.
2000 Severity : enhancement
2002 Description: remove fs_prep_san_write operation and related patches
2003 Details : remove the ext3-san-jdike patches which are no longer useful.
2008 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
2009 Details : a disk corruption can cause the mballoc code to assert on a
2010 double free or other extent corruptions. Handle these with
2011 ext3_error() instead of assertions.
2014 Frequency : only with mballoc3 code and deep extent trees
2016 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
2017 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
2018 can cause the filesystem to be remounted read-only.
2023 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
2024 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
2025 few external users of this ioctl number is lfsck's call to
2026 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
2027 Add the old ioctl number to the handler so both old and new
2032 Description: parallel lock callbacks
2033 Details : Instead of sending blocking and completion callbacks as separated
2034 requests, adding them to a set and sending in parallel.
2038 Description: Disable most debugging by default
2039 Details : To improve performance, disable most logging (for debug purposes)
2040 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
2041 default, and HA includes fewer messages.
2045 Frequency : only run on patchless client.
2046 Description: use do_facet on sanity.sh for test handling recoverables errors
2047 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
2051 Description: hash tables for lists of nids, connections and uuids
2052 Details : Hash tables noticeably help when a lot of clients connect to a
2053 server, to faster identify duplicate connections or reconnects,
2054 also to faster find export to evict in manual eviction case.
2058 Description: Sometimes, when the server evict a client, and the client will
2059 not be evicted as soon as possible.
2060 Details : In enqueue req, the error was returned by intent, instead of
2061 rq_status which make ptlrpc layer not detect this error, and
2062 does not evict the client. So enqueue error should be returned
2066 Frequency : only at startup
2068 Description: Delay client connections to MDT until fist MDT->OST connect
2069 Details : If a client tried to create a new file before the MDT had
2070 connected to any OSTs, the create would return EIO. Now
2071 the client will simply block until the MDT connects to the
2072 first OST and the create can succeed.
2075 Frequency : only for SLES9
2077 Description: Routinely utilize latest Quadrics drivers in CFS releases
2078 Details : Update patch qsnet-suse-2.6.patch.
2081 Frequency : only for sles10
2083 Description: Update patches for SLES 10 SP1 kernel.
2084 Details : Update the patch vfs_intent-2.6-sles10.patch.
2089 Description: Routinely utilize latest Quadrics drivers in CFS releases
2090 Details : Update patch qsnet-rhel4-2.6.patch.
2093 Frequency : at statup only
2095 Description: mds_lov_synchronize race leads to various problems
2096 Details : simultaneous MDT->OST connections at startup can cause the
2097 sync to abort, leaving the OSC in a bad state.
2102 Description: Using wrong pointer in osc_brw_prep_request
2103 Details : Access to array[-1] can produce panic if kernel compiled with
2104 CONFIG_PAGE_ALLOC enabled
2106 Severity : enhancement
2108 Description: Async OSC create to avoid the blocking unnecessarily.
2109 Details : If a OST has no remain object, system will block on the creating
2110 when need to create a new object on this OST. Now, ways use
2111 pre-created objects when available, instead of blocking on an
2112 empty osc while others are not empty. If we must block, we block
2113 for the shortest possible period of time.
2117 Frequency : only in recovery
2118 Description: Mark OST as early accessible if his start SYNC.
2119 Details : osc_precreate return flag early accessible if oscc marked as
2120 OSCC_FLAG_SYNC_IN_PROGRESS.
2125 Description: Sometimes precreate code can triger create object on wrong ost
2126 Details : Wrong protected or not not restored variables aftre precreate loop
2127 can produce creation object on wrong ost.
2130 Frequency : oss recovery
2132 Description: llog_commit_thread cleanup should sync with llog_commit_thread
2134 Details : llog_commit_thread_count should be synced between llog_commit
2135 start and cleanup, so new llog_commit thread should not be started
2136 when llog_commit threads being stopped to avoid accessing some
2139 Severity : enhancement
2141 Description: Add printing inode info into message about error in writepage.
2143 --------------------------------------------------------------------------------
2145 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2150 Frequency : on some architectures
2152 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2153 Details : Uninitialized flags sometimes cause configuration commands to
2157 Frequency : patchless clients only
2159 Description: missing __iget() symbol export
2160 Details : The __iget() symbol export is missing. To avoid the need for
2161 this on patchless clients the deathrow inode reaper is turned
2162 off, and we depend on the VM to clean up old inodes. This
2163 dependency was during via the fix for bug 12181.
2168 Description: sanity.sh fail: test_52b
2169 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
2170 incorrect inode's flags to client.
2172 --------------------------------------------------------------------------------
2174 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2176 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2177 INTEROPERATE with older versions automatically. In many cases a
2178 special upgrade step is needed. Please read the
2179 user documentation before upgrading any part of a 1.4.x system.
2180 * WARNING: Lustre configuration and startup changes are required with
2181 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2183 * Support for kernels:
2184 2.4.21-47.0.1.EL (RHEL 3),
2185 2.6.5-7.283 (SLES 9),
2186 2.6.9-42.0.10.EL (RHEL 4),
2187 2.6.12.6 vanilla (kernel.org),
2188 2.6.16.27-0.9 (SLES10)
2189 * Client support for unpatched kernels:
2190 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
2191 2.6.16 - 2.6.19 vanilla (kernel.org),
2192 2.6.9-42.0.8EL (RHEL 4)
2193 * Recommended e2fsprogs version: 1.39.cfs6
2194 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2197 Severity : enhancement
2199 Description: Async OSC create to avoid the blocking unnecessarily.
2200 Details : If a OST has no remain object, system will block on the creating
2201 when need to create a new object on this OST. Now, ways use
2202 pre-created objects when available, instead of blocking on an
2203 empty osc while others are not empty. If we must block, we block
2204 for the shortest possible period of time.
2206 Severity : enhancement
2208 Description: MountConf
2209 Details : Lustre configuration is now managed via mkfs and mount
2210 commands instead of lmc and lconf. New obd types (MGS, MGC)
2211 are added for dynamic configuration management. See
2212 https://mail.clusterfs.com/wikis/lustre/MountConf for
2215 Severity : enhancement
2217 Description: dynamic OST addition
2218 Details : OSTs can now be added to a live filesystem
2220 Severity : enhancement
2222 Description: startup order invariance
2223 Details : MDTs and OSTs can be started in any order. Clients only
2224 require the MDT to complete startup.
2226 Severity : enhancement
2228 Description: parallel, asynchronous orphan cleanup
2229 Details : orphan cleanup is now performed in separate threads for each
2230 OST, allowing parallel non-blocking operation.
2232 Severity : enhancement
2234 Description: optimized stripe assignment
2235 Details : stripe assignments are now made based on ost space available,
2236 ost previous usage, and OSS previous usage, in order to try
2237 to optimize storage space and networking resources.
2239 Severity : enhancement
2241 Description: Permanently set tunables
2242 Details : All writable /proc/fs/lustre tunables can now be permanently
2243 set on a per-server basis, at mkfs time or on a live
2246 Severity : enhancement
2248 Description: Lustre message v2
2249 Details : Add lustre message format v2.
2251 Severity : enhancement
2253 Description: client OST exclusion list
2254 Details : Clients can be started with a list of OSTs that should be
2255 declared "inactive" for known non-responsive OSTs.
2259 Description: ENOENT returned for valid filehandle during dbench.
2260 Details : Check if a directory has children when invalidating dentries
2261 associated with an inode during lock cancellation. This fixes
2262 an incorrect ENOENT sometimes seen for valid filehandles during
2263 testing with dbench.
2266 Frequency : SFS test only (otherwise harmless)
2268 Description: SPEC SFS validation failure on NFS v2 over lustre.
2269 Details : Changes the blocksize for regular files to be 2x RPC size,
2270 and not depend on stripe size.
2272 Severity : enhancement
2274 Description: fine-grained SMP locking inside DLM
2275 Details : Improve DLM performance on SMP systems by removing the single
2276 per-namespace lock and replace it with per-resource locks.
2278 Severity : enhancement
2280 Description: don't hold multiple extent locks at one time
2281 Details : To avoid client eviction during large writes, locks are not
2282 held on multiple stripes at one time or for very large writes.
2283 Otherwise, clients can block waiting for a lock on a failed OST
2284 while holding locks on other OSTs and be evicted.
2286 Severity : enhancement
2288 Description: Multiple MD RPCs in flight.
2289 Details : Further unserialise some read-only MDT RPCs - learn about intents.
2290 To avoid overly-overloading MDT, introduce a limit on number of
2291 MDT RPCs in flight for a single client and add /proc controls
2292 to adjust this limit.
2294 Severity : enhancement
2296 Description: client read/write statistics
2297 Details : Add client read/write call usage stats for performance
2298 analysis of user processes.
2299 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2300 file access. extents_stats shows chunk size distribution.
2301 extents_stats_per_process show chunk size distribution per
2304 Severity : enhancement
2306 Description: per-client statistics on server
2307 Details : Add ldlm and operations statistics for each client in
2308 /proc/fs/lustre/mds|obdfilter/*/exports/
2310 Severity : enhancement
2312 Description: improved MDT statistics
2313 Details : Add detailed MDT operations statistics in
2314 /proc/fs/lustre/mds/*/stats
2316 Severity : enhancement
2318 Description: VFS operations stats
2319 Details : Add client VFS call stats, trackable by pid, ppid, or gid
2320 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
2325 Description: Fix client-side osc byte counters
2326 Details : The osc read/write byte counters in
2327 /proc/fs/lustre/osc/*/stats are now working
2330 Frequency : always as root on SLES
2332 Description: Failure of copying files with lustre special EAs.
2333 Details : Client side always return success for setxattr call for lustre
2334 special xattr (currently only "trusted.lov").
2339 Description: Refcount LNET uuids
2340 Details : The global LNET uuid list grew linearly with every startup;
2341 refcount repeated list entries instead of always adding to
2344 Severity : enhancement
2346 Description: Dynamic service threads
2347 Details : Within a small range, start extra service threads
2348 automatically when the request queue builds up.
2351 Frequency : mixed-endian client/server environments
2353 Description: mixed-endian crashes
2354 Details : The new msg_v2 system had some failures in mixed-endian
2357 Severity : enhancement
2359 Description: Easy OST removal
2360 Details : OSTs can be permanently deactivated with e.g. 'lctl
2361 conf_param lustre-OST0001.osc.active=0'
2363 Severity : enhancement
2365 Description: MGS proc entries
2366 Details : Added basic proc entries for the MGS showing what filesystems
2369 Severity : enhancement
2371 Description: provide MGS failover
2372 Details : Added config lock reacquisition after MGS server failover.
2374 Severity : enhancement
2376 Description: add Linux 2.4 support
2377 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2381 Description: a large application tries to do I/O to the same resource and dies
2382 in the middle of it.
2383 Details : Check the req->rq_arrival time after the call to
2384 ost_brw_lock_get(), but before we do anything about
2385 processing it & sending the BULK transfer request. This
2386 should help move old stale pending locks off the queue as
2387 quickly as obd_timeout.
2390 Frequency : when an incorrect nid is specified during startup
2392 Description: ptlrpc connect to non-existant node causes kernel crash
2393 Details : LNET can't be re-entered from an event callback, which
2394 happened when we expire a message after the export has been
2395 cleaned up. Instead, hand the zombie cleanup off to another
2398 Severity : enhancement
2400 Description: plain/inodebits lock performance improvement
2401 Details : Grouping plain/inodebits in granted list by their request modes
2402 and bits policy, thus improving the performance of search through
2406 Frequency : only if OST filesystem is corrupted
2408 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2409 Details : for a short time RPCs with bulk IO are in the replay list,
2410 but replay of bulk IOs is unimplemented. If the OST filesystem
2411 is corrupted due to disk cache incoherency and then replay is
2412 started it is possible to trip an assertion. Avoid putting
2413 committed RPCs into the replay list at all to avoid this issue.
2416 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2418 Description: System hang on startup
2419 Details : This bug allowed the liblustre (e.g. catamount) client to
2420 return to the app before handling all startup RPCs. This
2421 could leave the node unresponsive to lustre network traffic
2422 and manifested as a server ptllnd timeout.
2424 Severity : enhancement
2426 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2427 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2428 boolean to control whether to print peer debug info when a
2429 client's RPC times out.
2432 Frequency : only for kernels with patches from Lustre below 1.4.3
2434 Description: Remove old rdonly API
2435 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2438 Frequency : only for devices with external journals
2440 Description: Set external device read-only also
2441 Details : During a commanded failover stop, we set the disk device
2442 read-only while the server shuts down. We now also set any
2443 external journal device read-only at the same time.
2446 Frequency : when upgrading from 1.4 while trying to change parameters
2448 Description: The wrong (new) MDC name was used when setting parameters for
2449 upgraded MDT's. Also allows changing of OSC (and MDC)
2450 parameters if --writeconf is specified at tunefs upgrade time.
2453 Frequency : when setting specific ost indicies
2455 Description: QOS code breaks on skipped indicies
2456 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2457 created with --index need not be sequential.
2459 Severity : enhancement
2461 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2462 Details : The uninit_groups feature works in conjunction with the kernel
2463 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2464 up the pass1 processing of e2fsck. This is a read-only feature
2465 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2466 mount filesystems that have had this feature enabled.
2468 Severity : enhancement
2470 Description: Improve multi-block allocation algorithm to avoid fragmentation
2471 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2472 allocation locality and avoid filesystem fragmentation.
2474 ------------------------------------------------------------------------------
2476 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2478 * Support for kernels:
2479 2.4.21-47.0.1.EL (RHEL 3)
2480 2.6.5-7.283 (SLES 9)
2481 2.6.9-42.0.10.EL (RHEL 4)
2482 2.6.12.6 vanilla (kernel.org)
2483 2.6.16.27-0.9 (SLES 10)
2484 * Recommended e2fsprogs version: 1.39.cfs5
2486 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2490 Frequency : occasional, depends on client load and configuration
2491 Bugzilla : 12181, 12203
2492 Description: data loss for recently-modified files
2494 Details : In some cases it is possible that recently written or created
2495 files may not be written to disk in a timely manner (this should
2496 normally be within 30s unless client IO load is very high).
2497 The problem appears as zero-length files or files that are a
2498 multiple of 1MB in size after a client crash or client eviction
2499 that are missing data at the end of the file.
2501 This problem is more likely to be hit on clients where files are
2502 repeatedly created and unlinked in the same directory, clients
2503 have a large amount of RAM, have many CPUs, the filesystem has
2504 many OSTs, the clients are rebooted frequently, and/or the files
2505 are not accessed by other nodes after being written.
2507 The presence of the problem can be detected by looking at
2508 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2509 smaller than the second (nr_unused) then dirty files will not
2510 be flushed automatically to disk. "sync; sleep 10" should be
2511 run several times on the node before unmounting it to update
2512 Lustre (this is also safe to run on nodes without this problem).
2514 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2515 that can cause this same problem, so customers using that kernel
2516 also need to update the kernel in addition to Lustre. In order
2517 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2519 It is normal that files written just before a client crash (less
2520 than 30s) may not yet have been flushed to disk, even for local
2524 Frequency : frequent on thin XT3 nodes
2526 Description: UUID collision on thin XT3 Linux nodes
2527 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2528 randomly, since we relied on an insufficiently-seeded PRNG.
2533 Description: OSS hangs after "All ost request buffers busy"
2534 Details : A deadlock between quota and journal operations caused OSS
2535 hangs after printing "All ost request buffers busy."
2538 Frequency : always on liblustre builds
2540 Description: Cleanup compiler warnings on liblustre
2543 Frequency : always on liblustre builds on XT3
2545 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2549 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2550 > 32000 subdirectories
2551 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2552 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2553 subdirectory count crosses 32000. This will aid e2fsck to
2554 correctly handle more than 32000 subdirectories.
2557 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2559 Description: System hang on startup
2560 Details : This bug allowed the liblustre (e.g. catamount) client to
2561 return to the app before handling all startup RPCs. This
2562 could leave the node unresponsive to lustre network traffic
2563 and manifested as a server ptllnd timeout.
2565 Severity : enhancement
2567 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2568 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2569 boolean to control whether to print peer debug info when a
2570 client's RPC times out.
2575 Description: make O_SYNC working on 2.6 kernels
2576 Details : 2.6 kernels use different method for mark pages for write,
2577 so need add a code to lustre for O_SYNC work.
2582 Description: Failure to close file and release space on NFS
2583 Details : Put inode details into lock acquired in ll_intent_file_open.
2584 Use mdc_intent_lock in ll_intent_open to properly
2585 detect all kind of errors unhandled by mdc_enqueue.
2590 Description: proc file read during shutdown sometimes raced obd removal,
2592 Details : Add lock to prevent obd access after proc file removal.
2595 Frequency : Only for files larger than 4GB on 32-bit clients.
2597 Description: improperly doing page alignment of locks
2598 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2599 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2604 Description: RPCs being resent when they shouldn't be
2605 Details : Some RPCs that should not be resent are being resent. This
2606 can cause inconsistencies in the RPC state machine. Do not
2607 resend such requests.
2610 Frequency : rare, only with NFS export
2612 Description: Crash on NFS re-export node
2613 Details : under very unusual load conditions an assertion is hit in
2614 ll_intent_file_open()
2617 Frequency : only if OST filesystem is corrupted
2619 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2620 Details : for a short time RPCs with bulk IO are in the replay list,
2621 but replay of bulk IOs is unimplemented. If the OST filesystem
2622 is corrupted due to disk cache incoherency and then replay is
2623 started it is possible to trip an assertion. Avoid putting
2624 committed RPCs into the replay list at all to avoid this issue.
2629 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2630 Details : Large single O_DIRECT read and write calls can fail to allocate
2631 a sufficiently large buffer to process the request. In case of
2632 allocation failure the allocation is retried with a smaller
2633 buffer and broken into smaller requests.
2635 Severity : enhancement
2637 Description: Add -o localflock option to simulate old noflock behaviour.
2638 Details : This will achieve local-only flock/fcntl locks coherentness.
2643 Description: versioning check is incomplete
2644 Details : Checking the version difference of client vs. server, report
2645 error if the gap is too big.
2650 Description: add support PG_writeback bit
2651 Details : add support for PG_writeback bit for Lustre, for more carefull
2652 work with page cache in 2.6 kernel. This also fix some deadlocks
2653 and remove hack for work O_SYNC with 2.6 kernel.
2655 Severity : enhancement
2657 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2658 Details : The uninit_groups feature works in conjunction with the kernel
2659 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2660 up the pass1 processing of e2fsck. This is a read-only feature
2661 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2662 mount filesystems that have had this feature enabled.
2664 Severity : enhancement
2666 Description: Improve multi-block allocation algorithm to avoid fragmentation
2667 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2668 allocation locality and avoid filesystem fragmentation.
2670 ------------------------------------------------------------------------------
2672 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2674 * Support for kernels:
2675 2.6.9-42.0.3.EL (RHEL 4)
2676 2.6.5-7.276 (SLES 9)
2677 2.4.21-47.0.1.EL (RHEL 3)
2678 2.6.12.6 vanilla (kernel.org)
2679 2.6.16.21-0.8 (SLES10)
2680 * Recommended e2fsprogs version: 1.39.cfs2-0
2682 * The backwards-compatible /proc/sys/portals symlink has been removed
2683 in this release. Before upgrading, please ensure that you change
2684 any configuration scripts or /etc/sysctl.conf files that access
2685 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2686 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2687 in advance of the upgrade on any system running Lustre 1.4.6 or
2688 newer, since /proc/sys/lnet was added in that version.
2689 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2693 Frequency : only when quota is used
2695 Description: avoid scanning export list for quota master
2696 Details : Change the algorithms to avoid scanning export list in order
2697 to improve the efficiency.
2700 Frequency : MDS failover only, very rarely
2702 Description: "went back in time" messages on mds failover
2703 Details : The greatest transno may be lost when the current operation
2704 finishes with an error (transno==0) and the client's last_rcvd
2705 record is over-written. Save the greatest transno in the
2706 mds_last_transno for this case.
2709 Frequency : always for specific kernels and striping counts
2711 Description: client may get "Matching packet too big" without ACL support
2712 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2713 when trying to access files in certain configurations. The
2714 clients should in fact be denied when mounting because they do
2715 not understand ACLs.
2718 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2720 Description: many clients connecting with IO in progress causes connect timeouts
2721 Details : Avoid synchronous journal commits to avoid delays caused by many
2722 clients connecting/disconnecting when bulk IO is in progress.
2723 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2724 OST_IO_PORTAL to avoid delays behind potentially many pending
2728 Frequency : occasionally with multiple writers to a single file
2730 Description: shared writes to file may result in wrong size reported by stat()
2731 Details : Allow growing of kms when extent lock is cancelled
2734 Frequency : always with random mmap IO to multi-striped file
2736 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2737 Details : Only if the hole is at the end of OST object so that kms is too
2738 small. Fix is to increase kms accordingly in ll_nopage.
2741 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2743 Description: writes to a missing object would leak memory on the OST
2744 Details : If there is an inconsistency between the MDS and OST filesystems,
2745 such that the MDS references an object that doesn't exist, writes
2746 to that object will leak memory due to incorrect cleanup in the
2747 error handling path, eventually running out of memory on the OST.
2752 Description: Creating too long symlink causes lustre errors
2753 Details : Check symlink and name lengths before sending requests to MDS.
2756 Frequency : only if flock is enabled (not on by default)
2758 Description: posix locks not released on fd closure on 2.6.9+
2759 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2760 kernels, this caused such locks not to be released on fd close and
2761 then assertions on fs unmount about still used locks.
2764 Frequency : MDS failover only, very rarely
2766 Description: clients may get ASSERTION(granted_lock != NULL)
2767 Details : When request was taking a long time, and a client was resending
2768 a getattr by name lock request. The were multiple lock requests
2769 with the same client lock handle and
2770 mds_getattr_name->fixup_handle_for_resent_request found one of the
2771 lock handles but later failed with ASSERTION(granted_lock != NULL).
2776 Description: handle->h_buffer_credits > 0, assertion failure
2777 Details : h_buffer_credits is zero after truncate, causing assertion
2778 failure. This patch extends the transaction or creates a new
2782 Frequency : NFS re-export or patchless client
2783 Bugzilla : 11179, 10796
2784 Description: Crash on NFS re-export node (__d_move)
2785 Details : We do not want to hash the dentry if we don't have a lock.
2786 But if this dentry is later used in d_move, we'd hit uninitialised
2787 list head d_hash, so we just do this to init d_hash field but
2788 leave dentry unhashed.
2791 Frequency : NFS re-export or patchless client
2793 Description: NFS exports has problem with symbolic link
2794 Details : lustre client didn't properly install dentry when re-exported
2795 to NFS or running patchless client.
2798 Frequency : NFS re-export or patchless client
2800 Description: Various nfs/patchless fixes.
2801 Details : fixes reuse disconected alias for lookup process - this fixes
2802 warning "find_exported_dentry: npd != pd",
2803 fix permission error with open files at nfs.
2804 fix apply umask when do revalidate.
2807 Frequency : occasional
2809 Description: Crash on NFS re-export node
2810 Details : calling clear_page() on the wrong pointer triggered oops in
2811 generic_mapping_read().
2814 Frequency : rarely, using O_DIRECT IO
2816 Description: unaligned directio crashes client with LASSERT
2817 Details : check for unaligned buffers before trying any requests.
2820 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2822 Description: stale data returned from RAID cache
2823 Details : If only a small amount of IO is done to the RAID device before
2824 reading it again it is possible to get stale data from the RAID
2825 cache instead of reading it from disk.
2828 Frequency : always for sles10 kernel
2830 Description: sles10 support
2831 Details : ll_follow_link: compile fixes and using of nd_set_link
2832 under newer kernels.
2835 Frequency : depends on arch, kernel and compiler version, always on sles10
2838 Description: recursive or deep enough symlinks cause stack overflow
2839 Details : getting rid of large stack-allocated variable in
2843 Frequency : depends on hardware
2845 Description: lustre write performance loss in the SLES10 kernel
2846 Details : the performance loss is caused by using of write barriers in the
2847 ext3 code. The SLES10 kernel turns barrier support on by
2848 default. The fix is to undo that change for ldiskfs.
2850 ------------------------------------------------------------------------------
2852 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2854 * Support for kernels:
2855 2.6.9-42.0.3EL (RHEL 4)
2856 2.6.5-7.276 (SLES 9)
2857 2.4.21-47.0.1.EL (RHEL 3)
2858 2.6.12.6 vanilla (kernel.org)
2862 Frequency : quota enabled and large files being deleted
2864 Description: releasing more than 4GB of quota at once hangs OST
2865 Details : If a user deletes more than 4GB of files on a single OST it
2866 will cause the OST to spin in an infinite loop. Release
2867 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2872 Description: statfs data retrieved from /proc may be stale or zero
2873 Details : When reading per-device statfs data from /proc, in the
2874 {kbytes,files}_{total,free,avail} files, it may appear
2875 as zero or be out of date.
2878 Frequency : systems with MD RAID1 external journal devices
2880 Description: lconf's call to blkid is confused by RAID1 journal devices
2881 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2882 of returning all block devices that match the journal UUID.
2885 Frequency : always, for aggregate stripe size over 4GB
2887 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2888 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2889 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2892 Frequency : always if "lfs find" used on a local file/directory
2894 Description: "lfs find" segfaults if used on a local file/directory
2895 Details : The case where a directory component was not specified wasn't
2896 handled correctly. Handle this properly.
2899 Frequency : always on ppc64
2901 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2902 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2903 perform endianness conversion.
2906 Frequency : rarely (truncate to non-zero file size after write under load)
2907 Bugzilla : 10730, 10687
2908 Description: Files padded with zeros to next 4K multiple
2909 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2910 it is possible that files that are truncated to a non-zero size
2911 immediately after being written are filled with zero bytes beyond
2912 the truncated size. No file data is lost.
2914 Severity : enhancement
2916 Description: Allow recovery/failover for liblustre clients.
2917 Details : liblustre clients were unaware of failover configurations until
2920 Severity : enhancement
2922 Description: user file locks should fail when not mounting with flock option
2923 Details : Set up an error-returning stub in ll_file_operations.lock field
2924 to prevent incorrect behaviour when client is mounted without
2925 flock option. Also, set up properly f_op->flock field for
2929 Frequency : always on ia64
2931 Description: "lfs df" loops on printing out MDS statfs information
2932 Details : The obd_ioctl_data was not initialized and in some systems
2933 this caused a failure during the ioctl that did not return
2934 an error. Initialize the struct and return an error on failure.
2937 Frequency : SLES 9 only
2939 Description: Error of copying files with lustre special EAs as root
2940 Details : Client side always return success for setxattr call for lustre
2941 special xattr (currently only "trusted.lov").
2944 Frequency : rarely on clusters with both ia64+i386 clients
2946 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2947 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2948 clients are doing concurrent writes to the same file it is
2949 possible that the ia64 clients may LASSERT because the OST
2950 extent locks are not PAGE_SIZE aligned. Ensure that grown
2951 locks are always aligned on the request boundary.
2954 Frequency : specific use, occasional
2956 Description: Overwriting in use executable truncates on-disk binary image
2957 Details : If one node attempts to overwrite an executable in use by
2958 another node, we now correctly return ETXTBSY instead of
2959 truncating the file.
2961 Severity : enhancement
2963 Description: Async OSC create to avoid the blocking unnecessarily.
2964 Details : If a OST has no remain object, system will block on the creating
2965 when need to create a new object on this OST. Now, ways use
2966 pre-created objects when available, instead of blocking on an
2967 empty osc while others are not empty. If we must block, we block
2968 for the shortest possible period of time.
2973 Description: chmod on Lustre root is propagated to other clients
2974 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2975 invalid by the follow_mount time.
2978 Frequency : liblustre clients only
2980 Description: Race in 'instant cancel' lock handling could lead to such locks
2981 never to be granted in case of SMP MDS
2982 Details : Do not destroy not yet granted but cbpending locks in
2986 Frequency : replay/resend of open
2988 Description: non null lock assetion failure in mds_intent_policy
2989 Details : Trying to replay/resend lockless open requests resulted in
2990 mds_open() returning 0 with no lock. Now it sets a flag if
2991 a lock is going to be returned.
2993 Severity : enhancement
2995 Description: Checksum enhancements
2996 Details : New checksum enhancements allow for resending RPCs that failed
2999 Severity : enhancement
3001 Description: Tunables on number of dirty pages in cacche
3002 Details : Allow to set limit on number of dirty pages cached.
3007 Description: client crash on unmount - lock still has references
3008 Details : In some error handling cases it was possible to leak a lock
3009 reference on a client while accessing a file. This was not
3010 harmful to the client during operation, but would cause the
3011 client to crash when the filesystem is unmounted.
3014 Frequency : specific case, rare
3016 Description: ETXTBSY on mds though file not in use by client
3017 Details : ETXTBSY is no longer incorrectly returned when attempting to
3018 chmod or chown a directory that the user previously tried to
3019 execute or a currently-executing binary.
3022 Frequency : extremely rare except on liblustre-based clients
3024 Description: Lustre space not freed when files are deleted
3025 Details : Clean up open-unlinked files after client eviction. Previously
3026 the unlink was skipped and the files remained as orphans.
3031 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3032 Details : In some cases it was possible to send a blocking callback to
3033 a client doing a glimpse, even though that client didn't get
3034 a lock granted. When the glimpse lock is cancelled on the OST
3035 the freed lock is left on the waiting list and corrupted the list.
3038 Frequency : all core dumps
3040 Description: Broke core dumps to lustre
3041 Details : Negative dentry may be unhashed if parent does not have UPDATE
3042 lock, but some callers, e.g. do_coredump, expect dentry to be
3043 hashed after successful create, hash it in ll_create_it.
3045 ------------------------------------------------------------------------------
3047 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3049 * Support for kernels:
3050 2.6.9-42.0.2.EL (RHEL 4)
3051 2.6.5-7.276 (SLES 9)
3052 2.4.21-40.EL (RHEL 3)
3053 2.6.12.6 vanilla (kernel.org)
3057 Frequency : always on RHEL 3
3059 Description: Number of open files grows over time
3060 Details : The number of open files grows over time, whether or not
3061 Lustre is started. This was due to a filp leak introduced
3062 by one of our kernel patches.
3064 ------------------------------------------------------------------------------
3066 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
3068 * Support for kernels:
3069 2.6.9-42.EL (RHEL 4)
3070 2.6.5-7.267 (SLES 9)
3071 2.4.21-40.EL (RHEL 3)
3072 2.6.12.6 vanilla (kernel.org)
3077 Bugzilla : 5719, 9635, 9792, 9684
3078 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3079 Details : If a server is under heavy load and cannot reply to new
3080 connection requests before the client resends the (re)connect,
3081 the connection handling code can behave badly if two service
3082 threads are concurrently handing separate (re)connections from
3083 the same client. Add better locking to the connection handling
3084 code, and ensure that only a single connection will be processed
3085 for a given client UUID, even if the lock is dropped.
3087 Severity : enhancement
3089 Description: add TCP zero-copy support to kernel
3090 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3091 sends if the hardware supports scatter-gather and checksumming.
3092 This allows socklnd to do client-write and server-read more
3093 efficiently and reduce CPU utilization from skbuf copying.
3096 Frequency : only if NFS exporting from client
3098 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3099 Details : If setting attributes on a file created under NFS that had
3100 never been opened it would be possible to oops the client
3101 if the file had no objects.
3105 Bugzilla : 9326, 10402, 10897
3106 Description: client crash in ptlrpcd_wake() thread when sending async RPC
3107 Details : It is possible that ptlrpcd_wake() dereferences a freed async
3108 RPC. In rare cases the ptlrpcd thread alread processed the RPC
3109 before ptlrpcd_wake() was called and the request was freed.
3112 Frequency : always for liblustre
3114 Description: liblustre client does MDS+OSTs setattr RPC for each write
3115 Details : When doing a write from a liblustre client, the client
3116 incorrectly issued an RPC to the MDS and each OST the file was
3117 striped over in order to update the timestamps. When writing
3118 with small chunks and many clients this could overwhelm the MDS
3119 with RPCs. In all cases it would slow down the write because
3120 these RPCs are unnecessary.
3122 Severity : enhancement
3124 Description: allow number of MDS service threads to be changed at module load
3125 Details : It is now possible to change the number of MDS service threads
3126 running. Adding "options mds mds_num_threads={N}" to the MDS's
3127 /etc/modprobe.conf will set the number of threads for the next
3128 time Lustre is restarted (assuming the "mds" module is also
3129 reloaded at that time). The default number of threads will
3130 stay the same, 32 for most systems.
3135 Description: OST crash if filesystem is unformatted or corrupt
3136 Details : If an OST is started on a device that has never been formatted
3137 or if the filesystem is corrupt and cannot even mount then the
3138 error handling cleanup routines would dereference a NULL pointer.
3143 Description: NULL pointer deref in llap_from_page.
3144 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3145 incorrect) mapping if the page was truncated/reclaimed while it was
3146 searched for. Check for this condition and skip such pages when
3147 doing readahead. Introduce extra check to llap_from_page() to
3148 verify page->mapping->host is non-NULL (so page is not anonymous).
3151 Frequency : Sometimes when using sys_sendfile
3153 Description: "page not covered by a lock" warnings from ll_readpage
3154 Details : sendfile called ll_readpage without right page locks present.
3155 Now we introduced ll_file_sendfile that does necessary locking
3156 around call to generic_file_sendfile() much like we do in
3160 Frequency : with certain MDS communication failures at client mount time
3162 Description: NULL pointer deref after failed client mount
3163 Details : a client connection request may delayed by the network layer
3164 and not be sent until after the PTLRPC layer has timed out the
3165 request. If the client fails the mount immediately it will try
3166 to clean up before the network times out the request. Add a
3167 reference from the request import to the obd device and delay
3168 the cleanup until the network drops the request.
3171 Frequency : occasionally during client (re)connect
3173 Description: assertion failure during client (re)connect
3174 Details : processing a client connection request may be delayed by the
3175 client or server longer than the client connect timeout. This
3176 causes the client to resend the connection request. If the
3177 original connection request is replied in this interval, the
3178 client may trip an assertion failure in ptlrpc_connect_interpret()
3179 which thought it would be the only running connect process.
3182 Frequency : only with obd_echo servers and clients that are rebooted
3184 Description: kernel BUG accessing uninitialized data structure
3185 Details : When running an obd_echo server it did not start the ping_evictor
3186 thread, and when a client was evicted an uninitialized data
3187 structure was accessed. Start the ping_evictor in the RPC
3188 service startup instead of the OBD startup.
3190 Severity : enhancement
3191 Bugzilla : 10193 (patchless)
3192 Description: Remove dependency on various unexported kernel interfaces.
3193 Details : No longer need reparent_to_init, exit_mm, exit_files,
3194 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3197 Frequency : rare (only users of deprecated and unsupported LDAP config)
3199 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3200 Details : LDAP apparently contains 'lustreName' attributes instead of
3201 'name'. A simple remapping of the name is sufficient.
3204 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3206 Description: waiting_locks_callback trips kernel BUG if client is evicted
3207 Details : Running with the dump_on_timeout debug flag turned on makes
3208 it possible that the waiting_locks_callback() can try to dump
3209 the Lustre kernel debug logs from an interrupt handler. Defer
3210 this log dumping to the expired_lock_main() thread.
3212 Severity : enhancement
3214 Description: Support NFS exporting on 2.6 kernels.
3215 Details : Implement non-rawops metadata methods for NFS server to use without
3216 changing NFS server code.
3219 Frequency : very rare (synthetic metadata workload only)
3221 Description: two racing renames might cause an MDS thread to deadlock
3222 Details : Running the "racer" program may cause one MDS thread to rename
3223 a file from being the source of a rename to being the target of
3224 a rename at exactly the same time that another thread is doing
3225 so, and the second thread has already enqueued these locks after
3226 doing a lookup of the target and is trying to relock them in
3227 order. Ensure that we don't try to re-lock the same resource.
3230 Frequency : only very large systems with liblustre clients
3232 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3233 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3234 all OSTs in parallel. This allows the eviction of stale liblustre
3235 clients to proceed much faster than if they were done in series,
3236 and also offers similar improvements for other set_info RPCs.
3241 Description: excessive CPU usage during initial read phase on client
3242 Details : During the initial read phase on a client, it would agressively
3243 retry readahead on the file, consuming too much CPU and impacting
3244 performance (since 1.4.5.8). Improve the readahead algorithm
3245 to avoid this, and also improve some other common cases (read
3246 of small files in particular, where "small" is files smaller than
3247 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3252 Description: MDS crash when receiving packet with unknown intent.
3253 Details : Do not LBUG in unknown intent case, just return -EFAULT
3255 Severity : enhancement
3256 Bugzilla : 9293, 9385
3257 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3258 Details : Do not serialize getattr (non-intent version) and statfs.
3261 Frequency : occasional, when OST network is overloaded/intermittent
3263 Description: client evicted by OST after bulk IO timeout
3264 Details : If a client sends a bulk IO request (read or write) the OST
3265 may evict the client if it is unresposive to its data GET/PUT
3266 request. This is incorrect if the network is overloaded (takes
3267 too long to transfer the RPC data) or dropped the OST GET/PUT
3268 request. There is no need to evict the client at all, since
3269 the pinger and/or lock callbacks will handle this, and the
3270 client can restart the bulk request.
3273 Frequency : Always when mmapping file with no objects
3275 Description: client crashes when mmapping file with no objects
3276 Details : Check that we actually have objects in a file before doing any
3277 operations on objects in ll_vm_open, ll_vm_close and
3283 Description: Request leak when working with deleted CWD
3284 Details : Introduce advanced request refcount tracking for requests
3285 referenced from lustre intent.
3287 Severity : Enhancement
3289 Description: Cache open file handles on client.
3290 Details : MDS now will return special lock along with openhandle, if
3291 requested and client is allowed to hold openhandle, even if unused,
3292 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3293 closing files for every read/write openration.
3295 Severity : Enhancement
3297 Description: Cache open negative dentries on client when possible.
3298 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3299 negative dentries on lock revocation.
3304 Description: Remounting a client read-only wasn't possible with a zconf mount
3305 Details : It wasn't possible to remount a client read-only with llmount.
3307 Severity : enhancement
3308 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3309 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3310 MPICH is included to add Lustre-specific ADIO interfaces.
3311 This is based closely on the UFS ADIO layer and only differs
3312 in file creation, in order to allow the OST striping to be set.
3313 This is user-contributed code and not supported by CFS.
3318 Description: extended inode attributes (immutable, append-only) work improperly
3319 when 2.4 and 2.6 kernels are used on client/server or vice versa
3320 Details : Introduce kernel-independent values for these flags.
3322 Severity : enhancement
3325 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3326 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3327 granularity. Now, Fractional MB granularity is be supported,
3328 this is very useful for low memory system.
3330 Severity : enhancement
3332 Description: Getattr by fid
3333 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3334 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3335 have locks and avoids getattr rpc requests for it on every stat.
3338 Frequency : Always, for filesystems larger than 2TB
3340 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3341 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3342 filesystems larger than 2TB to be mounted with mballoc enabled.
3345 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3347 Description: filesystem corruption for non-standard kernels and very large OSTs
3348 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3349 filesystems larger than 2TB is mounted then the kernel will
3350 silently corrupt the start of the filesystem. CONFIG_LBD is
3351 enabled for all CFS-supported kernels, but the possibility of
3352 this happening with a modified kernel config exists.
3354 Severity : enhancement
3356 Description: add client O_DIRECT support for 2.6 kernels
3357 Details : It is now possible to do O_DIRECT reads and writes to files
3358 in the Lustre client mountpoint on 2.6 kernel clients.
3360 Severity : enhancement
3362 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3363 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3364 parallel, not waiting for response from every OST before sending
3365 a rpc to the next OST.
3370 Description: setattr vs write race when updating file timestamps
3371 Details : Client processes that update a file timestamp into the past
3372 right after writing to the file (e.g. tar) it is possible that
3373 the updated file modification time can be reset to the current
3374 time due to a race between processing the setattr and write RPC.
3376 Severity : enhancement
3378 Description: Bring 'lfs find' closer in line with regular Linux find.
3379 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3380 -print0 options and obtains all the needed info through the lustre
3383 Severity : enhancement
3385 Description: support up to 1024 configured devices on one node
3386 Details : change obd_dev array from statically allocated to dynamically
3387 allocated structs as they are first used to reduce memory usage
3392 Description: Flush dirty partially truncated pages during truncate
3393 Details : Immediatelly flush partially truncated pages in filter_setattr,
3394 this way we completely avoid having any pages in page cache on OST
3395 and can retire ugly workarounds during writes to flush such pages.
3400 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3401 Details : protect inode from truncation within vfs_unlink() context
3402 just take a reference before calling vfs_unlink() and release it
3403 when parent's i_sem is free.
3406 Frequency : always, if extents are used on OSTs
3408 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3409 Details : OSTs using the extents format would not zero the high 16 bits of
3410 the index physical block number. This is not a problem for any
3411 OST filesystems smaller than 16TB, and no kernels support ext3
3412 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3413 new/modified files) and can be fixed for existing filesystems
3414 with e2fsprogs-1.39-cfs1.
3419 Description: import connection selection may be incorrect if timer wraps
3420 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3421 import connections to be ignored if the 32-bit jiffies counter
3422 wraps. Use a 64-bit jiffies counter.
3425 Frequency : during server recovery
3427 Description: crash after server is denying duplicate export
3428 Details : If clients are resending connect requests to the server, the
3429 server refuses to allow a client to connect multiple times.
3430 Fixed a bug in the handling of this case.
3433 Frequency : very large clusters immediately after boot
3435 Description: LNET request buffers exhausted under heavy short-term load
3436 Details : If a large number of client requests are generated on a service
3437 that has previously never seen so many requests it is possible
3438 that the request buffer growth cannot keep up with the spike in
3439 demand. Instead of dropping incoming requests, they are held in
3440 the LND until the RPC service can accept more requests.
3443 Frequency : Sometimes during replay
3445 Description: Assertion failure in ll_local_open after replay.
3446 Details : If replay happened on an open request reply before we were able
3447 to set replay handler, reply will become not swabbed tripping the
3448 assertion in ll_local_open. Now we set the handler right after
3449 recognising of open request
3452 Frequency : very rare
3454 Description: kernel reports "badness in vsnprintf"
3455 Details : Reading from the "recovery_status" /proc file in small chunks
3456 may cause a negative length in lprocfs_obd_rd_recovery_status()
3457 call to vsnprintf() (which is otherwise harmless). Exit early
3458 if there is no more space in the output buffer.
3460 Severity : enhancement
3462 Description: clear OBD RPC statistics by writing to them
3463 Details : It is now possible to clear the OBD RPC statistics by writing
3464 to the "stats" file.
3469 Description: Client mtime is not the same on different clients after utimes
3470 Details : In some cases, the client was using the utimes() syscall on
3471 a file cached on another node. The clients now validate the
3472 ctime from the MDS + OSTs to determine which one is right.
3477 Description: Inability to activate failout mode
3478 Details : lconf script incorrectly assumed that in python string's numeric
3479 value is used in comparisons.
3482 Frequency : always with multiple stripes per file
3484 Description: Inefficient object allocation for mutli-stripe files
3485 Details : When selecting which OSTs to stripe files over, for files with
3486 a stripe count that divides evenly into the number of OSTs,
3487 the MDS is always picking the same starting OST for each file.
3488 Return the OST selection heuristic to the original design.
3493 Description: mount failures may take full timeout to return an error
3494 Details : Under some heavy load conditions it is possible that a
3495 failed mount can wait for the full obd_timeout interval,
3496 possibly several minutes, before reporting an error.
3497 Instead return an error as soon as the status is known.
3499 ------------------------------------------------------------------------------
3501 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
3503 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3504 INTEROPERATE with older versions automatically. Please read the
3505 user documentation before upgrading any part of a live system.
3506 * WARNING: Lustre networking configuration changes are required with
3507 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3510 * Support for kernels:
3511 2.6.9-22.0.2.EL (RHEL 4)
3512 2.6.5-7.244 (SLES 9)
3513 2.6.12.6 vanilla (kernel.org)
3516 Severity : enhancement
3517 Bugzilla : 7981/8208
3518 Description: Introduced Lustre Networking (LNET)
3519 Details : LNET is new networking infrastructure for Lustre, it includes
3520 a reorganized network configuration mode (see the user
3521 documentation for full details) as well as support for routing
3522 between different network fabrics. Lustre Networking Devices
3523 (LNDs) for the supported network fabrics have also been
3524 created for this new infrastructure.
3526 Severity : enhancement
3527 Description: Introduced Access control lists
3528 Details : clients can set ACLs on files and directories in order to have
3529 more fine-grained permissions than the standard Unix UGO+RWX.
3530 The MDS must be started with the "-o acl" mount option.
3532 Severity : enhancement
3533 Description: Introduced filesystem quotas
3534 Details : Administrators may now establish per-user quotas on the
3537 Severity : enhancement
3539 Description: Configuration change for the XT3
3540 The PTLLND is now used to run Lustre over Portals on the XT3
3541 The configure option(s) --with-cray-portals are no longer used.
3542 Rather --with-portals=<path-to-portals-includes> is used to
3543 enable building on the XT3. In addition to enable XT3 specific
3544 features the option --enable-cray-xt3 must be used.
3549 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3550 Details : A race between allocating a new llcd and re-getting the llcd_lock
3551 allowed another thread to grab newly-allocated llcd.
3553 Severity : enhancement
3555 Description: 2.6 OST async journal commit and locking fix to improve performance
3556 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3557 2.6 kernels are now async as they already were in 2.4 kernels so
3558 that they can commit concurrently with the network bulk transfer.
3559 For block-allocated files the filter allocation semaphore is held
3560 to avoid filesystem fragmentation during allocation. BKL lock
3561 removed for 2.6 xattr operations where it is no longer needed.
3566 Description: lconf incorrectly determined whether two IP networks could talk
3567 Details : In some more complicated routing and multiple-network
3568 configurations, lconf will avoid trying to make a network
3569 connection to a disjoint part of the IP space. It was doing the
3570 math incorrectly for one set of cases.
3575 Description: Fix for potential infinite loop processing records in an llog.
3576 Details : If an llog record is corrupted/zeroed, it is possible to loop
3577 forever in llog_process(). Validate the llog record length
3578 and skip the remainder of the block on error.
3581 Frequency : occasional (liblustre only)
3583 Description: liblustre could not open files whose last component is a symlink
3584 Details : sysio_path_walk() would incorrectly pass the open intent to
3585 intermediate path components.
3588 Frequency : rare (liblustre only with non-standard tuning)
3589 Bugzilla : 7201 (7350)
3590 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3591 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3592 after it has been released, only do this for referenced locks
3594 Severity : enhancement
3596 Description: specifying an (invalid) directory default stripe_size of -1
3597 would reset the directory default striping
3598 Details : stripe_size -1 was used internally to signal directory stripe
3599 removal, now use "all default" to signal dir stripe removal
3600 as a directory striping of "all default" is not useful
3603 Frequency : common for large clusters running liblustre clients
3605 Description: doing an ls when liblustre clients are running is slow
3606 Details : sending a glimpse AST to a liblustre client waits for every AST
3607 to time out, as liblustre clients will not respond. Since they
3608 cannot cache data we refresh the OST lock LVB from disk instead.
3610 Severity : enhancement
3612 Description: doing an ls at the same time as file IO can be slow
3613 Details : enqueue and other "small" requests can be blocked behind many
3614 large IO requests. Create a new OST IO portal for non-IO
3615 requests so they can be processed faster.
3618 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3620 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3621 created file with mode 000
3622 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3623 allow an ftruncate() as a non-root user to a file with mode 000.
3624 Lustre now respects this flag to disable mode checks when
3625 truncating a file owned by the user
3628 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3631 Description: Revoking locks from clients that went dead or catatonic might take
3633 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3634 cancellation of such locks instant on servers without waiting for
3635 any reply from clients. Clients drops these locks when cancel
3636 notification from server is received without replying.
3639 Frequency : liblustre-only, when liblustre client dies or becomes busy
3641 Description: Doing ls on Linux clients can take a long time with active
3643 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3644 granting such locks to it in the first place if possible. Locks
3645 are taken by proxy on the OST during the read or write and
3646 dropped immediately afterward. Add connect flags handling, do
3647 not grant locks to liblustre clients for glimpse ASTs.
3649 Severity : enhancement
3651 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3652 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3653 sequential reads and behaves badly with applications doing random
3654 reads. Improve it by reading ahead at least read region, and
3655 avoiding excessive large RPC for small reads.
3657 Severity : enhancement
3659 Description: Creating more than 1000 files for a single job may cause a load
3660 imbalance on the OSTs if there are also a large number of OSTs.
3661 Details : qos_prep_create() uses an OST index reseed value that is an
3662 even multiple of the number of available OSTs so that if the
3663 reseed happens in the middle of the object allocation it will
3664 still utilize the OSTs as uniformly as possible.
3669 Description: OST or MDS may oops in ping_evictor_main()
3670 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3671 but doesn't restart at beginning of obd_exports_timed list
3674 Severity : enhancement
3676 Description: improve by-nid export eviction on the MDS and OST
3677 Details : allow multiple exports with the same NID to be evicted at one
3678 time without re-searching the exports list.
3681 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3683 Description: MDS may oops in groups_free()
3684 Details : in rare race conditions a newly allocated group_info struct is
3685 freed again, and this can be NULL. The 2.4 compatibility code
3686 for groups_free() checked for a NULL pointer, but 2.6 did not.
3689 Frequency : common for liblustre clients doing little filesystem IO
3690 Bugzilla : 9352, 7313
3691 Description: server may evict liblustre clients accessing contended locks
3692 Details : if a client is granted a lock or receives a completion AST
3693 with a blocking AST already set it would not reply to the AST
3694 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3697 Frequency : lfs setstripe, only systems with more than 160 OSTs
3699 Description: unable to set striping with a starting offset beyond OST 160
3700 Details : llapi_create_file() incorrectly limited the starting stripe
3701 index to the maximum single-file stripe count.
3704 Frequency : LDAP users only
3706 Description: lconf did not handle in-kernel recovery with LDAP properly
3707 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3709 Severity : enhancement
3711 Description: bind OST threads to NUMA nodes to improve performance
3712 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3713 node and do their allocations there to localize memory access
3715 Severity : enhancement
3717 Description: llmount can determine client NID directly from Myrinet (GM)
3718 Details : the client NID code from gmnalnid was moved directly into
3719 llmount, removing the need to use this or specifying the
3720 client NID explicitly when mounting GM clients with zeroconf
3723 Frequency : if client is started with down MDS
3725 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3726 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3727 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3728 ptlrpc_import_delayed_req() because we didn't check if the
3729 request was interrupted, and we also didn't break out of the
3730 event loop if there was no timeout
3735 Description: data loss during non-page-aligned writes to a single file from
3736 both multiple nodes and multiple threads on one node at same time
3737 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3738 inconsistency led to false short-reads, that were cached and later
3739 used by ->prepare_write() to fill in partially written page,
3740 leading to data loss.
3743 Frequency : always, if lconf --abort_recovery used
3745 Description: lconf --abort_recovery fails with 'Operation not supported'
3746 Details : lconf was attempting to abort recovery on the MDT device and not
3749 Severity : enhancement
3751 Description: remove cleanup logs
3752 Details : replace lconf-generated cleanup logs with lustre internal
3753 cleanup routines. Eliminates the need for client-cleanup and
3756 Severity : enhancement
3758 Description: add support for EAs (user and system) on lustre filesystems
3759 Details : it is now possible to store extended attributes in the Lustre
3760 client filesystem, and with the user_xattr mount option it
3761 is possible to allow users to store EAs on their files also
3763 Severity : enhancement
3765 Description: Add possibility (config option) to show minimal available OST free
3767 Details : When compiled with --enable-mindf configure option, statfs(2)
3768 (and so, df) will return least minimal free space available from
3769 all OSTs as amount of free space on FS, instead of summary of
3770 free spaces of all OSTs.
3772 Severity : enhancement
3774 Description: do not expand extent locks acquired on OST-side
3775 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3776 by server: they are not cached anyway.
3779 Frequency : when mmap is used/binaries executed from Lustre
3781 Description: Unmmap pages before throwing them away from read cache.
3782 Details : llap_shrink cache now attempts to unmap pages before discarding
3783 them (if unmapping failed - do not discard). SLES9 kernel has
3784 extra checks that trigger if this unmapping is not done first.
3789 Description: lconf didn't resolve symlinks before checking to see whether a
3790 given mountpoint was already in use
3793 Frequency : when migrating failover services
3794 Bugzilla : 6395, 9514
3795 Description: When migrating a subset of services from a node (e.g. failback
3796 from a failover service node) the remaining services would
3797 time out and evict clients.
3798 Details : lconf --force (implied by --failover) sets the global obd_timeout
3799 to 5 seconds in order to quickly disconnect, but this caused
3800 other RPCs to time out too quickly. Do not change the global
3801 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3803 Severity : enhancement
3804 Frequency : if MDS is started with down OST
3805 Bugzilla : 9439,5706
3806 Description: Allow startup/shutdown of an MDS without depending on the
3807 availability of the OSTs.
3808 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3809 Add appropriate locking and lov-osc refcounts for safe
3810 cleaning. Add osc abort_inflight calls in case the
3811 synchronize never started.
3814 Frequency : occasional (Cray XT3 only)
3816 Description: root not authorized to access files in CRAY_PORTALS environment
3817 Details : The client process capabilities were not honoured on the MDS in
3818 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3819 been accessed by an authorized user then root was able to access
3820 the file on the local client also. The root user capabilities
3821 are now allowed on the MDS, as this environment has secure UID.
3824 Frequency : occasional
3826 Description: ldiskfs "too long searching" message happens too often
3827 Details : A debugging message (otherwise harmless) prints too often on
3828 the OST console. This has been reduced to only happen when
3829 there are fragmentation problems on the filesystem.
3834 Description: Division by zero in statfs when all OSCs are inactive
3835 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3836 lov_statfs divides by value returned by lov_get_stripecnt().
3840 Bugzilla : 9489, 3273
3841 Description: First write from each client to each OST was only 4kB in size,
3842 to initialize client writeback cache, which caused sub-optimal
3843 RPCs and poor layout on disk for the first writen file.
3844 Details : Clients now request an initial cache grant at (re)connect time
3845 and so that they can start streaming writes to the cache right
3846 away and always do full-sized RPCs if there is enough data.
3847 If the OST is rebooted the client also re-establishes its grant
3848 so that client cached writes will be honoured under the grant.
3853 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3854 Details : Now I/O RPCs go to different portal number and (presumably) fast
3855 lock requests (and glimses) and other RPCs get their own service
3856 threads pool that should be able to service those RPCs
3859 Severity : enhancement
3861 Description: Ability to exchange lustre version between client and servers and
3862 issue warnings at client side if client is too old. Also for
3863 liblustre clients there is ability to refuse connection of too old
3865 Details : New 'version' field is added to connect data structure that is
3866 filled with version info. That info is later checked by server and
3870 Frequency : rare, liblustre only.
3871 Bugzilla : 9296, 9581
3872 Description: Two simultaneous writes from liblustre at offset within same page
3873 might proceed at the same time overwriting eachother with stale
3875 Details : I/O lock withing llu_file_prwv was released too early, before data
3876 actually was hitting the wire. Extended lock-holding time until
3877 server acknowledges receiving data.
3880 Frequency : extremely rare. Never observed in practice.
3882 Description: avoid generating lustre_handle cookie of 0.
3883 Details : class_handle_hash() generates handle cookies by incrementing
3884 global counter, and can hit 0 occasionaly (this is unlikely, but
3885 not impossible, because initial value of cookie counter is
3886 selected randonly). Value of 0 is used as a sentinel meaning
3887 "unassigned handle" --- avoid it. Also coalesce two critical
3888 sections in this function into one.
3890 Severity : enhancement
3892 Description: allow liblustre clients to delegate truncate locking to OST
3893 Details : To avoid overhead of locking, liblustre client instructs OST to
3894 take extent lock in ost_punch() on client's behalf. New connection
3895 flag is added to handle backward compatibility.
3897 Severity : enhancement
3898 Bugzilla : 4928, 7341, 9758
3899 Description: allow number of OST service threads to be specified
3900 Details : a module parameter allows the number of OST service threads
3901 to be specified via "options ost ost_num_threads={N}" in the
3902 OSS's /etc/modules.conf or /etc/modprobe.conf.
3906 Bugzilla : 6146, 9635, 9895
3907 Description: servers crash with bad pointer in target_handle_connect()
3908 Details : In rare cases when a client is reconnecting it was possible that
3909 the connection request was the last reference for that export.
3910 We would temporarily drop the export reference and get a new
3911 one, but this may have been the last reference and the export
3912 was just destroyed. Get new reference before dropping old one.
3914 Severity : enhancement
3915 Frequency : if client is started with failover MDS
3917 Description: Allow multiple MDS hostnames in the mount command
3918 Details : Try to read the configuration from all specified MDS
3919 hostnames during a client mount in case the "primary"
3922 Severity : enhancement
3924 Description: Stop sending data to evicted clients as soon as possible.
3925 Details : Check if the client we are about to send or are sending data to
3926 was evicted already. (Check is done every second of waiting,
3927 for which l_wait_event interface was extended to allow checking
3928 of exit condition at specified intervals).
3931 Frequency : rare, normally only when NFS exporting is done from client
3933 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3935 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3938 Frequency : always, if srand() is called before liblustre initialization
3940 Description: Liblustre uses system PRNG disturbing its usage by user application
3941 Details : Introduce internal to lustre fast and high-quality PRNG for
3942 lustre usage and make liblustre and some other places in generic
3943 lustre code to use it.
3945 Severity : enhancement
3946 Bugzilla : 9477, 9557, 9870
3947 Description: Verify that the MDS configuration logs are updated when xml is
3948 Details : Check if the .xml configuration logs are newer than the config
3949 logs stored on the MDS and report an error if this is the case.
3950 Request --write-conf, or allow starting with --old_conf.
3952 Severity : enhancement
3954 Description: Handle symlinks in the path when checking if Lustre is mounted.
3955 Details : Resolve intermediate symlinks when checking if a client has
3956 mounted a filesystem to avoid duplicate client mounts.
3961 Description: lconf can hit an error exception but still return success.
3962 Details : The lconf command catches the Command error exception at the top
3963 level script context and will exit with the associated exit
3964 status, but doesn't ensure that this exit status is non-zero.
3969 Description: failure of ptlrpc thread startup can cause oops
3970 Details : Starting a ptlrpc service thread can fail if there are a large
3971 number of threads or the server memory is very fragmented.
3972 Handle this without oopsing.
3975 Frequency : always, only if liblustre and non-default acceptor port was used
3977 Description: liblustre cannot connect to servers with non-default acceptor port
3978 Details : tcpnal_set_default_params() was not called and was therefore
3979 ignoring the environment varaible TCPNAL_PORT, as well as other
3980 TCPNAL_ environment variables
3985 Description: two objects could be created on the same OST for a single file
3986 Details : If an OST is down, in some cases it was possible to create two
3987 objects on a single OST for a single file. No problems other
3988 than potential performance impact and spurious error messages.
3992 Bugzilla : 5681, 9562
3993 Description: Client may oops in ll_unhash_aliases
3994 Details : Client dcache may become inconsistent in race condition.
3995 In some cases "getcwd" can fail if the current directory is
4001 Description: Inode refcounting problems in NFS export code
4002 Details : link_raw functions used to call d_instantiate without obtaining
4003 extra inode reference first.
4007 Bugzilla : 9942, 9903
4008 Description: Referencing freed requests leading to crash, memleaks with NFS.
4009 Details : We used to require that call to ll_revalidate_it was always
4010 followed by ll_lookup_it. Also with revalidate_special() it is
4011 possible to call ll_revalidate_it() twice for the same dentry
4012 even if first occurence returned success. This fix changes semantic
4013 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4014 reference on a request referred from the intent.
4015 ll_intent_release() then releases such a request.
4018 Frequency : rare, normally benchmark loads only
4020 Description: unlinked inodes were kept in memory on the client
4021 Details : If a client is repeatedly creating and unlinking files it
4022 can accumulate a lot of stale inodes in the inode slab cache.
4023 If there is no other client load running this can cause the
4024 client node to run out of memory. Instead flush old inodes
4025 from client cache that have the same inode number as a new inode.
4028 Frequency : SLES9 2.6.5 kernel and long filenames only
4029 Bugzilla : 9969, 10379
4030 Description: utime reports stale NFS file handle
4031 Details : SLES9 uses out-of-dentry names in some cases, which confused
4032 the lustre dentry revalidation. Change it to always use the
4036 Frequency : rare, unless heavy write-truncate concurrency is continuous
4037 Bugzilla : 4180, 6984, 7171, 9963, 9331
4038 Description: OST becomes very slow and/or deadlocked during object unlink
4039 Details : filter_destroy() was holding onto the parent directory lock
4040 while truncating+unlinking objects. For very large objects this
4041 may block other threads for a long time and slow overall OST
4042 responsiveness. It may also be possible to get a lock ordering
4043 deadlock in this case, or run out of journal credits because of
4044 the combined truncate+unlink. Solution is to do object truncate
4045 first in one transaction without parent lock, and then do the
4046 final unlink in a new transaction with the parent lock. This
4047 reduces the lock hold time dramatically.
4050 Frequency : rare, 2.4 kernels only
4052 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4053 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4054 before unmounting their filesystem in order to avoid deadlock.
4055 The kernel_locked() function in 2.4 kernels only checks whether
4056 the kernel lock is held, not whether it is this process that is
4057 holding it as 2.6 kernels do.
4062 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4063 Details : The client ptlrpc code may be trying to reconnect to a down
4064 server before a previous connection attempt has timed out.
4065 Increase the reconnect interval to be longer than the connection
4066 timeout interval to avoid sending duplicate connections to
4070 Frequency : echo_client brw_test command
4072 Description: fix echo_client to work with OST preallocated code
4073 Details : OST preallocation code (5137) didn't take echo_client IO path
4074 into account: echo_client calls filter methods outside of any
4075 OST thread and, hence, there is no per-thread preallocated
4076 pages and buffers to use. Solution: hijack pga pages for IO. As
4077 a byproduct, this avoids unnecessary data copying.
4081 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4082 Description: Client can oops in mdc_commit_close() after open replay
4083 Details : It was possible for the MDS to return an open request with no
4084 transaction number in mds_finish_transno() if the client was
4085 evicted, but without actually returning an error. Clients
4086 would later try to replay that open and may trip an assertion
4087 Simplify the client close codepath, and always return an error
4088 from the MDS in case the open is not successful.
4091 Frequency : rare, 2.6 OSTs only
4093 Description: OST may deadlock under high load on fragmented files
4094 Details : If there was a heavy load and highly-fragmented OST filesystems
4095 it was possible to have all the OST threads deadlock waiting on
4096 allocation of biovecs, because the biovecs were not released
4097 until the entire RPC IO was completed. Instead, release biovecs
4098 as soon as they are complete to ensure forward IO progress.
4100 Severity : enhancement
4102 Description: Support for specifying external journal device at mount
4103 Details : If an OST or MDS device is formatted with an external journal
4104 device, this device major/minor is stored in the ext3 superblock
4105 and may not be valid for failover. Allow detecting and
4106 specifying the external journal at mount time.
4111 Description: Mounting an MDS with pending unlinked files may cause oops
4112 Details : target_finish_recovery() calls mds_postrecov() which returned
4113 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4114 considers this an error and immediately begins cleaning up the
4115 lov, just after starting the mds_lov process
4117 Severity : enhancement
4119 Description: Implement 'lfs df' to report actual free space on per-OST basis
4120 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4121 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4122 usage of objects; '-h' to report in human readable format.
4124 ------------------------------------------------------------------------------
4126 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
4133 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4134 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4135 at startup during orphan unlinking. Instead, skip buddy update
4136 before setup, as it will be regenerated after recovery is complete.
4141 Description: If an OST is inactive, its locks might reference stale inodes.
4142 Details : lov_change_cbdata() must iterate over all namespaces, even if
4143 they are inactive to clear inode references from the lock.
4145 Severity : enhancement
4146 Frequency : occasional, if non-standard max_dirty_mb used
4148 Description: Client will block write RPCs if not enough grant
4149 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4150 then the client will block writes while waiting for another RPC
4151 to complete instead of consuming its dirty limit. With change
4152 we get improved performance when max_dirty_mb is small.
4154 Severity : enhancement
4155 Bugzilla : 3389, 6253
4156 Description: Add support for supplementary groups on the MDS.
4157 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4158 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4159 lookups for user supplementary groups into a cache.
4163 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4164 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4165 for liblustre O_EXCL is enforced by MDS.
4169 Description: readdir never returns NULL in liblustre.
4170 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4171 dentry in struct dirent.
4175 Frequency : liblustre only, depends on application IO pattern
4176 Description: liblustre clients evicted if not contacting servers
4177 Details : Don't put liblustre clients into the ping_evictor list, so
4178 they will not be evicted by the pinger ever.
4180 Severity : enhancement
4182 Description: Add ability to evict clients by NID from MDS.
4183 Details : By echoing "nid:$NID" string into
4184 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4185 $NID would be instantly evicted from this MDS and from all active
4186 OSTs connected to it.
4190 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4191 Details : lookup_it_finish() used to query file size from OSTs that was not
4196 Description: service threads change working directory to that of init
4197 Details : Starting lustre service threads may pin the working directory
4198 of the parent thread, making that filesystem busy. Threads
4199 now change to the working directory of init to avoid this.
4203 Frequency : during shutdown only
4204 Description: shutdown with a failed MDS or OST can cause unmount to hang
4205 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4210 Frequency : 2.6 only
4211 Description: chmod/chown may include an extra supplementary group
4212 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4213 supplementary group and if none is specified this is used.
4216 Bugzilla : 5479 (6816)
4218 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4219 Details : If lookup is in progress on a file that is unlinked we might try
4220 to revalidate the inode and fail in revalidate after lookup is
4221 complete and ll_file_open() enqueues the open again but
4222 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4225 Frequency : always, if lconf --abort_recovery used
4227 Description: lconf --abort_recovery fails with 'Operation not supported'
4228 Details : lconf was attempting to abort recovery on the MDT device and not
4231 ------------------------------------------------------------------------------
4233 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4238 Frequency : rare (only unsupported configurations with a node running as an
4240 Bugzilla : 6514, 5137
4241 Description: Mounting a Lustre file system on a node running as an OST could
4243 Details : OSTs now preallocates memory needed to write out data at
4244 startup, instead of when needed, to avoid having to
4245 allocate memory in possibly low memory situations.
4246 Specifically, if the file system is mounted on on OST,
4247 memory pressure could force it to try to write out data,
4248 which it needed to allocate memory to do. Due to the low
4249 memory, it would be unable to do so and the node would
4250 become unresponsive.
4252 Severity : enhancement
4254 Description: Addition of lconf --service command line option
4255 Details : lconf now accepts a '--service <arg>' option, which is
4256 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4258 Severity : enhancement
4260 Description: Failover mode is now the default for OSTs.
4261 Details : By default, OSTs will now run in failover mode. To return to
4262 the old behaviour, add '--failout' to the lmc line for OSTs.
4264 Severity : enhancement
4266 Description: Health checks are now provided for MDS and OSTs
4267 Details : Additional detailed health check information on MSD and OSTs
4268 is now provided through the procfs health_check value.
4271 Frequency : occasional, depends on IO load
4273 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4274 after numerous create/delete cycles
4275 Details : The ext3 inode allocation policy would not allocate new inodes
4276 very well on the OSTs because there are no new directories
4277 being created. Instead we look for groups with free space if
4278 the parent directories are nearly full.
4283 Description: Network or server problems during mount may cause partially
4284 mounted clients instead of returning an error.
4285 Details : The config llog parsing code may overwrite the error return
4286 code during mount error handling, returning success instead
4292 Description: MDS can fail to allocate large reply buffers
4293 Details : After long uptimes the MDS can fail to allocate large reply
4294 buffers (e.g. zconf client mount config records) due to memory
4295 fragmentation or consumption by the buffer cache. Preallocate
4296 some large reply buffers so that these replies can be sent even
4297 under memory pressure.
4301 Frequency : rare (liblustre)
4302 Description: fsx running with liblustre complained that using truncate() to
4303 extend the file doesn't work. This patch corrects that issue.
4304 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4305 ATTR_SIZE and lsm use in llu_setattr_raw.
4309 Frequency : rare, only 2.6 kernels
4310 Description: Unusual file access patterns on the MDS may result in inode
4311 data being lost in very rare circumstances.
4312 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4313 code in the 2.6 kernel caused the inode and/or EA data not to
4314 be read from disk, causing single-file corruption.
4318 Frequency : rare, only 2.6 filesystems using extents
4319 Description: Heavy concurrent write and delete load may cause data corruption.
4320 Details : It was possible under high-load situations to have an extent
4321 metadata block in the block device cache from a just-unlinked
4322 file overwrite a newly-allocated data block. We now unmap any
4323 metadata buffers that alias just-allocated data blocks.
4327 Frequency : filesystems with default stripe_count larger than 77
4328 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4329 Details : lconf specifies an inode size of 4096 bytes when the default
4330 stripe_count is larger than 77. This conflicts with the default
4331 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4332 this case to avoid pinning too much memory for large EAs.
4334 ------------------------------------------------------------------------------
4336 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4341 Frequency : rare (extremely heavy IO load with hundreds of clients)
4343 Description: Client is evicted, gets IO error writing to file
4344 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4345 caused clients to be evicted for AST timeouts. The fixes in
4346 bug 5192 mean we no longer need to have such short AST timeouts
4347 so ldlm_timeout has been increased.
4350 Frequency : occasional during --force or --failover shutdown under load
4351 Bugzilla : 5949, 4834
4352 Description: Server oops/LBUG if stopped with --force or --failover under load
4353 Details : a collection of import/export refcount and cleanup ordering
4354 issues fixed for safer force cleanup
4357 Frequency : only filesystems larger than 120 OSTs
4358 Bugzilla : 5990, 6223
4359 Description: lfs getstripe would oops on a very large filesystem
4360 Details : lov_getconfig used kfree on vmalloc'd memory
4363 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4364 Bugzilla : 6242, 6243
4365 Description: reading from files that had been truncated to a non-zero size
4366 but never opened returned no data
4367 Details : ll_file_read() reads zeros from no-object files to EOF
4372 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4373 Details : There's little harm in aborting MDS/OSS recovery and letting it
4374 try again, so I removed the LASSERT and return an error instead.
4376 Severity : enhancement
4378 Description: New debugging infrastructure for tracking down data corruption
4379 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4380 (b) cover more of the client-side code path, and (c) try to narrow
4381 down where problems occurred
4385 Bugzilla : 3819, 4364, 4397, 6313
4386 Description: Racing close and eviction MDS could cause assertion in mds_close
4387 Details : It was possible to get multiple mfd references during close and
4388 client eviction, leading to one thread referencing a freed mfd.
4390 Severity: : enhancement
4391 Bugzilla : 3262, 6359
4392 Description: Attempts to reconnect to servers are now more aggressive.
4393 Details : This builds on the enhanced upcall-less recovery that was added
4394 in 1.4.2. When trying to reconnect to servers, clients will
4395 now try each server in the failover group every 10 seconds. By
4396 default, clients would previously try one server every 25 seconds.
4401 Description: After recovery, certain operations trigger a failed
4402 assertion on a client.
4403 Details : Failing over an mds, using lconf -d --failover, while a
4404 client was doing a readdir() call would cause the client to
4405 LBUG after recovery completed and the readdir() was resent.
4407 Severity : enhancement
4409 Description: Default groups are now added by lconf
4410 Details : You can now run lconf --group <servicename> without having to
4411 manually add groups with lmc.
4414 Frequency : occasional
4416 Description: Nodes with an elan id of 0 trigger a failed assertion
4419 Frequency : always when accessing e.g. tty/console device nodes
4421 Description: tty and some other devices nodes cannot be used on lustre
4422 Details : file's private_data field is used by device data and lustre
4423 values in there got lost. New field was added to struct file to
4424 store fs-specific private data.
4427 Frequency : when exporting Lustre via NFS
4429 Description: NFSD failed occasionally when looking up a path component
4430 Details : NFSD is looking up ".." which was broken in ext3 directories
4431 that had grown large enough to become hashed.
4434 Frequency : Clusters with multiple interfaces not on the same subnet
4436 Description: Nodes will repeatedly try to reconnect to an interface which it
4437 cannot reach and report an error to the log.
4438 Details : Extra peer list entries will be created by lconf with some peers
4439 unreachable. lconf now validates the peer before adding it.
4442 Frequency : Only if a default stripe is set on the filesystem root.
4444 Description: Setting a default stripe on the filesystem root prevented the
4445 filesystem from being remounted.
4446 Details : The client was sending extra request flags in the root getattr
4447 request and did not allocate a reply buffer for the dir EA.
4450 Frequency : occasional, higher if lots of files are accessed by one client
4451 Bugzilla : 6159, 6097
4452 Description: Client trips assertion regarding lsm mismatch/magic
4453 Details : While revalidating inodes the VFS looks up inodes with ifind()
4454 and in rare cases can find an inode that is being freed.
4455 The ll_test_inode() code will free the lsm during ifind()
4456 when it finds an existing inode and then the VFS later attaches
4457 this free lsm to a new inode.
4461 Bugzilla : 6422, 7030
4462 Description: MDS deadlock between mkdir and client eviction
4463 Details : Creating a new file via mkdir or mknod (starting a transaction
4464 and getting the ns lock) can deadlock with client eviction
4465 (gets ns lock and trying to finish a synchronous transaction).
4468 Frequency : occasional
4469 Description: While starting a server, the fsfilt_ext3 module could not be
4471 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4472 kernels. Previously, lconf would still use the ext3 name
4473 when trying to load modules. Now, it will correctly use
4474 ext3 on 2.4 and ldiskfs on 2.6.
4476 Severity : enhancement
4477 Description: The default stripe count has been changed to 1
4478 Details : The interpretation of the default stripe count (0, to lfs
4479 or lmc) has been changed to mean striping across a single
4480 OST, rather than all available. For general usage we have
4481 found a stripe count of 1 or 2 works best.
4483 Severity : enhancement
4484 Description: Add support for compiling against Cray portals.
4485 Details : Conditional compiling for some areas that are different
4489 Frequency : occasional
4490 Bugzilla : 6409, 6834
4491 Description: Creating files with an explicit stripe count may lead to
4492 a failed assertion on the MDS
4493 Details : If some OSTs are full or unavailable, creating files may
4494 trigger a failed assertion on the MDS. Now, Lustre will
4495 try to use other servers or return an error to the
4499 Frequency : occasional
4501 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4502 caused lock timeout during readahead (since 1.4.2).
4503 Details : Processes doing readahead might match a lock that hasn't been
4504 granted yet if there are overlapping and conflicting lock
4505 requests. The readahead process waits on ungranted lock
4506 (original lock is CBPENDING), while OST waits for that process
4507 to cancel CBPENDING read lock and eventually evicts client.
4509 Severity : enhancement
4511 Description: Initial enabling of flock support for clients
4512 Details : Implements fcntl advisory locking and file status functions.
4513 This feature is provided as an optional mount flag (default
4514 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4515 locking are implemented yet, and those that are are not guaranteed
4516 to be completely correct in production environments.
4517 mount -t lustre -o [flock|noflock] ...
4520 Frequency : occasional
4522 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4523 assertion in the ext3 JBD (journaling) layer.
4524 Details : The b_committed_data struct is protected by the big kernel lock
4525 in 2.4 kernels, serializing journal_commit_transaction() and
4526 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4527 access to this struct. In 2.6 kernels there is finer grained
4528 locking to improve SMP performance of the JBD layer.
4532 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4534 -----------------------------------------------------------------------------
4536 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4538 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4539 versions of Lustre. Please update all servers and clients to
4540 version 1.4.2 or later at the same time. You must also run
4541 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4542 to update the configuration logs.
4544 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4545 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4546 - swab llog records of type '0' so we get proper header size/idx (5861)
4547 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4548 - fix rename of one directory over another leaking an inode (5953)
4549 - avoid SetPageDirty on 2.6 (5981)
4550 - don't re-add just-being-destroyed locks to the waiting list (5653)
4551 - when creating new directories, inherit the parent's custom
4552 striping settings if present parent (3048)
4553 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4554 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4555 - don't hold client locks on temporary worklist from l_lru (5666)
4556 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4557 - automatically evict dead clients (5921)
4558 - Update file size properly in create+truncate+fstat case (6196)
4559 - Do not unhash mountpoint dentries, do not allow removal of
4561 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4562 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4563 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4564 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4565 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4566 - Correctly handle reads of files with no objects (6243)
4567 - lctl recover will also mark a device active if deactivate used (5933)
4569 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4570 - allow --write-conf on an MDS with different nettype than client (5619)
4571 - don't write config llogs to MDS for mounts not from that MDS (5617)
4572 - lconf should create multiple TCP connections from a client (5201)
4573 - init scripts are now turned off by default; run chkconfig --on
4574 lustre and chkconfig --on lustrefs to use them
4575 - upcalls are no longer needed for clients to recover to failover
4577 - add --abort-recovery option to lconf to abort recovery on device
4579 - add support for an arbitrary number of OSTs (3026)
4580 - Quota support protocol changes.
4581 - forward compatibility changes to wire structs (6007)
4582 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4583 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4584 - improved "lustre status" script
4585 - initialize blocksize for non-regular files (6062)
4586 - added --disable-server and --disable-client configure options (5782)
4587 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4588 - Vanilla 2.4.29 support
4589 - increase maximum number of obd devices to 520 (6242)
4590 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4591 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4592 SLES 9 2.6.5 kernels
4593 - limit stripes per file to 160 (the maximum EA size) (6093)
4595 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4598 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4599 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4600 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4601 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4602 - small ext3 extents cleanups and fixes (5733)
4603 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4604 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4605 - avoid needless client->OST connect, fix handle mismatch (5317)
4606 - fix DLM error path that led to out-of-sync client, long delays (5779)
4607 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4608 - fix several locking issues related to i_size (5492,5624,5654,5672)
4609 - don't move pending lock onto export if it is already evicted (5683)
4610 - fix kernel oops when creating .foo in unlinked directory (5548)
4611 - fix deadlock in obdfilter statistics vs. object create (5811)
4612 - use time_{before,after} to avoid timer jiffies wrap (5882)
4613 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4614 - Do not leak request if server was not able to process it (5154)
4615 - If mds_open unable to find parent dir, make that negative lookup(5154)
4616 - don't create new directories with extent-mapping (5909, 5936)
4618 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4619 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4620 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4621 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4622 - support for RHEL4 kernel on i686 (5773)
4623 - provide error messages when incompatible logs are encountered (5898)
4625 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4626 * version 1.4.0.10 (1.4.1 release candidate 1)
4628 - don't keep a lock reference when lock is not granted (4238)
4629 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4630 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4631 - drop import inflight refcount on signal_completed_replay error (5255)
4632 - unlock page after async write error during send (3677)
4633 - handle missing objects in filter_preprw_read properly (5265)
4634 - no transno return for symlink open, don't save no-trasno open (3440)
4635 - don't try to complete elan receive that already failed (4012)
4636 - free RPC server reply state on error (5406)
4637 - clean up thread from ptlrpc_start_thread() on error (5160)
4638 - readahead could read extra page into cache that wasn't ejected (5388)
4639 - prevent races in class_attach/setup/cleanup/detach (5260)
4640 - don't dereference de->d_inode after l_dput of de (5458)
4641 - use "int" for stripe value returned from lock_to_stripe (5544)
4642 - mballoc allocation and error-checking fixes in 2.6 (5504)
4643 - block device patches to fix I/O request sizes in 2.6 (5482)
4644 - look up hostnames for IB nals (5602)
4645 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4646 - don't start multiple acceptors for the same port (5277)
4647 - fix incorrect LASSERT in mds_getattr_name (5635)
4648 - export a proc file for general "ping" checking (5628)
4649 - fix "lfs check" to not block when the MDS is down (5628)
4651 - service request history (4965)
4652 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4653 - create an "evict_client" /proc entry on OSTs, like the MDS has
4654 - fix mount usage message, return errors per mount(8) (5168)
4655 - change grep [] to grep "[]" in tests so they work in more UMLs
4656 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4657 - remove ancient LOV_MAGIC_V0 EA support (5047)
4658 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4659 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4660 - allow readahead allocations to fail when low on memory (5383)
4661 - mmap locking landed again, after considerable improvement (2828)
4662 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4664 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4667 - send OST transaction number in read/write reply to free req (4966)
4668 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4669 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4670 - the watchdog thread now runs as interruptible (5246)
4671 - flock/lockf fixes (but it's still disabled, pending 5135)
4672 - don't use EXT3 constants in llite code (5094)
4673 - memory shortage at startup could cause assertion (5176)
4675 - reorganization of lov code
4676 - single portals codebase
4678 - add extents/mballoc support (5025)
4679 - direct I/O reads in the obdfilter (4048)
4680 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4682 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4685 - send OST transaction number in read/write reply to free req (4966)
4686 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4687 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4688 - the watchdog thread now runs as interruptible (5246)
4689 - handle missing objects in filter_preprw_read properly (5265)
4690 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4691 - drop import inflight refcount on signal_completed_replay error (5255)
4692 - unlock page after async write error during send (3677)
4693 - return original error code on reconstructed replies (3761)
4694 - no transno return for symlink open, don't save no-trasno open (3440)
4696 - add pid to ldlm debugging output (4922)
4697 - bump the watchdog timeouts -- we can't handle 30sec yet
4698 - extra debugging for orphan dentry/inode bug (5259)
4700 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4703 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4704 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4705 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4706 - fix NULL dereference in /proc/sys/portals/routes (4827)
4707 - allow failed mdc_close() operations to be interrupted (4561)
4708 - stop precreate on OST before MDS would time out on it (4778)
4709 - don't send partial-page writes before EOF from client (4410)
4710 - discard client grant for sub-page writes on large-page clients (4520)
4711 - don't free dentries not owned by NFS code, check generation (4806)
4712 - fix lsm leak if mds_create_objects() fails (4801)
4713 - limit debug_daemon file size, always print CERROR messages (4789)
4714 - use transno after validating reply (3892)
4715 - process timed out requests if import state changes (3754)
4716 - update mtime on OST during writes, return in glimpse (4829)
4717 - add mkfsoptions to LDAP (4679)
4718 - use ->max_readahead method instead of zapping global ra (5039)
4719 - don't interrupt __l_wait_event() during strace
4721 - add software watchdogs to catch hung threads quickly (4941)
4722 - make lustrefs init script start after nfs is mounted
4723 - fix CWARN/ERROR duplication (4930)
4724 - return async write errors to application if possible (2248)
4725 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4726 - print NAL number in %x format (4645)
4727 - update barely-supported suse-2.4.21-171 series (4842)
4728 - support for sles 9 %post scripts
4729 - support for building 2.6 kernel-source packages
4730 - support for sles km_* packages
4732 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4735 - ignore -ENOENT errors in osc_destroy (3639)
4736 - notify osc create thread that OSC is being cleaned up (4600)
4737 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4738 - reconstruct ost_handle() like mds_handle() (4657)
4739 - create a new thread to do import eviction to avoid deadlock (3969)
4740 - let lconf resolve symlinked-to devices (4629)
4741 - don't unlink "objects" from directory with default EA (4554)
4742 - hold socknal file ref over connect in case target is down (4394)
4743 - allow more than 32000 subdirectories in a single directory (3244)
4744 - fix blocks count for O_DIRECT writes (3751)
4745 - OST returns ENOSPC from object create when no space left (4539)
4746 - don't send truncate RPC if file size isn't changing (4410)
4747 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4748 - bind to privileged port in socknal and tcpnal (3689)
4750 - rate limit CERROR/CWARN console message to avoid overload (4519)
4751 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4752 - basic mmap support (3918)
4753 - kernel patch series update from b1_4 (4711)
4755 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4758 - avoid crash during MDS cleanup with OST shut down (2775)
4759 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4760 - don't use bad inodes on the MDS (3744)
4761 - dynamic object preallocation to improve recovery speed (4236)
4762 - don't hold spinlock over lock dumping or change debug flags (4401)
4763 - don't zero obd_dev when it is force cleaned (3651)
4764 - print grants to console if they go negative (4431)
4765 - "lctl deactivate" will stop automatic recovery attempts (3406)
4766 - look for existing locks in ldlm_handle_enqueue() (3764)
4767 - don't resolve lock handle twice in recovery avoiding race (4401)
4768 - revalidate should check working dir is a directory (4134)
4770 - don't always mark "slow" obdfilter messages as errors (4418)
4772 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4775 - don't close LustreDB during write_conf until it is done (3860)
4776 - fix typo in lconf for_each_profile (3821)
4777 - allow dumping logs from multiple threads at one time (3820)
4778 - don't allow multiple threads in OSC recovery (3812)
4779 - fix debug_size parameters (3864)
4780 - fix mds_postrecov to initialize import for llog ctxt (3121)
4781 - replace config semaphore with spinlock (3306)
4782 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4783 - don't allow enqueue to complete on a destroyed export (3822)
4784 - down write_lock before checking llog header bitmap (3825)
4785 - recover from lock replay timeout (3764)
4786 - up llog sem before sending rpc (3652)
4787 - reduce ns lock hold times when setting kms (3267)
4788 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4789 - fix NULL deref and obd_dev leak on setup error (3312)
4790 - replace some LBUG about llog ops with error handling (3841)
4791 - don't match INVALID dentries from d_lookup and spin (3784)
4792 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4793 - fix invalid assertion in ptlrpc_set_wait (3880)
4795 - add libwrap support for the TCP acceptor (3996)
4796 - add /proc/sys/portals/routes for non-root route listing (3994)
4797 - allow setting MDS UUID in .xml (2580)
4798 - print the stack of a process that LBUGs (4228)
4800 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4803 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4804 - make sure to unset replay flag from failed open requests (3440)
4805 - if default stripe count is 0, use OST count for inode size (3636)
4806 - update parent mtime/ctime on client for create/unlink (2611)
4807 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4808 - free recovery state on server during a forced cleanup (3571)
4809 - unregister_reply for resent reqs (3063)
4810 - loop back devices mounting and status check on 2.6 (3563)
4811 - fix resource-creation race that can provoke i_size == 0 (3513)
4812 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4813 - more debugging for page-accounting assertion (3746)
4814 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4815 - avoid deadlock after precreation failure (3758)
4816 - fix race and lock order deadlock in orphan handling (3450, 3750)
4817 - add validity checks when grabbing inodes from l_ast_data (3599)
4819 - add /proc/.../recovery_status to obdfilter (3428)
4820 - lightweight CDEBUG infrastructure, debug daemon (3668)
4821 - change default OSC RPC parameters to be better on small clusters
4822 - turn off OST read cache for files smaller than 32MB
4823 - install man pages and include them in rpms (3100)
4824 - add new init script for (un)mounting lustre filesystems (2593)
4825 - run chkconfig in %post for init scripts (3701)
4826 - drop scimac NAL (unmaintained)
4828 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4831 - clean kiobufs before and after use (3485)
4832 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4833 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4834 - change init script to fail more clearly if not run as root (1528)
4835 - allow clients to reconnect during replay (1742)
4836 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4837 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4838 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4839 - fix cond_resched (3554)
4840 - only evict unfinished clients after recovery (3515)
4841 - allow bulk resend, prevent data loss (3570)
4842 - dynamic ptlrpc request buffer allocation (2102)
4843 - don't allow unlinking open directory if it isn't empty (2904)
4844 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4845 - remove extraneous obd dereference causing LASSERT failure (3334)
4846 - don't use get_cycles() when creating temp. files on the mds (3156)
4847 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4848 - handle EEXIST for set-stripe, set proper directory name (3336)
4850 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4851 - fix ksocknal_fmb_callback() error messages (2918)
4853 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4856 - don't copy lvb into (possibly NULL) reply on error (2983)
4857 - don't deref dentry after dput, don't free lvb on error (2922)
4858 - use the kms to determine writeback rpc length (2947)
4859 - increment oti_logcookies when osc is inactive (2948)
4860 - update client's i_blocks count via lvb messages (2543)
4861 - handle intent open/close of special files properly (1557)
4862 - mount MDS with errors=remount-ro, like obdfilter (2009)
4863 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4864 - don't use cancelling-locks' kms values (2947)
4865 - use highest lock extent for kms, not last one (2925)
4866 - don't dereference ERR_PTR() dentry in error handling path (3107)
4867 - fix thread race in portals_debug_dumplog() (3122)
4868 - create lprocfs device entries at setup instead of at attach (1519)
4869 - common AST error handler, don't evict client on completion race (3145)
4870 - zero nameidata in detach_mnt in 2.6 (3118)
4871 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4872 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4873 - initialize RPC timeout timer earlier for 2.6 (3219)
4874 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4875 - print nal/nid for unknown nid (3258)
4876 - additional checks for oscc recovery before doing precreate (3284)
4877 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4878 - don't crash in mdc_close for bad permissions on open (3285)
4879 - zero i_rdev for non-device files (3147)
4880 - clear page->private before handing to FS, better assertion (3119)
4881 - tune the read pipeline (3236)
4882 - fix incorrect decref of invalidated dentry (2350)
4883 - provide read-ahead stats and refine rpc in flight stats (3328)
4884 - don't hold journal transaction open across create RPC (3313)
4885 - update atime on MDS at close time (3265)
4886 - close LDAP connection when recovering to avoid server load (3315)
4887 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4888 - don't leak open file on MDS after open resend (3325)
4889 - serialize filter_precreate and filter_destroy_precreated (3329)
4890 - loop device shouldn't call sync_dev() for nul device (3092)
4891 - clear page cache after eviction (2766)
4892 - resynchronize MDS->OST in background (2824)
4893 - refuse to mount the same filesystem twice on same mountpoint (3394)
4894 - allow llmount to create routes for mounting behind routers (3320)
4895 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4896 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4897 - fix rare problem with rename on htree directories (3417)
4899 - allow default OST striping configuration per directory (1414)
4900 - fix compilation for qswnal for 2.6 kernels (3125)
4901 - increase maximum number of MDS request buffers for large systems
4902 - change liblustreapi to be useful for external progs like lfsck (3098)
4903 - increase local configuration timeout for slow disks (3353)
4904 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4906 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4909 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4910 - don't overwrite extent policy data in reply if lock was blocked (2901)
4911 - drop filter export grants atomically with removal from device (2663)
4912 - del obd_self_export from work_list in class_disconnect_exports (2908)
4913 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4914 - swab reply message in mdc_close, other PPC fixes (2464)
4915 - fix destroying of named logs (2325)
4916 - overwrite old logs when running lconf --write_conf (2264)
4917 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4918 - fix race in target_handle_connect (2898)
4919 - mds_reint_create() should take same inode create lock (2926)
4920 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4921 - don't close files for self_export to avoid uninitialized obd (2936)
4922 - allow MDS with the same name as client node (2939)
4923 - hold dentry reference for closed log files for unlink (2325)
4924 - reserve space for all logs during transactions (2059)
4925 - don't evict page beyond end of stripe extent (2925)
4926 - don't oops on a deleted current working directory (2399)
4927 - handle hard links to targets without a parent properly (2517)
4928 - don't dereference NULL lock when racing during eviction (2867)
4929 - don't grow lock extents when lots of conflicting locks (2919)
4931 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4934 - account for cache space usage on clients to avoid data loss (974)
4935 - lfsck support in lustre kernel code (2349)
4936 - reduce journal credits needed for BRW writes (2370)
4937 - orphan handling to avoid losing space on client/server crashes
4938 - ptlrpcd can be blocked, stopping ALL progress (2477)
4939 - use lock value blocks to assist in proper KMS, faster stat (1021)
4940 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4941 - recovery for initial connections (2355)
4942 - fixes for mds_cleanup_orphans (1934)
4943 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4944 - block all file creations until orphan recovery completes (1901)
4945 - client remove rq_connection from request struct (2423)
4946 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4947 - recovery timer race (2670)
4948 - mdc_close recovey bug (2532)
4949 - ptlrpc cleanup bug (2710)
4950 - mds timeout on local locks (2588)
4951 - namespace lock held during RPCs (2431)
4952 - handle interrupted sync write properly (2503)
4953 - don't try to handle a message that hasn't been replied to (2699)
4954 - client assert failure during cleanup after abort recovery (2701)
4955 - leak mdc device after failed mount (2712)
4956 - ptlrpc_check_set allows timedout requests to complete (2714)
4957 - wait for inflight reqs when ptlrpcd finishes (2710)
4958 - make sure unregistered services are removed from the srv_list
4959 - reset bulk XID's when resending them (caught by 1138 test)
4960 - unregister_bulk after timeout
4961 - fix lconf error (2694)
4962 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4963 - readahead locks pages, leaves pending causing memory pressure (2673)
4964 - increase OST request buffers to 4096 on large machines (2729)
4965 - fix up permission of existing directories in simple_mkdir (2661)
4966 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4967 - don't assign transno to errored transactions (2742)
4968 - don't delete objects on OST if given a bogus objid from MDS (2751)
4969 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4970 - if rq_no_resend, then timeout request after recovery (2432)
4971 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4972 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4973 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4974 - MDS recovery shouldn't delete objects that it has given out (2730)
4975 - if enqueue arrives after completion, don't clobber LVB (2819)
4976 - don't unlock pages twice when trigger_group_io returns error (2814)
4977 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4978 - don't write pages to disk if there was an error (1450)
4979 - don't ping imports that have recovery disabled (2676)
4980 - take buffered bytes into account when balancing socknal conn (2817)
4981 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4982 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4983 - remove little-endian swabbing of llog records (1987)
4984 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4985 - retry reposting request buffers if they fail (1191)
4986 - grow extent at grant time to avoid granting a revoked lock (2809)
4987 - lock revoke doesn't evict page if covered by a second lock (2765)
4988 - disable VM readahead to avoid reading outside lock extents (2805)
4990 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4991 - updated kernel patches for hp-2.4.20 kernel (2681)
4993 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4996 - fix truncated write corruption (2366)
4997 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4999 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5000 - ensure len(uuid) < 37 in lmc (1171)
5001 - fix ia64 OOPS in llog_test (2255)
5002 - zero end of page at obdfilter for partial page writes (2648)
5003 - don't leave stale dentries around after renames (bug 2428)
5004 - fix timeouts when evicting a client with a single lock held (2642)
5005 - set deadline for the initial HELLO message to drain (2634)
5006 - print out dotted-quad IP addresses in the socknal (2302)
5008 - additional debugging for MDS client eviction problem (2443)
5009 - fix mkfsoptions support for osts (2603, 2604)
5011 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5014 - add series for the vanilla 2.6.0 kernel
5015 - add series for the vanilla 2.4.24 kernel
5016 - add series for a cray x86/64 UL kernel drop
5017 - fix xattr patches for the vanilla 2.4.19 series
5019 - generate true UUIDs in lmc (1171)
5020 - have portals stack dumping break in UML (2466)
5021 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5022 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5023 - wake callback waiting threads on client eviction (2460)
5024 - Add --ptldebug and --subsystem to lmc (1719)
5025 - update assertion to allow safe interrupt allocation
5026 - set rq_no_resend for cancel requests (2432)
5027 - recalculate ptlrpcd timeout after resend (2494)
5028 - call vfs_rmdir when removing pending directories (2368)
5029 - fix renaming a file to itself (2429)
5030 - lmc creates a default one-stripe lov (2454)
5031 - expand procfs space to handle large clusters (2326)
5032 - increase UML stack to avoid overflow
5033 - update lconf's list of debug and subsystem masks
5034 - fix lfs find --obd (2510)
5035 - /proc tunable for disabling filter read caching (2591)
5036 - stop rpm packages from altering slapd.conf (2301)
5037 - disable nagle in the socknal under 0conf (2578)
5038 - choose mds inode size based on stripe count (2572)
5039 - fix kernel-source rpm problems (2516)
5041 - add --disable-doc to avoid pdf generation (2421)
5042 - update documentation, tests, type-os, comments
5043 - avoid format warnings on ia64
5044 - remove the TOE NAL
5045 - tiny code cleanups by removing unused fields
5047 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5050 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5051 - remove the most blatant lies from BUILDING file (2371)
5052 - change default debug level to reasonable production setting
5053 - reduce client side cache size to reduce cache flush time
5054 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5055 - make TCP zerocopy and pinger support enabled by default (2476)
5056 - sync writes completed after process exits caused crashes (2319)
5057 - maintain correct mount count on the MDS (2356)
5058 - backout 1557, because 2316 wasn't really fixed
5059 - better file I/O statistics gathering in /proc
5060 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5061 - another kernel patch to fix zero-copy TCP function export
5062 - don't take duplicate lock when processing re-sent getattr (2420)
5063 - lctl uses obd_self_export instead of creating new conn (2353)
5064 - MDS/OST recovery case which requires object creation asserted (2425)
5065 - move lfs from /usr/sbin to /usr/bin in packages
5066 - fix race between mds_client_add and mds_client_free (2417)
5067 - use kmalloc instead of slabs in portals (2430)
5068 - don't create duplicate records when a failover MDS is present (2442)
5069 - remove unnecessary mount age check (2332)
5070 - don't remove directory inodes from locks prematurely (2451)
5071 - don't break if MDS service name is the same as hostname (2103)
5072 - fix races in client write RPC generation when cache full (2482)
5074 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5077 - remove now-unused request->rq_obd (278)
5078 - if an allocation fails, print out how much memory we've used (1933)
5079 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5080 - add the "configurable stack size" patch to most series files (1256)
5081 - ability to write large log records, for 100+ OST configs (2306)
5082 - fix NULL deref when filter_prep fails (2314)
5083 - fix operator precedence error in filter_sync
5084 - dynamic allocation of socknal TX descriptors (2315)
5085 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5086 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5087 - ensure that CWARN messages in Portals always get to the syslog
5088 - __init/__exit are not for prototype decls (ldlm_init/exit)
5089 - x86-64 compile warning fixes
5090 - fix gateway LMC keyword conflict (2318)
5091 - fix MDS lock inversions in getattr/reint paths (1844)
5092 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5093 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5094 - DLM race condition prevented some lock evictions (2328)
5095 - ENOMEM detection and retry on socknal sends (2230)
5096 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5097 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5098 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5099 - better error messages when a client is rejected during recovery (1505)
5100 - avoid cancelling locks which were never granted, after failure (2330)
5101 - fix i_sem/journal inversion in mds_client_add (2333)
5102 - fix truncate/getattr lock cycle deadlock (2334)
5103 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5104 - fix two rare exit paths which could leak an l_lock() ref (2321)
5105 - fencepost error in MDS/OST orphan recovery (2226)
5106 - make log record alignment 8 bytes (1988)
5107 - lstripe now fails when requested offset > ost_count (2237)
5108 - ensure that all kernel series have a complete list.h (1607)
5109 - fix crashes in special-file operations (2316)
5110 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5112 - allow configurable automake binary, for testing new versions
5113 - small update to the lfs documentation
5115 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5117 * fix negative export reference count in fsfilt_sync (2312)
5119 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5120 * release candidate 0.9.1
5122 - orphans are moved into the PENDING directory for possible recovery
5123 - replayed opens now open by fid for orphan/rename safety (1042)
5124 - last close of an orphan inode generates a transno (683)
5125 - chdir() and mount() now pin the directory entry (1020)
5126 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5127 - discard very old requests without processing them (1502)
5128 - remove some common, well-understood CERRORs (1505)
5129 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5130 - clear "grant" flags in OST replies until OST grant code lands (1644)
5131 - fix read performance by not clobbering i_blksize on client (1598)
5132 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5133 - make LRU size a /proc tunable, clears locks when reduced (707)
5134 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5135 - server locks take references on exports now (1558)
5136 - build fixes for 2.4.20-rh trees (1663)
5137 - return an error from lov_create if all OSCs are inactive (1751)
5138 - fix import levels when a reconnect happens without a timeout (1597)
5139 - exit early from mds_open if we get a lookup error (1749)
5140 - partial page read at EOF wouldn't wait for disk before sending (1642)
5141 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5142 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5143 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5144 - write/truncate lock inversion (1639)
5145 - Don't auto-load obdclass, portals modules during cleanup (1495)
5146 - fix timestamps from jumping to "now" (1763)
5147 - extra journal assertions (1648)
5148 - add an extra multiunlink test (1771)
5149 - fix read_record/write_record API (1776)
5150 - fix leak of offset_extent, possible incorrect i_size later (1772)
5151 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5152 - Debugging for the kqswnal_get_idle_tx problems (1820)
5153 - Allow recovery to be attempted multiple times (1536)
5154 - Write out MDS last_rcvd file after it is first created (1600)
5155 - Fix tx_descriptor leak in failed transmit situations (1827)
5156 - ext3 journaling fixes for assertion failure after IO error (1871)
5157 - class_export_put() on freed export after completion AST error (1896)
5158 - Fix revalidate looping in VFS (1322)
5159 - Don't access a freed export during MDS_REINT timeout (1521)
5160 - Add open-unlink recovery support on the MDS (1673,1764)
5161 - Return an error if no MDS data was read from last_rcvd (1946)
5162 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5163 - Don't setup a disk device that doesn't match exported UUID (317)
5164 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5165 - avoid committing NULL handle in force close
5166 - local.sh is now a one-stripe LOV configuration
5167 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5168 - fix ext3 htree duplicate directory entry corruption (1516)
5169 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5170 - update to new LOV EA format (2097)
5171 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5173 2003-06-15 Phil Schwan <phil@clusterfs.com>
5176 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5177 - per-import/export recovery handling (958, 931, 959)
5178 - multiple last-rcvd slots, for serving multiple FSes (949)
5179 - connections are again shared between multiple imp/exports (963, 964)
5180 - "umount -f" would hang if any requests needed to be sent (393, 978)
5181 - avoid pinning large req buffer by copying for queued messages (989)
5182 - add "uuid" to "lctl device" command to help upcalls (991)
5183 - "open" RPCs with transnos would confuse recovery counters (1037)
5184 - do proper endian conversion of all wire messages (288, 340, 891)
5185 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5186 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5187 - fix ptlrpc_connection leak in target_handle_connect (1174)
5188 - fix import refcounting bug in OST and MDS cleanup (1134)
5189 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5190 - fix very unlikely obd_types race condition (501)
5191 - remove osc_open hack for echo_client (1187)
5192 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5193 - a failure in read_inode2 leads to deadlock (1139)
5194 - cancel ack-locks as soon as transaction is committed (1072)
5195 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5196 - make sure to commitrw after any preprw to avoid deadlock (1162)
5197 - failing to execute a file in a lustre FS would lock inode (1203)
5198 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5199 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5200 - obd_uuid2tgt would walk off the end of the list (1255)
5201 - on IA64 the osc would give portals incorrect bulk size (1258)
5202 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5203 - fix lock inversion caused by new llite matching code (1282)
5204 - limit the number of dirty pages on a client to 10MB (1286)
5205 - timed out locks were not being corrected cancelled (1289)
5206 - fix O_DIRECT above 4GB on IA-32 (1292)
5207 * major user-visible changes
5208 - fail out/fail over policy now controlled by the upcall (993)
5210 - add OBD_PING to check server availability and failure (954)
5211 - lustre messages are now sent in sending host order (288, 340, 891)
5212 - add eadatalen to MDS getattr reply (340)
5213 - OST read replies may contain second buffer, with per-page status (593)
5215 2003-03-11 Phil Schwan <phil@clusterfs.com>
5218 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5219 - failed open()s could cause deadlock; fixed (867, 869)
5220 - stop cancelling OST locks when files are closed (481)
5221 - overlapping XID spaces caused network corruption (851, 853)
5222 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5223 - setattr_raw added, to do single-RPC, server-side setattrs
5224 - lmc/lconf syntax change for OST UUIDs
5225 - fix crashy race condition between ptlrpc_free_req and osc_close
5226 - don't use request in mdc_enqueue if we hit a timeout (889)
5227 - don't set the inode i_size for regular files from the MDS (896)
5228 - handle out of order completion AST (842)
5229 - don't LBUG if a lock request times out after receiving AST (913)
5230 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5231 - if a bad lock AST arrives, send an error instead of dropping entirely
5232 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5233 - fix leak in bulk IO when only partially completed (899, 900, 926)
5234 - fix O_DIRECT for ia64 (55)
5235 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5236 - C-z after timeout could hang a process forever; fixed (977)
5238 - client-side I/O cache (678, 924, 929, 941, 970)
5240 - READPAGE and SETATTRs which don't take server-side locks get
5243 2003-02-11 Phil Schwan <phil@clusterfs.com>
5246 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5247 - Fix fsfilt_extN_readpage() to read a full page of directory
5248 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5249 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5250 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5251 - Most intent operations were removed, in favour of a new RPC mode
5252 that does a single RPC to the server and bypasses most of the VFS
5253 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5254 - Aggressively cancel local locks on DLM servers
5255 - mds_reint_unlink sends EA to the client if it's the last nlink.
5256 client uses that EA to unlink OST objects.
5257 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5258 - recursive symlinks were fixed (439)
5259 - fixed NULL deref in DEBUG_REQ
5260 - filter_update_lastobjid no longer calls sync, which annoyed extN
5261 - fixed multi-client small-writes to a single file problem (445)
5262 - fixed mtime updates during file writes (607)
5263 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5264 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5265 - fixed Linux OST size reporting problem (444, 656)
5266 - OST now updates object mtime with writes or setattr (607, 619)
5267 - client verifies file size before zeroing page past EOF (445)
5268 - OST now writes last allocated objid to disk with allocation (108)
5269 - LOV on echo now works (409)
5271 - mds_reint_unlink sends a new buffer, with the EA included. this
5272 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5273 set if a regular file was being unlinked, and it was the last link
5274 - use PtlGet from the target for bulk writes (315)
5275 - OST now updates object mtime with writes or setattr (607, 619)
5276 - LDLM now has a grant-time callback to revalidate locked items, if
5278 - Many MDS operations were reorganized to combat race conditions
5280 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5281 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5283 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5286 - Fully reactivate OST imports after reconnection (512, others)
5287 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5288 - More graceful error handling for truncating on dead OST (515)
5289 - Don't error out unless we're actually accessing dead stripes (474)
5290 - Fix garbage sizes when stripes are missing (410)
5291 - LRU counters were broken, causing constant lock purge (433, 432)
5292 - garbage on read from stripes with failed OSTs (441)
5293 - mark OSCs as active before reconnecting during recovery (438)
5294 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5295 - lfind did not preserve OST order in output (443)
5296 - symlinks cause hung clients, incorrect data (439)
5297 - stop dereferencing request after dropping refcount (457)
5298 - don't LASSERT(spin_is_locked) on non-SMP (455)
5299 - fixes for many rename() bugs
5300 - fstat didn't correctly synchronize attributes (399)
5301 - server must handle lock cancellation during blocking AST prep (487)
5302 - bulk descriptors were free()d too soon (511)
5303 - fix paths in lconf, which would load incorrect modules (451, 507)
5304 - fix confusing lconf 'host not found' error message (386)
5305 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5306 - fix race condition in mdc_blocking_ast() for inode access (526)
5307 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5308 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5309 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5310 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5311 - fix lstripe in several ways (kernel side) (527)
5312 - fix request leak in ldlm_cli_enqueue (262)
5313 - incorrect OSC was marked inactive after OST failure
5314 - call mds_fs_cleanup before unmounting filesystem (524)
5315 - fix races between taking ns_lock and ldlm_lock_change_resource
5316 - fix races updating LOV export open file list
5317 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5318 - fix recovery NULL deref in ldlm_cli_cancel_unused
5319 - fix some DLM races by using new hash table for lock handles (419)
5320 - permit the client to specify desired inodes, at replay
5321 - duplicate requests when we queue them for replay reintegration
5322 - fix last_rcvd offset calculation
5323 - sync after each recovered transaction, so we always make progress
5324 - never, not always, ERESTART requests without transnos
5325 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5327 - skip replay if the MDS says that the client is already connected
5328 - don't check for a recovery-enabled export to match lctl's UUID
5329 - don't INC_USE_COUNT for phantom exports
5330 - don't crash when cleaning up phantom exports (567)
5331 - don't double-finish or set replay data for errored mdc_open requests
5332 - abort requests when they time out, so we don't get old replies
5333 - send/receive replies for AST messages again
5334 - if the client says that it doesn't have the lock, cancel it on the
5336 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5337 mark it as destroyed, it will all work out when decref is called
5338 - fix module use counts (22, 581)
5340 - ASTs now expect a reply (server cancels lock on error reply)
5342 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5345 - fix many simultaneous client startup (392)
5346 - fix dentry->d_it clobbering
5347 - credentials weren't being shipped for readdir/getattr operations
5348 - remove invalid assertions triggered during some concurrent MD
5350 - proper Lustre versions added (336, 389)
5351 - fix memory leak for create error case (398)
5352 - fix LOV locking bug that would get cli/srv out of sync
5353 - fix echo client over LOV (409)
5354 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5355 - fix double-O_EXCL intent crash (424)
5356 - avoid sending multiple lock CANCELs (352)
5358 - MDS can do multi-client recovery (modulo bugs in new code)
5360 - many updates, edits, cleanups
5362 2002-11-18 Phil Schwan <phil@clusterfs.com>
5365 - fix null d_it dereference (346)
5366 - fix full OST/dbench hang (333)
5367 - fix permission problem with file removal (286)
5368 - fix removal of OSCs from LOV when they fail
5369 - fix NULL deref during bulk timeout (214)
5370 - fix problems related to multiple filesystems on one MDS (241)
5371 - fixed serious subtle metadata locking bugs
5372 - free locks on clients when inodes are removed due to memory
5374 - fix inode pointer in lock data (285)
5375 - partial support for multiple MDS on a single host (241)
5376 - data locks weren't cancelled at clear_inode time (290, 311)
5377 - intent locks could lead to unbounded lock growth (205)
5378 - added a maximum lock count, an LRU list, and a flusher
5379 - fix multiple rename (365)
5380 - properly abstracted the echo client
5381 - OSC locked 1 byte too many; fixed
5382 - rewrote brw callback code:
5383 - fixed recovery bugs related to LOVs (306)
5384 - fixed too-many-pages-in-one-write crash (191)
5385 - fixed (again) crash in sync_io_timeout (214)
5386 - probably fixed callback-related race (385)
5388 - Add capability to MDS protocol
5389 - LDLM cancellations and callbacks on different portals
5391 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5394 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5395 - defer open object destruction to close time (601981, 138)
5396 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5397 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5398 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5399 - don't call obd_setattr in ll_file_release for destroyed objects
5400 * protocol change to lustre_msg: move |version| and add |flags|
5401 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5402 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5403 * added replay of create, unlink, link and rename operations during
5404 MDS failover; recovery should be much more robust now
5405 * remove failed OSCs from LOVs (only lov_create uses this so far)
5406 * the lustre-HOWTO was brought (more) up to date (582544)
5408 2002-10-23 Phil Schwan <phil@clusterfs.com>
5411 - in-use dentries weren't being reused properly (617851)
5412 - prevent multiple LDLM setup (599178)
5413 - fix LOV size calculations for truncate (617853)
5414 - fix client handling of MDS intent errors (POSIX)
5415 - fix permission bug in lovstripe.c test (624321)
5416 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5417 - truncate past end of file could corrupt data
5418 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5419 - a race in recovery could return ETIMEDOUT to apps (623947)
5420 - building outside the source directory was fixed
5421 * the lustre-HOWTO was brought (more) up to date (582544)
5422 * major progress was made on recovery functionality
5424 2002-10-10 Phil Schwan <phil@clusterfs.com>
5427 - recovery deadlock fix
5428 - rm -rf causes LBUG fix (617817)
5429 - file open by multiple tasks fix (618962)
5430 - directory permissions bugs (602707 and 620007)
5431 - journal_stop fixed with locking (611313)
5432 - O_APPEND failures resolved (618273, perhaps 614459)
5433 - lconf PATH fix (619770)
5434 - IA64 build fix (621450)
5435 - RPC buffer sizes scale with amount of memory
5437 2002-10-01 Phil Schwan <phil@clusterfs.com>
5440 - locks would be cancelled without throwing away data pages,
5441 resulting in inconsistent data (605627)
5442 - inode attributes were not always being refreshed (605627, 612449)
5443 - lconf now continues to cleanup after lctl reports an error
5444 - MDS now enforces user permissions (602707)
5445 - lprocfs cleanup fixed, but not yet enabled (614157)
5446 - fixed infinite server hang, should a client not respond to an AST
5447 - avoid going into recovery if user calls readlink() with a buffer
5448 that's too small (613941)
5449 - AST RPCs no longer require replies (614867) -- this may be changed
5450 - don't crash server if client sends an IOV that's too big (611336)
5451 - fixed lock conversion deadlock (611892)
5452 - fixed the following of symlinks (614622)
5453 * recovery: the server can remove locks from a client that dies, other
5454 clients can make progress
5455 * more extN patch fixes
5456 * compile-time configurable ptlrpc buffer allocations
5458 - collaborative read cache document
5459 - Lustre Lite Performance CDR document-in-progress
5461 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5464 - fix typo in patch-2.4.18
5466 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5469 - clear ptlrpc request each time in handle_incoming_request()
5470 - unlink of files now destroys the object on the OST
5472 2002-09-19 Peter Braam <braam@clusterfs.com>
5474 * add hard link support
5475 * change obdfile creation method
5476 * kernel patch changed
5478 2002-09-19 Peter Braam <braam@clusterfs.com>
5481 - stack overflow bug in extN fixed
5483 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5485 * documentation updates
5486 - add man pages for config tools
5487 - update tests/README to describe testing with new config tools
5488 - finish metadata API descriptions
5489 * bug fixes and cleanups
5490 - statfs workaround for 16TB limit
5491 - LOV stripe allocation improved, can stripe on subset of OSTs
5492 - LOV file size/IO offset was wrong for files > 4GB in size
5493 - object EA data was being dropped, caused files to be unreadable
5494 - memory overflow with non-LOV OST caused memory corruption
5495 - fixed regression tests to work with new config tools, obdfilter
5496 - fixed bug when directory size became larger than 1 block
5497 - fixed bug (for single client case) when PWD was deleted
5498 - invalidate local directory pages when doing intent-based ops
5499 - avoid LDLM oops when lock callback contained bad data
5501 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5503 * documentation updates
5504 * bug fixes and cleanups
5505 - configuration tools
5508 - 64-bit compile warnings
5509 - 64-bit internal statfs data
5511 * test_brw on persistent OST devices
5513 * lprocfs (disabled)
5515 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5517 * documentation updates
5518 * bug fixes and cleanups
5519 * configuration tools
5521 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5524 * many small fixes to 0_5_4
5525 * io/network handling
5526 * thinkos in MDS operations
5528 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5531 * crucial basic fixes to 0.5.3
5532 * IOR, Iozone work over Elan
5535 2002-08-07 Phil Schwan <phil@clusterfs.com>
5536 * version 0_5_3, our first alpha
5537 * we use the new Portals iovs
5538 * documentation updates
5539 * bug fixes and cleanups
5540 * small changes in the DLM wire protocol
5542 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5543 * version 0_5_1 with some initial stability,
5544 * locking on MD and file I/O.
5545 * documentation updates
5546 * several bug fixes since 0.5.0
5547 * small changes in wire protocol
5549 2002-07-18 Phil Schwan <phil@clusterfs.com>
5551 * delivered as Lustre Light Alpha
5552 * fixed a crash after handling invalid MDS requests
5553 * fixed directory pages for architectures with non-4k pages sizes
5555 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5556 * release version v0_4_4
5557 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5559 * Fixes a number of interruption problems with OST operations.
5560 * Update documentation for portals header changes
5561 * Move all wire protocol structs/defines to lustre_idl.h
5562 * Fixes symlink length bug.
5563 * Add tcpdump to repository.
5565 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5566 * release version v0_4_3
5567 * Fixes statfs for inodes on extN.
5568 * Fixes bug in runtests which would delete /etc/hosts.
5569 * Use 64-bit object IDs wherever possible (not into VFS though)
5570 Remove ost_get_info, which is unused by lustre, and out of date.
5572 2002-07-03 Peter Braam <braam@clusterfs.com>
5573 * release version v0_4_2 Fixes a lookup error (type not passed)
5574 * move forward to head of Portals
5575 * move forward to latest Lustre kernel
5577 2002-06-25 Peter Braam <braam@clusterfs.com>
5578 * release version v0_4_1. Hopefully stable on single node use.