1 tbd Sun Microsystems, Inc.
5 2.6.9-67.0.7.EL (RHEL 4),
6 2.6.16.54-0.2.5 (SLES 10),
7 2.6.18-53.1.14.el5 (RHEL 5),
8 2.6.18.8 vanilla (kernel.org)
9 2.6.22.14 vanilla (kernel.org)
10 * Client support for unpatched kernels:
11 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
12 2.6.16 - 2.6.22 vanilla (kernel.org)
13 * Client support for unpatched kernels:
14 we do not recommend using patchless RHEL4 clients with kernels
15 prior to 2.6.9-55EL (RHEL4U5).
16 * Recommended e2fsprogs version: 1.40.7-sun1
17 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
18 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
19 removed cwd "./" (refer to Bugzilla 14399).
20 * A new quota file format has been introduced in 1.6.5.
21 The format conversion from prior releases is handled transparently,
22 but releases older than 1.4.12/1.6.5 will not understand this new
23 format. The automatic format conversion can be avoided by running
24 the following command on the MDS:
25 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
26 For more information, please refer to bugzilla 13904.
30 Frequency : rare, on shutdown ost
31 Description: don't hit live lock with umount ost.
32 Description: shrink_dcache_parent can be in long loop with destroy dentries,
33 use shrink_dcache_sb instead.
37 Description: don't panic with use echo client
38 Details : echo client pass NULL as client nid pointer and this produce null
43 Description: fix build on ppc32
44 Details : compile code with -m64 flag produce wrong object file for ppc32.
48 Description: Don't update lov_desc members until making sure they are valid
49 Details : When updating lov_desc members via proc fs, need fix their
50 validities before doing the real update.
52 -------------------------------------------------------------------------------
55 04-26-2008 Sun Microsystems, Inc.
57 * Support for kernels:
59 2.6.9-67.0.7.EL (RHEL 4),
60 2.6.16.54-0.2.5 (SLES 10),
61 2.6.18-53.1.14.el5 (RHEL 5),
62 2.6.18.8 vanilla (kernel.org)
63 2.6.22.14 vanilla (kernel.org)
64 * Client support for unpatched kernels:
65 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
66 2.6.16 - 2.6.22 vanilla (kernel.org)
67 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
68 we do not recommend using patchless RHEL4 clients with kernels
69 prior to 2.6.9-55EL (RHEL4U5).
70 * Recommended e2fsprogs version: 1.40.7-sun1
71 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
72 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
73 removed cwd "./" (refer to Bugzilla 14399).
74 * A new quota file format has been introduced in 1.6.5.
75 The format conversion from prior releases is handled transparently,
76 but releases older than 1.4.12/1.6.5 will not understand this new
77 format. The automatic format conversion can be avoided by running
78 the following command on the MDS:
79 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
80 For more information, please refer to bugzilla 13904.
84 Description: add message levels for liblustreapi
88 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
89 Details : Move the 'good_osts' check before the 'total_bavail' check. This
90 will result in an -EAGAIN and in the exit call path we call
91 alloc_rr() which will with increasing aggressiveness attempt to
92 aquire precreated objects on the minimum number of required OSCs.
96 Description: Use old size assignment to avoid deadlock
97 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
98 the scheduling while holding a spinlock. We do not need locking
99 for size in ll_update_inode() because size is only updated from
100 the MDS for directories or files without objects, so there is no
101 other place to do the update, and concurrent access to such inodes
102 are protected by the inode lock.
106 Description: Use __u64 instead of int for valid bits
110 Description: resolve "_IOWR redefined" build error on SLES10
114 Description: dump the memory debugging after all modules are unloaded to
115 suppress false negative in conf_sanity test 39
119 Description: the recovery timer never expires
120 Details : for new client connect request, the recovery timer should not be
121 reset, otherwise recovery timer will never expired, if the old
122 client never come. Only old client connect and first connection
123 req should trigger recovery timer reset.
127 Description: the min numbers of lproc stats are wrong
128 Details : adding a new constant LC_MIN_INIT and use it for initialization
133 Description: misc fixes (including bug 15169 and 15175) for directory readahead.
134 Details : prevent previous statahead async RPC callback from processing the
135 current "statahead_info", race condition between async RPC callback
136 add dentry into dentry hash table and "ls" thread revalidate such
137 dentry, statahead his/miss control for hidden items, and so on.
139 Severity : enhancement
141 Description: build kernel-ib packages for OFED 1.3 in our release cycle
145 Description: incore types cleaning in quota code (with respect to 64-bit limits)
146 Details : several u32 variables declarations are replaced with u64 declarations
151 Description: fix SLES kernel versioning
152 Details : the kernel version for our SLES 10 kernel did not include a "-"
153 before the "smp" at the end. while this was not a problem in
154 general, it did mean that software trying to use the kernel
155 version to try to detect a vendor specific kernel would fail.
156 this was most evident by the OFED build scripts.
160 Description: don't put request into delay list while invalidate in flight.
161 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
162 import in flight. this produce timeout for invalidate and sometimes
163 can cause stale data.
165 Severity : enhancement
167 Description: Update kernel to SLES9 2.6.5-7.311.
169 Severity : enhancement
171 Description: Update kernel to RHEL4 2.6.9-67.0.7.
175 Frequency : on ppc only
176 Description: not convert ost objects for directory because it's not exist.
177 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
181 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
182 __u64 as usingied long long type.
186 Frequency : rare, on shutdown
187 Description: race process ast vs remove callback
188 Details : removing callback before disconnect import open race with processing
191 Severity : enhancement
193 Description: Update kernel to SLES9 2.6.5-7.311.
197 Description: Add FMODE_EXEC to SLES10 SP1 series.
199 Severity : enhancement
201 Description: Add server support for vanilla-2.6.22.14.
205 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
206 Details : If an OST is down the MDS will hang indefinitely in
207 obd_statfs_rqset() waiting for the statfs data. While for
208 MDS QOS usage of statfs, it should not stuck in waiting.
212 Description: Disable adaptive timeouts by default
216 Frequency : on network error
217 Description: panic with double free request if network error
218 Details : mdc_finish_enqueue is finish request if any network error occuring,
219 but it's true only for synchronus enqueue, for async enqueue
220 (via ptlrpcd) this incorrect and ptlrpcd want finish request
225 Frequency : rare, on recovery
226 Description: read procfs can produce deadlock in some situation
227 Details : Holding lprocfs lock with send rpc can produce block for destroy
228 obd objects and this also block reconnect with -EALREADY.
229 This isn't fix all lprocfs bugs - but make it rare.
231 Severity : enhancement
233 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
236 Frequency : frequent on X2 node
238 Description: mdc_set_open_replay_data LBUG
239 Details : Set replay data for requests that are eligible for replay.
243 Description: lustre_mgs: operation 101 on unconnected MGS
244 Details : When MGC is disconnected from MGS long enough, MGS will evict the
245 MGC, and late on MGC cannot successfully connect to MGS and a lot
246 of the error messages complaining that MGS is not connected.
249 Frequency : rare, depends on device drivers and load
251 Description: MDS or OSS nodes crash due to stack overflow
252 Details : Code changes in 1.6.4 increased the stack usage of some functions.
253 In some cases, in conjunction with device drivers that use a lot
254 of stack, the MDS (or possibly OSS) service threads could overflow
255 the stack. One change which was identified to consume additional
256 stack has been reworked to avoid the extra stack usage.
258 Severity : enhancement
260 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
262 Severity : enhancement
264 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
266 Severity : enhancement
268 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
270 Serverity : enhancement
272 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
275 Frequency : rare on shutdown ost
277 Description: Don't allow skipping OSTs if index has been specified.
278 Details : Don't allow skipping OSTs if index has been specified, make locking
279 in internal create lots better.
283 Description: ASSERTION(!PageDirty(page)) failed
284 Details : Wrong check could lead to an assertion failure under specific
289 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
290 Details : access to bitfield in structure is always rounded to long
291 and this produce problem with not atomic change any bit.
295 Description: Lustre make rpms failed.
296 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
297 builds Lustre rpms from tarball.
299 Severity : enhancement
301 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
304 Frequency : rare on shutdown ost
306 Description: If llog cancel was not send before clean_exports phase, this can
307 produce deadlock in llog code.
308 Details : If llog thread has last reference to obd and call class_import_put
309 this produce deadlock because llog_cleanup_commit_master wait when
310 last llog_commit_thread exited, but this never success because was
311 called from llog_commit_thread.
315 Description: NULL lov_tgts causing MDS oops
316 Details : more safe checks for NULL lov_tgts for avoid oops.
318 Severity : enhancement
320 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
322 Severity : enhancement
324 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
328 Description: make mgs_setparam() handle fsname containing dash
329 Details : fsname containing a dash does not work with lctl conf_param
331 Severity : enhancement
333 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
337 Frequency : rare, in recovery and (or) destroy lovobjid file.
338 Description: rewrite lov objid code.
339 Details : Cleanup for lov objid code, remove scability problems and wrong
340 locking. Fix sending last_id into ost.
342 Severity : enhancement
344 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
346 Severity : enhancement
348 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
349 Details : Use d_move_locked instead of __d_move.
353 Frequency : rare, at shutdown
354 Description: access already free / zero obd_namespace.
355 Details : if client_disconnect_export was called without force flag set,
356 and exist connect request in flight, this can produce access to
357 NULL pointer (or already free pointer) when connect_interpret
358 store ocd flags in obd_namespace.
362 Frequency : only at startup
363 Description: not alloc memory with spinlock held.
364 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
365 if any spinlock held.
370 Description: lfs find does not continue on file error
371 Details : Continue other files processing when a file/dir is absent.
375 Description: Inconsistent usage of lustre_pack_reply()
376 Details : Standardize the usage of lustre_pack_reply() such that it
377 always generate a CERROR on failure.
380 Frequency : very rare
382 Description: Fix replay if there is an un-replied request and open
383 Details : In some cases, older replay request will revert the
384 mcd->mcd_last_xid on MDS which is used to record the client's
387 Severity : enhancement
389 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
391 Severity : enhancement
393 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
395 Severity : enhancement
397 Description: Update to SLES9 kernel-2.6.5-7.287.3.
399 Severity : enhancement
401 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
403 Severity : enhancement
405 Description: Build SLES10 patchless client fails
406 Details : The configure was broken by run ./configure with
407 --with-linux-obj=.... argument for patchless client. When the
408 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
409 can't search header adequately. Use absolute path such as
410 -I($LINUX)/include instead.
412 Severity : enhancement
414 Description: Lustre Page Accounting
415 Details : New macros for page alloc and free which enable accounting
416 of page allocation of Lustre. Use percpu counters to store memory
420 Frequency : only if debugging is disabled
422 Description: LASSERT_{REQ,REP}SWAB macros are buggy
423 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
424 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
425 these macros and replace them with their difinitions instead.
430 Description: interrupt oig_wait produce painc on resend.
431 Details : brw_redo_request can be used for resend requests from ptlrpcd and
432 private set, and this produce situation when rq_ptlrpcd_data not
433 copyed to new allocated request and triggered LBUG on assert
434 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
435 ptlrpcd set if request is changed and can be safety changed to use
438 Severity : enhancement
440 Description: organize the server-side client stats on per-nid basis
441 Details : Change the structure of stats under obdfilter and mds to
449 The "uuid"s file would list the uuids of _active_ exports.
450 And the clear entry is to clear all stats and stale nids.
455 Description: Processes looping in ll_readdir() on Lustre clients finally causing
456 a full node pseudo-hang
457 Details : Concurrent access to the same directory from multiple clients with
458 intensive file creation/removal can cause a client node to spin in
459 ll_readdir(). i_version must be increased every time the lock is
460 cancelled to ensure a revalidate is done.
465 Description: touch file failed when fs is not full
466 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
467 otherwise we can get ENOSP while fs is not full.
470 Frequency : only for Cray XT3
471 Bugzilla : 12829/13455
472 Description: Changing primary group doesn't change the group lustre assigns to
474 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
475 overridden with the primary group provided by the group upcall,
476 whereas the supplied fsgid can be trusted if it is in the list of
477 supplementary groups returned by the group upcall.
481 Description: Root Squash Functionality
482 Details : Implementation of NFS-like root squash capability. Specifically,
483 don't allow someone with root access on a client node to be able
484 to manipulate files owned by root on a server node.
488 Description: Slow trucate/writes to huge files at high offsets.
489 Details : Directly associate cached pages to lock that protect those pages,
490 this allows us to quickly find what pages to write and remove
491 once lock callback is received.
495 Description: Too many locks accumulating on client during NFS usage
496 Details : mds_open improperly used accmode to find out access mode to a
497 file. Also mdc_intent_lock logic to find out if we already have
498 lock similar to just received was flawed since introduction of
499 skiplists - locks are now added to the front of the granted
504 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
506 Details : In obd setup failure handler, obd_stopping will not necessarily to
507 be 1, and obd_set_up should also be checked to make sure whether
508 obd is completely setup.
510 Severity : enhancement
512 Description: Allow masking D_WARNING, D_ERROR messages from console
513 Details : Console messages can now be disabled via lnet.printk.
517 Description: User code with malformed file open parameter crashes client node
518 Details : Before packing join_file req, all the related reference should be
519 checked carefully in case some malformed flags cause fake join_file
525 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
526 grain of quota doing harm to quota's accuracy
527 Details : qunit size will be changed when quota limitation is too low/high;
528 record the pending quota write in order to get more accureate
529 quota; delete the patch for bug12588, which is unnecessary when
530 this patch is landed. This bug also contains fixes for bug 14526,
531 14299, 14601 and 13794.
535 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
536 Details : ldlm_completion_ast() assumes that a lock is granted when the req
537 mode is equal to the granted mode. However, it should also check
538 that LDLM_FL_CP_REQD is not set.
542 Description: Heavy nfs access might result in deadlocks
543 Details : After ELC code landed, it is now improper to enqueue any mds
544 locks under och_sem, because enqueue might want to decide to
545 cancel open locks for same inode we are holding och_sem for.
549 Description: 35% write performance drop with ldiskfs2 when quotas are on
550 Details : Enable ext3 journalled quota by default to improve performance
551 when quotas are turned on.
555 Description: Client eviction while running blogbench
556 Details : A lot of unlink operations with concurrent I/O can lead to a
557 deadlock causing evictions. To address the problem, the number of
558 oustanding OST_DESTROY requests is now throttled to
559 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
560 callbacks are processed in priority.
563 Frequency : RHEL4 only
565 Description: mkfs is very slow on IA64/RHEL4
566 Details : A performance regression has been discovered in the MPT Fusion
567 driver between versions 3.02.73rh and 3.02.99.00rh. As a
568 consequence, we have downgraded the MPT Fusion driver in the RHEL4
569 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
572 Frequency : PPC/PPC64 only
574 Description: conflicts between asm-ppc64/types.h and lustre_types.h
575 Details : fix duplicated definitions between asm-ppc64/types.h and
576 lustre_types.h on PPC.
579 Frequency : PPC/PPC64 only
581 Description: asm-ppc/segment.h does not exist
582 Details : fix compile issue on PPC.
586 Description: data checksumming impacts single node performance
587 Details : add support for several checksum algorithms. Currently, CRC32 and
588 Adler-32 are supported. The checksum type can be changed on the fly
589 through /proc/fs/lustre/osc/*/checksum_type.
593 Description: use adler32 for page checksums
594 Details : when available, use the Adler-32 algorithm instead of CRC32 for
599 Description: better handle error messages in extents code
601 Severity : enhancement
603 Description: SNMP support enhancement
604 Details : Adding total number of sampled request for an MDS node in snmp
607 Severity : enhancement
609 Description: Optimize ldlm waiting list processing for PR extent locks
610 Details : When processing waiting list for read extent lock and meeting read
611 lock that is same or wider to it that is not contended, skip
612 processing rest of the list and immediatelly return current
613 status of conflictness, since we are guaranteed there are no
614 conflicting locks in the rest of the list.
618 Description: Time out and refuse to reconnect
619 Details : When the failover node is the primary node, it is possible
620 to have two identical connections in imp_conn_list. We must
621 compare not conn's pointers but NIDs, otherwise we
622 can defeat connection throttling.
626 Description: Client not clear own cache if answer to reconnect is lost.
627 Details : Client gets evicted from server. Now client also thinks it is
628 disconnected (or gets ENOTCONN on its operation) and decides to
629 reconnect. Server receives reconnect message, but cannot find
630 export. New export is created that is fully valid (new cookie!),
631 but reply is lost and not reported to client. Client reconnects
632 again and gets back a just-created connection, but it is not new
633 so client thinks it was not evicted and does not do recovery.
637 Description: Detect stride IO mode in read-ahead
638 Details : When a client does stride read, read-ahead should detect that and
639 read-ahead pages according to the detected stride pattern.
643 Description: build for x2 fails
644 Details : fix compile issue on Cray systems.
646 Severity : enhancement
648 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
649 Details : This greatly improves speed of NFS writes on 2.6 kernels.
652 Frequency : only on PPC/SLES10
654 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
655 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
656 be defined. Add a hack in mkfs_lustre.c to work around this compile
661 Description: LASSERT on MDS when client holding flock lock dies
662 Details : ldlm pool logic depends on number of granted locks equal to
663 number of released locks which is not true for flock locks, so
664 just exclude such locks from consideration.
668 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
669 Details : Use fsfilt_sync() for both the whole filesystem sync and
670 individual file sync to eliminate dangerous inode locking
671 with I_LOCK that can lead to a deadlock.
675 Description: Update an obsolete wirecheck.c generator
676 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
680 Description: Client can panic on open sometimes
681 Details : It is possible that we try to free already freed request in
682 ll_file_open in some error cases when we send request from
687 Description: performance in 1.6.3
688 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
689 sure that 1MB requests are not fragmented by the block layer.
693 Description: LDLM soft lockups - improvement
694 Details : It is be possible to send the lock handle along with each read
695 or write request because the client is already doing a lock match
696 itself so there isn't any reason the OST should have to re-do that
699 --------------------------------------------------------------------------------
701 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
703 * Support for kernels:
704 2.6.5-7.286 (SLES 9),
705 2.6.9-55.0.9.EL (RHEL 4),
706 2.6.16.53-0.8 (SLES 10),
707 2.6.18-8.1.14.el5 (RHEL 5),
708 2.6.18.8 vanilla (kernel.org)
709 * Client support for unpatched kernels:
710 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
711 2.6.16 - 2.6.22 vanilla (kernel.org)
712 * Due to recently discovered recovery problems, we do not recommend
713 using patchless RHEL 4 clients with this or any earlier release.
714 * Recommended e2fsprogs version: 1.40.2-cfs1
715 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
718 Frequency : occasional
720 Description: MDS hang or stay in waiting lock
721 Details : If client receive lock with CBPENDING flag ldlm need send lock
722 cancel as separate rpc, to avoid situation when cancel request
723 can't processed due all i/o threads stay in wait lock.
726 Frequency : occasional
728 Description: Do not fail import if osc_interpret_create gets -EAGAIN
729 Details : If osc_interpret_create got -EAGAIN it immediately exits and
730 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
731 oscc_has_objects and see OSC has no objests and call
732 oscc_internal_create to resend create request.
734 Severity : enhancement
736 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
737 Details : Update which_patch & target file for SLES10 latest kernel.
739 Severity : enhancement
741 Description: add --type and --size parameters to lfs find
742 Details : Enhance lfs find by adding filetype and filesize parameters. Also
743 multiple OBDs can now be specified for the --obd option.
745 Severity : enhancement
747 Description: eliminate client locks in face of contention
748 Details : file contention detection and lockless i/o implementation
751 Severity : enhancement
753 Description: Remove client patches from SLES 10 kernel.
754 Details : This causes SLES 10 clients to behave as patchless clients
755 even on a Lustre-patched (server) kernel.
757 Severity : enhancement
759 Description: use i_size_read and i_size_write in 2.6 port
760 Details : replace inode->i_size access with i_size_read/write()
763 Frequency : when removing large files
765 Description: scheduling issue during removal of large Lustre files
766 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
767 It causes scheduling issues when removing large files (17TB in the
773 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
774 Details : Impossible (EPROTO is returned) to access a directory that has a
775 non-default striping and ACLs.
778 Frequency : only on ppc
780 Description: /proc/fs/lustre/devices broken on ppc
781 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
782 We should make sure the type of 'index' is loff_t and then cast
783 explicitly as needed below. Do not assign an explicitly cast
787 Frequency : only for rhel5
789 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
790 Details : Modify the target file & which_kernel.
793 Frequency : if the uninit_groups feature is enabled on ldiskfs
795 Description: e2fsck reports "invalid unused inodes count"
796 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
797 feature and only a single inode is created in a group then the
798 "bg_unused_inodes" count is incorrectly updated. Creating a
799 second inode in that group would update it correctly.
804 Description: buffer overruns could theoretically occur
805 Details : llapi_semantic_traverse() modifies the "path" argument by
806 appending values to the end of the origin string, and a buffer
807 overrun may occur. Adding buffer overrun check in liblustreapi.
809 Severity : enhancement
811 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
814 Frequency : only if filesystem is inconsistent
816 Description: handle "serious error: objid * already exists" more gracefully
817 Details : If LAST_ID value on disk is smaller than the objects existing in
818 the O/0/d* directories, it indicates disk corruption and causes an
819 LBUG(). If the object is 0-length, then we should use the existing
820 object. This will help to avoid a full fsck in most cases.
822 Severity : enhancement
824 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
825 Details : Modify vm-tunables-rhel4.patch.
827 Severity : enhancement
829 Description: Kernel config for 2.6.18-vanilla.
830 Details : Modify targets/2.6-vanilla.target.in.
831 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
832 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
833 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
834 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
837 Frequency : occasional
839 Description: improve handling recoverable errors
840 Details : If request processed with error which can be recoverable on server
841 request should be resend, otherwise page released from cache and
847 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
848 Details : Modify target file & which_patch.
849 A flaw was found in the IA32 system call emulation provided
850 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
851 value could be stored in the %RAX register, which could trigger an
852 out-of-bounds system call table access. An untrusted local user
853 could exploit this flaw to run code in the kernel
854 (ie a root privilege escalation). (CVE-2007-4573).
858 Description: change order of libsysio includes
859 Details : '#include sysio.h' should always come before '#include xtio.h'
861 Severity : enhancement
863 Description: adapt the lustre_config script to support the upgrade case
864 Details : Add "-u" option for lustre_config script to support upgrading 1.4
865 server targets to 1.6 in parallel.
870 Description: To avoid grant space > avaible space when the disk is almost
871 full. Without this patch you might see the error "grant XXXX >
872 available" or some LBUG about grant, when the disk is almost
874 Details : In filter_check_grant, for non_grant cache write, we should
875 check the left space by if (*left > ungranted + bytes), instead
876 of (*left > ungranted), because only we are sure the left space
877 is enough for another "bytes", then the ungrant space should be
878 increase. In client, we should update cl_avail_grant only there is
879 OBD_MD_FLGRANT in the reply.
884 Description: Update RHEL 4 kernel to fix local root privilege escalation.
885 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
886 described in CVE-2007-4573. This problem could allow untrusted
887 local users to gain root access.
890 Frequency : when using O_DIRECT and quotas
892 Description: Incorrect file ownership on O_DIRECT output files
893 Details : block usage reported by 'lfs quota' does not take into account
894 files that have been written with O_DIRECT.
896 --------------------------------------------------------------------------------
898 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
900 * Support for kernels:
901 2.6.5-7.286 (SLES 9),
902 2.6.9-55.0.2.EL (RHEL 4),
903 2.6.16.46-0.14 (SLES 10),
904 2.6.18-8.1.8.el5 (RHEL 5),
905 2.6.18.8 vanilla (kernel.org)
906 * Client support for unpatched kernels:
907 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
908 2.6.16 - 2.6.21 vanilla (kernel.org)
909 * Due to recently discovered recovery problems, we do not recommend
910 using patchless RHEL 4 clients with this or any earlier release.
911 * Recommended e2fsprogs version: 1.40.2-cfs1
912 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
916 Description: Fix errors in lfs documentation
917 Details : Fixes man pages
919 Severity : enhancement
921 Description: Adaptive timeouts
922 Details : RPC timeouts adapt to changing server load and network
923 conditions to reduce resend attempts and improve recovery time.
925 Severity : enhancement
927 Description: llapi_file_create() does not allow some changes
928 Details : add llapi_file_open() that allows specifying the file creation
929 mode and open flags, and also returns an open file handle.
931 Severity : enhancement
933 Description: df doesn't work properly if diskfs blocksize != 4K
934 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
937 Frequency : always on directories with default striping set
939 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
940 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
942 Severity : enhancement
944 Description: merge and cleanup kernel patches.
945 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
947 Severity : enhancement
949 Description: RedHat Update kernel for RHEL5
950 Details : Kernel config file for RHEL5.
952 Severity : enhancement
954 Description: OSS needs mutliple precreate threads
955 Details : Add ability to start more than one create thread per OSS.
958 Frequency : only with quota on the root user
960 Description: mds_obd_create error creating tmp object
961 Details : When the user sets quota on root, llog will be affected and can't
962 create files and write files.
965 Frequency : Always on ia64 patchless client, and possibly others.
967 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
968 Details : This allows the patchless client to be loaded on architectures
974 Description: disorder of journal start and llog_add cause deadlock.
975 Details : in llog_origin_connect, journal start should happen before
976 llog_add keep the same order as other functions to avoid
979 Severity : enhancement
981 Description: RedHat Update kernel for RHEL5
982 Details : Modify the kernel config file more closer RHEL5.
984 Severity : enhancement
986 Description: Build failure against Centos5 (RHEL5)
987 Details : Define PAGE_SIZE when it isn't present.
990 Frequency : occasionally when using NFS
992 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
993 Details : with NFS, the anon dentry's parent was set to itself in
994 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
995 corresponding dentry other than use rec->ur_name.
997 Severity : enhancement
999 Description: enable data checksumming by default
1000 Details : enable checksum by default, allow --disable-checksum
1001 configure option and "-o nochecksum" mount option. Checksums
1002 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1005 Frequency : Occasionally with failover
1007 Description: Client eviction due to failover config
1008 Details : after a connection loss, the lustre client should attempt to
1009 reconnect to the last active server first before trying the
1010 other potential connections.
1012 Severity : enhancement
1014 Description: client-side metadata stat-ahead during readdir(directory readahead)
1015 Details : perform client-side metadata stat-ahead when the client detects
1016 readdir and sequential stat of dir entries therein
1021 Description: when mds and osts use different quota unit(32bit and 64bit),
1022 quota will be released repeatly.
1023 Details : void sending multiple quota reqs to mds, which will keep the
1024 status between the reqs.
1027 Frequency : only with liblustre clients on XT3
1029 Description: evictions taking too long
1030 Details : allow llrd to evict clients directly on OSTs
1034 Description: osts not allocated evenly to files
1035 Details : change the condition to increase offset_idx
1037 Severity : enhancement
1039 Description: self-adjustable client's lru lists
1040 Details : use adaptive algorithm for managing client cached locks lru
1041 lists according to current server load, other client's work
1042 pattern, memory activities, etc. Both, server and client
1043 side namespaces provide number of proc tunables for controlling
1048 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
1049 Details : rewrite inherited code (uuid parsing code from ext2 utils and
1050 readdir code from ext3) from scratch preserving functionality.
1054 Description: Only those disconnect error should be returned by rq_status.
1055 Details : In open/enqueue processs, Some errors, which will cause client
1056 disconnected, should be returned by rq_status, while other
1057 errors should still be returned by intent, then mdc or llite will
1060 Severity : enhancement
1062 Description: Tune the kernel for good SCSI performance.
1063 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1064 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1068 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1069 Bugzilla : 13547, 13627
1070 Description: Data corruption for OSTs that are formatted larger than 2TB
1072 Details : When generating the bio request for lustre file writes the
1073 sector number would overflow a temporary variable before being
1074 used for the IO. The data reads correctly from Lustre (which
1075 will overflow in a similar manner) but other file data or
1076 filesystem metadata may be corrupted in some cases.
1080 Description: "lfs find -obd UUID" prints directories
1081 Details : "lfs find -obd UUID" will return all directory names instead
1082 of just file names. It is incorrect because the directories
1083 do not reside on the OSTs.
1086 Bugzilla : 13596, 13614
1087 Description: MDS hang after unclean shutdown of lots of clients
1088 Details : Never resend AST requests.
1092 Frequency : Always, for kernels after 2.6.16
1093 Description: Fix warning idr_remove called for id=.. which is not allocated.
1094 Details : Last kernels save old s_dev before kill super and not allow
1095 to restore from callback - restore it before call kill_anon_super.
1099 Description: Fix error on 'ls .' at the top of the Lustre mount.
1100 Details : Don't revalidate dentry if it is a root dentry.
1102 --------------------------------------------------------------------------------
1104 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
1106 * Support for kernels:
1107 2.6.5-7.286 (SLES 9),
1108 2.6.9-55.0.2.EL (RHEL 4),
1109 2.6.16.46-0.14 (SLES 10),
1110 2.6.18-8.1.8.el5 (RHEL 5),
1111 2.6.18.8 vanilla (kernel.org)
1112 * Client support for unpatched kernels:
1113 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1114 2.6.16 - 2.6.22 vanilla (kernel.org)
1115 * Due to recently discovered recovery problems, we do not recommend
1116 using patchless RHEL 4 clients with this or any earlier release.
1117 * Recommended e2fsprogs version: 1.39.cfs8
1118 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1123 Description: block reactivating mgc import until all deactivates complete
1124 Details : Fix race when failing back MDT/MGS to itself (testing)
1126 Severity : enhancement
1128 Description: lfs setstripe enhancement
1129 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1132 Frequency : mds/oss recovery
1134 Description: llog ctxt is refrenced after it has been freed.
1135 Details : llog ctxt refcount was added to avoide the race between ctxt free
1136 and llog recovery process. Each llog user must hold ctxt refcount
1137 before it access the llog. And the llog ctxt can only be freed
1138 when its refcount is zero.
1140 Severity : enhancement
1142 Description: randomly memory allocation failure util
1143 Details : Make lustre randomly failed allocating memory for testing purpose.
1146 Frequency : only for SLES10
1148 Description: Update kernel patch for SLES10 SP1
1149 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
1151 Severity : enhancement
1153 Description: omit set fsid for export NFS
1154 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
1155 as FS_REQUIRES_DEV for avoid problems with generate fsid.
1158 Frequency : after network failures
1160 Description: Add sync option to mount_lustre.c
1161 Details : Client loses data written to lustre after a network interruption.
1163 Severity : enhancement
1165 Description: Error message improvement.
1166 Details : Merging of two LCONSOLE_ERROR_MSG into one.
1168 Severity : enhancement
1170 Description: don't use GFP_* in generic Lustre code.
1171 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1173 Severity : enhancement
1175 Description: obdclass is limited by single OBD_ALLOC(idarray)
1176 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1178 Severity : enhancement
1180 Description: updated patchess for new RHEL4 kernel
1181 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
1183 Severity : enhancement
1185 Description: warnings with build patchless client with vanila 2.6.19 and up
1186 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1187 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1189 Severity : enhancement
1191 Description: O_DIRECT bypasses client statistics.
1192 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1193 read_ahead_stats were updated. Copied stats section from
1194 osc_send_oap_rpc() into async_internal().
1197 Frequency : only for Cray XT3
1199 Description: peer credits not enough on many OST per OSS systems.
1200 Details : Use new lnet way to add credits as we need those for pings and ASTs
1203 Frequency : only with liblustre
1205 Description: Liblustre is not releasing flock locks on file close.
1206 Details : Release flock locks on file close.
1209 Frequency : only for RHEL4
1211 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
1212 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
1213 Update target file and kernel config.
1217 Description: lustre support for RHEL5
1218 Details : Add support for RHEL5.
1223 Description: ASSERTION(export != NULL) failed in target_handle_connect
1224 Details : Assetion hit is result of rare race between disconnect and connet
1225 to same nid. target_handle_connect found old connect cockie and
1226 tried to reconnect, but can't find export for this cockie.
1231 Description: umount blocks forever on error
1232 Details : In result of wrong using obd_no_recov and obd_force flags client
1233 can hand if cancel or some other requests is lost.
1236 Frequency : Only for SLES
1238 Description: sanity_quota fail test_1
1239 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1240 file, which makes TSTID[2] inunique.
1242 Severity : enhancement
1244 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1245 Details : Update target/ChangeLog/which_patch .
1247 Severity : enhancement
1249 Description: jbd statistics
1250 Details : Port older jbd statistics patch for sles10
1255 Description: lvbo_init failed for resource with missing objects.
1256 Details : Fix returning error if we do stat for file with missing/corrupted
1257 objects and i_size set to all sum of size of all avaible objects.
1258 if we to truncate/write to missing object - it is recreated.
1263 Description: Oops in read and write path when failing to allocate lock.
1264 Details : Check if lock allocation failed and return error back.
1267 Frequency : When flocks are used.
1269 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1270 Details : Flock locks might destroy just granted lock if it could be merged
1271 with another existing flock, this is done in completion handler,
1272 so teach ldlm_cli_enquque_fini that this is a valid case for
1278 Description: reply_lock_interpret crash due to race with it and lock cancel.
1279 Details : Do not replay locks that are being cancelled. Do not reference
1280 locks by their address during replay, just by their handle.
1283 Frequency : only with deactivated OSTs
1285 Description: lstripe command fails for valid OST index
1286 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
1287 instead of lov->desc.ld_active_tgt_count.
1289 Severity : enhancement
1291 Description: Build failure against Centos5 (RHEL5)
1292 Details : Use getpagesize() instead of PAGE_SIZE.
1294 --------------------------------------------------------------------------------
1296 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
1298 * Support for kernels:
1299 2.6.5-7.283 (SLES 9),
1300 2.6.9-55.EL (RHEL 4),
1301 2.6.16.46-0.14 (SLES 10),
1302 2.6.18.8 vanilla (kernel.org)
1303 * Client support for unpatched kernels:
1304 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1305 2.6.16 - 2.6.22 vanilla (kernel.org)
1306 * Due to recently discovered recovery problems, we do not recommend
1307 using patchless RHEL 4 clients with this or any earlier release.
1308 * Recommended e2fsprogs version: 1.39.cfs8
1309 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1310 * Starting with this release, the ldiskfs backing filesystem required
1311 by Lustre is now in its own package, lustre-ldiskfs. This package
1312 should be installed. It is versioned separately from Lustre and
1313 may be released separately in future.
1318 Description: server LBUG when shutting down
1319 Details : Block umount forever until the mount refcount is zero rather
1320 than giving up after an arbitrary timeout.
1322 Severity : enhancement
1324 Description: add optional extra BUILD_VERSION info
1325 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1326 to override the lustre version.
1329 Frequency : 2.6.18 servers only
1331 Description: ll_kern_mount() doesn't release the module reference
1332 Details : The ldiskfs module reference count never drops down to 0
1333 because ll_kern_mount() doesn't release the module reference.
1338 Description: server LBUG when using old ost_num_threads parameter
1339 Details : Accept the old ost_num_threads parameter but warn that it
1340 is deprecated, and fix an off-by-one error that caused an LBUG.
1345 Description: Transient SCSI error results in persistent IO issue
1346 Details : iobuf->dr_error is not reinitialized to 0 between two
1350 Frequency : sometimes when underlying device returns I/O errors
1352 Description: OSTs not going read-only during write failures
1353 Details : OSTs are not remounted read-only when the journal commit threads
1354 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1355 instead of the ext3 wrappers.
1358 Frequency : only with 10000 clients or more
1360 Description: poor connect scaling with increasing client count
1361 Details : Don't run filter_grant_sanity_check for more than 100 exports
1362 to improve scaling for large numbers of clients.
1365 Frequency : SLES10 only
1367 Description: sanity-quota.sh quotacheck failed: rc = -22
1368 Details : Quotas cannot be enabled on SLES10.
1371 Frequency : liblustre clients only
1373 Description: getdirentries does not give error when run on compute nodes
1374 Details : getdirentries does not fail when the size specified as an argument
1375 is too small to contain at least one entry
1377 Severity : enhancement
1379 Description: Add LNET router traceability for debug purposes
1380 Details : If a checksum failure occurs with a router as part of the
1381 IO path, the NID of the last router that forwarded the bulk data
1382 is printed so it can be identified.
1387 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1388 Details : Due to a race condition, liblustre clients were occasionally
1389 evicted incorrectly.
1391 Severity : enhancement
1393 Description: lfs setstripe use optional parameters instead of postional
1396 Severity : enhancement
1398 Description: Nanosecond timestamp support for ldiskfs
1399 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1400 resolution timestamps. There is not yet support for this at
1401 the Lustre filesystem level.
1404 Frequency : during server recovery
1406 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1407 Details : request with rq_no_resend flag not awake l_wait_event if they get a
1411 Frequency : nfs export on patchless client
1413 Description: connectathon hang when test nfs export over patchless client
1414 Details : Disconnected dentry cannot be found with lookup, so we do not need
1415 to unhash it or make it invalid
1418 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1419 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1420 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1425 Description: open req refcounting wrong on reconnect
1426 Details : If reconnect happened between getting open reply from server and
1427 call to mdc_set_replay_data in ll_file_open, we will schedule
1428 replay for unreferenced request that we are about to free.
1429 Subsequent close will crash in variety of ways.
1430 Check that request is still eligible for replay in
1431 mdc_set_replay_data().
1436 Description: disable writes to filesystem when reading health_check file
1437 Details : the default for reading the health_check proc file has changed
1438 to NOT do a journal transaction and write to disk, because this
1439 can cause reads of the /proc file to hang and block HA state
1440 checking on a healthy but otherwise heavily loaded system. It
1441 is possible to return to the previous behaviour during configure
1442 with --enable-health-write.
1444 Severity : enhancement
1446 Description: 64-bit inode version
1447 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1448 to the inode. This will be required for version-based recovery.
1453 Description: MDS fails to start if a duplicate client export is detected
1454 Details : in some rare cases it was possible for a client to connect to
1455 an MDS multiple times. Upon recovery the MDS would detect this
1456 and fail during startup. Handle this more gracefully.
1458 Severity : enhancement
1460 Description: Add -o localflock option to simulate old noflock behaviour.
1461 Details : This will achieve local-only flock/fcntl locks coherentness.
1466 Description: log_commit_thread vs filter_destroy race leads to crash
1467 Details : Take import reference before releasing llog record semaphore
1472 Description: Wrong request locking in request set processing
1473 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1474 imp_delayed_list, in this place should be used imp_lock.
1477 Frequency : when reconnecting
1479 Description: Grant leak when OSC reconnect to OST
1480 Details : When osc reconnect ost, OST(filter) should check whether it
1481 should grant more space to client by comparing fed_grant and
1482 cl_avail_grant, and return the granted space to client instead
1483 of "new granted" space, because client will call osc_init_grant
1484 to update the client grant space info.
1487 Frequency : when client reconnects to OST
1489 Description: Grant leak when OSC does a resend and replays bulk write
1490 Details : When osc reconnect to OST, OST(filter) should clear grant info of
1491 bulk write request, because the grant info while be sync between
1492 OSC and OST when reconnect, and we should ignore the grant info
1493 these of resend/replay write req.
1498 Description: Grant space more than avaiable space sometimes.
1499 Details : When then OST is about to be full, if two bulk writing from
1500 different clients came to OST. Accord the avaliable space of the
1501 OST, the first req should be permitted, and the second one
1502 should be denied by ENOSPC. But if the seconde arrived before
1503 the first one is commited. The OST might wrongly permit second
1504 writing, which will cause grant space > avaiable space.
1507 Frequency : when client is evicted
1509 Description: Grant might be wrongly erased when osc is evicted by OST
1510 Details : when the import is evicted by server, it will fork another
1511 thread ptlrpc_invalidate_import_thread to invalidate the
1512 import, where the grant will be set to 0. While the original
1513 thread will update the grant it got when connecting. So if
1514 the former happened latter, the grant will be wrongly errased
1515 because of this race.
1520 Description: Checking Stale with correct fid
1521 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1522 to check whether it is stale, because sometimes, we want the
1523 enqueue happened anyway, and op_data.fid2 will not be initialized.
1525 Severity : enhancement
1527 Description: update patchless client
1528 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1531 Frequency : only with 2.4 kernel
1533 Description: random memory corruption
1534 Details : size of struct ll_inode_info is to big for union inode.u and this
1535 can be cause of random memory corruption.
1540 Description: Memory leak in recovery
1541 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1542 It should also check obd_fail before fsfilt_start, otherwise if
1543 fsfilt_start return -EROFS,(failover mds during mds recovery).
1544 then the req will return with repmsg->transno = 0 and rc = EROFS.
1545 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1546 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1547 be freed no matter whether fsfilt_commit success or not.
1550 Frequency : only with huge numbers of clients
1552 Description: Prevents from taking the superblock lock in llap_from_page for
1554 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1555 for llap_from_page call in ll_removepage() prevents from taking
1556 the superblock lock for a soon died page.
1561 Description: Not check open intent error before release open handle
1562 Details : in some rare cases, the open intent error is not checked before
1563 release open handle, which may cause
1564 ASSERTION(open_req->rq_transno != 0), because it tries to release
1565 the failed open handle.
1570 Description: Set cat log bitmap only after create log success.
1571 Details : in some rare cases, the cat log bitmap is set too early. and it
1572 should be set only after create log success.
1576 Description: Accessing a block bevice can re-enable I/O when Lustre is
1577 tearing down a device.
1578 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1583 Description: service threads may hog cpus when there are a lot of requests
1584 Details : Insert cond_resched to give other threads a chance to use some CPU
1589 Description: the cat log was not initialized in recovery
1590 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1591 unlink log on mds will not be initialized until mds post
1592 recovery. And also in mds post recovery, the unlink log will
1593 initialization will be done asynchronausly, so there will be race
1594 between add unlink log and unlink log initialization.
1598 Description: brw_stats were being printed incorrectly
1599 Details : brw_stats were being printed as log2 but all of them were not
1600 recorded as log2. Also remove some code duplication arising from
1601 filter_tally_{read,write}.
1605 Frequency : rare, only in recovery.
1606 Description: ASSERTION(req->rq_type != LI_POISON) failed
1607 Details : imp_lock should be held while iterating over imp_sending_list for
1608 prevent destroy request after get timeout in ptlrpc_queue_wait.
1612 Description: replay-single.sh test 52 fails
1613 Details : A lock's skiplist need to be cleanup when it being unlinked
1614 from its resource list.
1619 Description: Short directio read returns full requested size rather than
1621 Details : Direct I/O operations should return actual amount of bytes
1622 transferred rather than requested size.
1624 Severity : enhancement
1626 Description: metadata RPC reduction (e.g. for rm performance)
1627 Details : decrease the amount of synchronous RPC between clients and servers
1628 by canceling conflicing lock before the operation on the client
1629 and packing thier handles into the main operation RPC to server.
1631 Severity : enhancement
1633 Description: add #ifdef HAVE_KERNEL_CONFIG_H
1634 Details : kernels from 2.6.19 not need include linux/config.h, but add
1635 include linux/autoconf.h in commpiler command line.
1637 Severity : enhancement
1639 Description: patchless client support for 2.6.22 kernel
1640 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
1641 removed. In this case we need drop using os depended interface to
1642 kmem_cache and use cfs_mem_cache API.
1647 Description: fix mal-formatted messages
1648 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
1652 Frequency : always in liblustre
1653 Description: wrong IS_ERR implementation in liblustre.h
1654 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
1659 Description: Correct condition for output debug message.
1660 Details : inode i_nlink equal zero is not enough for output message about
1661 disk corruption, i_ctime and i_mode should be also checked.
1665 Frequency : always in patchless client
1666 Description: add configure check for truncate_complete_page
1667 Details : improve checks for exported symbols. This allow run check without
1668 sources, but with Module.symvers shipped with kernel distribution.
1669 add check for truncate_complete_page used by patchless client.
1673 Description: sanity.sh test_77h fails with "test_77h file compare failed"
1674 Details : test_77h uses a file which was messed by other test case.
1678 Description: Not Check whether lov_tgts is NULL in some lov functions
1679 Details : Checking whether lov_tgts is NULL in some functions.
1683 Description: replace obdo_alloc() with OBDO_ALLOC macro
1684 Details : nothing special is done in obdo_alloc() function, and for
1685 debugging purpose, it needs to be replaced with macros.
1689 Description: bad return value and errno from fcntl call
1690 Details : In liblustre API, errno should be a negative value if error
1695 Description: ptlrpc_check_set() LBUG
1696 Details : In case of positive reply from server and failed client bulk
1697 callback after bulk transfer shouldn't LBUG, but process this
1698 request as erroneous.
1700 Severity : enhancement
1702 Description: VFS operations stats tool.
1703 Details : Tool which collects stats by tracking value written in pid,
1704 ppid, gid and uses llstat to generate output to plot graph using
1706 Updated lustre/utils/Makefile.am
1707 Added lustre/utils/ltrack_stats.c
1709 Severity : enhancement
1711 Description: 2.6.18 server support (lustre 1.6.1)
1712 Details : Support for 2.6.18 kernels on the server side.
1717 Description: ASSERTION(imp->imp_conn_current) failed
1718 Details : an assertion failure is hit if a client node boots and attempts to
1719 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
1722 Frequency : only for i686
1724 Description: 1.4.11 RC1 build fails for RHEL 4, i686
1725 Details : Fixed config variable for build.
1730 Description: Updated patchess for new RHEL4 kernel
1731 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
1732 Updated patch jbd-stats-2.6.9.patch
1733 Updated patch qsnet-rhel4-2.6.patch
1734 Updated patch quota-deadlock-on-pagelock-core.patch
1735 Updated patch vfs_intent-2.6-rhel4.patch
1736 Updated patch vfs_races-2.6-rhel4.patch
1737 Updated series file 2.6-rhel4-titech.series
1738 Updated series file 2.6-rhel4.series
1739 Updated kernel config files
1744 Description: lquota slave complains LBUG when reconnecting with mds
1746 Details : quota slave depends qctxt->lqc_import to send its quota request.
1747 This pointer will be invalid if mds did failover or broke its
1748 connect to osts, which leads to LBUG.
1751 Frequency : when qunit size is too small(less than 20M)
1753 Description: write is stopped by improper -EDQUOT
1754 Details : If the master is busy and qunit size is small enough(let's say 1M),
1755 the slave can not get quota from master on time, which will lead
1756 slave to trigger a -EQUOTA to client.
1761 Description: Deadlock during metadata tests
1762 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
1763 called with the per-dentry lock held.
1766 Frequency : SLES9 only
1768 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
1769 Details : set CONFIG_NR_CPUS to 128 instead of 8.
1771 Severity : enhancement
1773 Description: remove fs_prep_san_write operation and related patches
1774 Details : remove the ext3-san-jdike patches which are no longer useful.
1779 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
1780 Details : a disk corruption can cause the mballoc code to assert on a
1781 double free or other extent corruptions. Handle these with
1782 ext3_error() instead of assertions.
1785 Frequency : only with mballoc3 code and deep extent trees
1787 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
1788 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
1789 can cause the filesystem to be remounted read-only.
1794 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
1795 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
1796 few external users of this ioctl number is lfsck's call to
1797 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
1798 Add the old ioctl number to the handler so both old and new
1803 Description: parallel lock callbacks
1804 Details : Instead of sending blocking and completion callbacks as separated
1805 requests, adding them to a set and sending in parallel.
1809 Description: Disable most debugging by default
1810 Details : To improve performance, disable most logging (for debug purposes)
1811 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
1812 default, and HA includes fewer messages.
1816 Frequency : only run on patchless client.
1817 Description: use do_facet on sanity.sh for test handling recoverables errors
1818 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
1822 Description: hash tables for lists of nids, connections and uuids
1823 Details : Hash tables noticeably help when a lot of clients connect to a
1824 server, to faster identify duplicate connections or reconnects,
1825 also to faster find export to evict in manual eviction case.
1829 Description: Sometimes, when the server evict a client, and the client will
1830 not be evicted as soon as possible.
1831 Details : In enqueue req, the error was returned by intent, instead of
1832 rq_status which make ptlrpc layer not detect this error, and
1833 does not evict the client. So enqueue error should be returned
1837 Frequency : only at startup
1839 Description: Delay client connections to MDT until fist MDT->OST connect
1840 Details : If a client tried to create a new file before the MDT had
1841 connected to any OSTs, the create would return EIO. Now
1842 the client will simply block until the MDT connects to the
1843 first OST and the create can succeed.
1846 Frequency : only for SLES9
1848 Description: Routinely utilize latest Quadrics drivers in CFS releases
1849 Details : Update patch qsnet-suse-2.6.patch.
1852 Frequency : only for sles10
1854 Description: Update patches for SLES 10 SP1 kernel.
1855 Details : Update the patch vfs_intent-2.6-sles10.patch.
1860 Description: Routinely utilize latest Quadrics drivers in CFS releases
1861 Details : Update patch qsnet-rhel4-2.6.patch.
1864 Frequency : at statup only
1866 Description: mds_lov_synchronize race leads to various problems
1867 Details : simultaneous MDT->OST connections at startup can cause the
1868 sync to abort, leaving the OSC in a bad state.
1873 Description: Using wrong pointer in osc_brw_prep_request
1874 Details : Access to array[-1] can produce panic if kernel compiled with
1875 CONFIG_PAGE_ALLOC enabled
1877 Severity : enhancement
1879 Description: Async OSC create to avoid the blocking unnecessarily.
1880 Details : If a OST has no remain object, system will block on the creating
1881 when need to create a new object on this OST. Now, ways use
1882 pre-created objects when available, instead of blocking on an
1883 empty osc while others are not empty. If we must block, we block
1884 for the shortest possible period of time.
1888 Frequency : only in recovery
1889 Description: Mark OST as early accessible if his start SYNC.
1890 Details : osc_precreate return flag early accessible if oscc marked as
1891 OSCC_FLAG_SYNC_IN_PROGRESS.
1896 Description: Sometimes precreate code can triger create object on wrong ost
1897 Details : Wrong protected or not not restored variables aftre precreate loop
1898 can produce creation object on wrong ost.
1901 Frequency : oss recovery
1903 Description: llog_commit_thread cleanup should sync with llog_commit_thread
1905 Details : llog_commit_thread_count should be synced between llog_commit
1906 start and cleanup, so new llog_commit thread should not be started
1907 when llog_commit threads being stopped to avoid accessing some
1910 Severity : enhancement
1912 Description: Add printing inode info into message about error in writepage.
1914 --------------------------------------------------------------------------------
1916 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1921 Frequency : on some architectures
1923 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1924 Details : Uninitialized flags sometimes cause configuration commands to
1928 Frequency : patchless clients only
1930 Description: missing __iget() symbol export
1931 Details : The __iget() symbol export is missing. To avoid the need for
1932 this on patchless clients the deathrow inode reaper is turned
1933 off, and we depend on the VM to clean up old inodes. This
1934 dependency was during via the fix for bug 12181.
1939 Description: sanity.sh fail: test_52b
1940 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
1941 incorrect inode's flags to client.
1943 --------------------------------------------------------------------------------
1945 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1947 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1948 INTEROPERATE with older versions automatically. In many cases a
1949 special upgrade step is needed. Please read the
1950 user documentation before upgrading any part of a 1.4.x system.
1951 * WARNING: Lustre configuration and startup changes are required with
1952 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1954 * Support for kernels:
1955 2.4.21-47.0.1.EL (RHEL 3),
1956 2.6.5-7.283 (SLES 9),
1957 2.6.9-42.0.10.EL (RHEL 4),
1958 2.6.12.6 vanilla (kernel.org),
1959 2.6.16.27-0.9 (SLES10)
1960 * Client support for unpatched kernels:
1961 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
1962 2.6.16 - 2.6.19 vanilla (kernel.org),
1963 2.6.9-42.0.8EL (RHEL 4)
1964 * Recommended e2fsprogs version: 1.39.cfs6
1965 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1968 Severity : enhancement
1970 Description: Async OSC create to avoid the blocking unnecessarily.
1971 Details : If a OST has no remain object, system will block on the creating
1972 when need to create a new object on this OST. Now, ways use
1973 pre-created objects when available, instead of blocking on an
1974 empty osc while others are not empty. If we must block, we block
1975 for the shortest possible period of time.
1977 Severity : enhancement
1979 Description: MountConf
1980 Details : Lustre configuration is now managed via mkfs and mount
1981 commands instead of lmc and lconf. New obd types (MGS, MGC)
1982 are added for dynamic configuration management. See
1983 https://mail.clusterfs.com/wikis/lustre/MountConf for
1986 Severity : enhancement
1988 Description: dynamic OST addition
1989 Details : OSTs can now be added to a live filesystem
1991 Severity : enhancement
1993 Description: startup order invariance
1994 Details : MDTs and OSTs can be started in any order. Clients only
1995 require the MDT to complete startup.
1997 Severity : enhancement
1999 Description: parallel, asynchronous orphan cleanup
2000 Details : orphan cleanup is now performed in separate threads for each
2001 OST, allowing parallel non-blocking operation.
2003 Severity : enhancement
2005 Description: optimized stripe assignment
2006 Details : stripe assignments are now made based on ost space available,
2007 ost previous usage, and OSS previous usage, in order to try
2008 to optimize storage space and networking resources.
2010 Severity : enhancement
2012 Description: Permanently set tunables
2013 Details : All writable /proc/fs/lustre tunables can now be permanently
2014 set on a per-server basis, at mkfs time or on a live
2017 Severity : enhancement
2019 Description: Lustre message v2
2020 Details : Add lustre message format v2.
2022 Severity : enhancement
2024 Description: client OST exclusion list
2025 Details : Clients can be started with a list of OSTs that should be
2026 declared "inactive" for known non-responsive OSTs.
2030 Description: ENOENT returned for valid filehandle during dbench.
2031 Details : Check if a directory has children when invalidating dentries
2032 associated with an inode during lock cancellation. This fixes
2033 an incorrect ENOENT sometimes seen for valid filehandles during
2034 testing with dbench.
2037 Frequency : SFS test only (otherwise harmless)
2039 Description: SPEC SFS validation failure on NFS v2 over lustre.
2040 Details : Changes the blocksize for regular files to be 2x RPC size,
2041 and not depend on stripe size.
2043 Severity : enhancement
2045 Description: fine-grained SMP locking inside DLM
2046 Details : Improve DLM performance on SMP systems by removing the single
2047 per-namespace lock and replace it with per-resource locks.
2049 Severity : enhancement
2051 Description: don't hold multiple extent locks at one time
2052 Details : To avoid client eviction during large writes, locks are not
2053 held on multiple stripes at one time or for very large writes.
2054 Otherwise, clients can block waiting for a lock on a failed OST
2055 while holding locks on other OSTs and be evicted.
2057 Severity : enhancement
2059 Description: Multiple MD RPCs in flight.
2060 Details : Further unserialise some read-only MDT RPCs - learn about intents.
2061 To avoid overly-overloading MDT, introduce a limit on number of
2062 MDT RPCs in flight for a single client and add /proc controls
2063 to adjust this limit.
2065 Severity : enhancement
2067 Description: client read/write statistics
2068 Details : Add client read/write call usage stats for performance
2069 analysis of user processes.
2070 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2071 file access. extents_stats shows chunk size distribution.
2072 extents_stats_per_process show chunk size distribution per
2075 Severity : enhancement
2077 Description: per-client statistics on server
2078 Details : Add ldlm and operations statistics for each client in
2079 /proc/fs/lustre/mds|obdfilter/*/exports/
2081 Severity : enhancement
2083 Description: improved MDT statistics
2084 Details : Add detailed MDT operations statistics in
2085 /proc/fs/lustre/mds/*/stats
2087 Severity : enhancement
2089 Description: VFS operations stats
2090 Details : Add client VFS call stats, trackable by pid, ppid, or gid
2091 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
2096 Description: Fix client-side osc byte counters
2097 Details : The osc read/write byte counters in
2098 /proc/fs/lustre/osc/*/stats are now working
2101 Frequency : always as root on SLES
2103 Description: Failure of copying files with lustre special EAs.
2104 Details : Client side always return success for setxattr call for lustre
2105 special xattr (currently only "trusted.lov").
2110 Description: Refcount LNET uuids
2111 Details : The global LNET uuid list grew linearly with every startup;
2112 refcount repeated list entries instead of always adding to
2115 Severity : enhancement
2117 Description: Dynamic service threads
2118 Details : Within a small range, start extra service threads
2119 automatically when the request queue builds up.
2122 Frequency : mixed-endian client/server environments
2124 Description: mixed-endian crashes
2125 Details : The new msg_v2 system had some failures in mixed-endian
2128 Severity : enhancement
2130 Description: Easy OST removal
2131 Details : OSTs can be permanently deactivated with e.g. 'lctl
2132 conf_param lustre-OST0001.osc.active=0'
2134 Severity : enhancement
2136 Description: MGS proc entries
2137 Details : Added basic proc entries for the MGS showing what filesystems
2140 Severity : enhancement
2142 Description: provide MGS failover
2143 Details : Added config lock reacquisition after MGS server failover.
2145 Severity : enhancement
2147 Description: add Linux 2.4 support
2148 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2152 Description: a large application tries to do I/O to the same resource and dies
2153 in the middle of it.
2154 Details : Check the req->rq_arrival time after the call to
2155 ost_brw_lock_get(), but before we do anything about
2156 processing it & sending the BULK transfer request. This
2157 should help move old stale pending locks off the queue as
2158 quickly as obd_timeout.
2161 Frequency : when an incorrect nid is specified during startup
2163 Description: ptlrpc connect to non-existant node causes kernel crash
2164 Details : LNET can't be re-entered from an event callback, which
2165 happened when we expire a message after the export has been
2166 cleaned up. Instead, hand the zombie cleanup off to another
2169 Severity : enhancement
2171 Description: plain/inodebits lock performance improvement
2172 Details : Grouping plain/inodebits in granted list by their request modes
2173 and bits policy, thus improving the performance of search through
2177 Frequency : only if OST filesystem is corrupted
2179 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2180 Details : for a short time RPCs with bulk IO are in the replay list,
2181 but replay of bulk IOs is unimplemented. If the OST filesystem
2182 is corrupted due to disk cache incoherency and then replay is
2183 started it is possible to trip an assertion. Avoid putting
2184 committed RPCs into the replay list at all to avoid this issue.
2187 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2189 Description: System hang on startup
2190 Details : This bug allowed the liblustre (e.g. catamount) client to
2191 return to the app before handling all startup RPCs. This
2192 could leave the node unresponsive to lustre network traffic
2193 and manifested as a server ptllnd timeout.
2195 Severity : enhancement
2197 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2198 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2199 boolean to control whether to print peer debug info when a
2200 client's RPC times out.
2203 Frequency : only for kernels with patches from Lustre below 1.4.3
2205 Description: Remove old rdonly API
2206 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2209 Frequency : only for devices with external journals
2211 Description: Set external device read-only also
2212 Details : During a commanded failover stop, we set the disk device
2213 read-only while the server shuts down. We now also set any
2214 external journal device read-only at the same time.
2217 Frequency : when upgrading from 1.4 while trying to change parameters
2219 Description: The wrong (new) MDC name was used when setting parameters for
2220 upgraded MDT's. Also allows changing of OSC (and MDC)
2221 parameters if --writeconf is specified at tunefs upgrade time.
2224 Frequency : when setting specific ost indicies
2226 Description: QOS code breaks on skipped indicies
2227 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2228 created with --index need not be sequential.
2230 Severity : enhancement
2232 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2233 Details : The uninit_groups feature works in conjunction with the kernel
2234 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2235 up the pass1 processing of e2fsck. This is a read-only feature
2236 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2237 mount filesystems that have had this feature enabled.
2239 Severity : enhancement
2241 Description: Improve multi-block allocation algorithm to avoid fragmentation
2242 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2243 allocation locality and avoid filesystem fragmentation.
2248 Description: lfs quota fails with deactivated OSTS
2249 Details : With this patch, three improvements are included:
2250 1. detete the softlimit in mds and osts when use "lfs quota".
2251 2. display the inaccurate data in the output of "lfs quota".
2252 3. try to get quota info when "lfs quota" is executed.
2254 ------------------------------------------------------------------------------
2256 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2258 * Support for kernels:
2259 2.4.21-47.0.1.EL (RHEL 3)
2260 2.6.5-7.283 (SLES 9)
2261 2.6.9-42.0.10.EL (RHEL 4)
2262 2.6.12.6 vanilla (kernel.org)
2263 2.6.16.27-0.9 (SLES 10)
2264 * Recommended e2fsprogs version: 1.39.cfs5
2266 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2270 Frequency : occasional, depends on client load and configuration
2271 Bugzilla : 12181, 12203
2272 Description: data loss for recently-modified files
2274 Details : In some cases it is possible that recently written or created
2275 files may not be written to disk in a timely manner (this should
2276 normally be within 30s unless client IO load is very high).
2277 The problem appears as zero-length files or files that are a
2278 multiple of 1MB in size after a client crash or client eviction
2279 that are missing data at the end of the file.
2281 This problem is more likely to be hit on clients where files are
2282 repeatedly created and unlinked in the same directory, clients
2283 have a large amount of RAM, have many CPUs, the filesystem has
2284 many OSTs, the clients are rebooted frequently, and/or the files
2285 are not accessed by other nodes after being written.
2287 The presence of the problem can be detected by looking at
2288 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2289 smaller than the second (nr_unused) then dirty files will not
2290 be flushed automatically to disk. "sync; sleep 10" should be
2291 run several times on the node before unmounting it to update
2292 Lustre (this is also safe to run on nodes without this problem).
2294 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2295 that can cause this same problem, so customers using that kernel
2296 also need to update the kernel in addition to Lustre. In order
2297 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2299 It is normal that files written just before a client crash (less
2300 than 30s) may not yet have been flushed to disk, even for local
2304 Frequency : frequent on thin XT3 nodes
2306 Description: UUID collision on thin XT3 Linux nodes
2307 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2308 randomly, since we relied on an insufficiently-seeded PRNG.
2313 Description: OSS hangs after "All ost request buffers busy"
2314 Details : A deadlock between quota and journal operations caused OSS
2315 hangs after printing "All ost request buffers busy."
2318 Frequency : always on liblustre builds
2320 Description: Cleanup compiler warnings on liblustre
2323 Frequency : always on liblustre builds on XT3
2325 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2329 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2330 > 32000 subdirectories
2331 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2332 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2333 subdirectory count crosses 32000. This will aid e2fsck to
2334 correctly handle more than 32000 subdirectories.
2337 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2339 Description: System hang on startup
2340 Details : This bug allowed the liblustre (e.g. catamount) client to
2341 return to the app before handling all startup RPCs. This
2342 could leave the node unresponsive to lustre network traffic
2343 and manifested as a server ptllnd timeout.
2345 Severity : enhancement
2347 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2348 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2349 boolean to control whether to print peer debug info when a
2350 client's RPC times out.
2355 Description: make O_SYNC working on 2.6 kernels
2356 Details : 2.6 kernels use different method for mark pages for write,
2357 so need add a code to lustre for O_SYNC work.
2362 Description: Failure to close file and release space on NFS
2363 Details : Put inode details into lock acquired in ll_intent_file_open.
2364 Use mdc_intent_lock in ll_intent_open to properly
2365 detect all kind of errors unhandled by mdc_enqueue.
2370 Description: proc file read during shutdown sometimes raced obd removal,
2372 Details : Add lock to prevent obd access after proc file removal.
2375 Frequency : Only for files larger than 4GB on 32-bit clients.
2377 Description: improperly doing page alignment of locks
2378 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2379 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2384 Description: RPCs being resent when they shouldn't be
2385 Details : Some RPCs that should not be resent are being resent. This
2386 can cause inconsistencies in the RPC state machine. Do not
2387 resend such requests.
2390 Frequency : rare, only with NFS export
2392 Description: Crash on NFS re-export node
2393 Details : under very unusual load conditions an assertion is hit in
2394 ll_intent_file_open()
2397 Frequency : only if OST filesystem is corrupted
2399 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2400 Details : for a short time RPCs with bulk IO are in the replay list,
2401 but replay of bulk IOs is unimplemented. If the OST filesystem
2402 is corrupted due to disk cache incoherency and then replay is
2403 started it is possible to trip an assertion. Avoid putting
2404 committed RPCs into the replay list at all to avoid this issue.
2409 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2410 Details : Large single O_DIRECT read and write calls can fail to allocate
2411 a sufficiently large buffer to process the request. In case of
2412 allocation failure the allocation is retried with a smaller
2413 buffer and broken into smaller requests.
2415 Severity : enhancement
2417 Description: Add -o localflock option to simulate old noflock behaviour.
2418 Details : This will achieve local-only flock/fcntl locks coherentness.
2423 Description: versioning check is incomplete
2424 Details : Checking the version difference of client vs. server, report
2425 error if the gap is too big.
2430 Description: add support PG_writeback bit
2431 Details : add support for PG_writeback bit for Lustre, for more carefull
2432 work with page cache in 2.6 kernel. This also fix some deadlocks
2433 and remove hack for work O_SYNC with 2.6 kernel.
2435 Severity : enhancement
2437 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2438 Details : The uninit_groups feature works in conjunction with the kernel
2439 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2440 up the pass1 processing of e2fsck. This is a read-only feature
2441 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2442 mount filesystems that have had this feature enabled.
2444 Severity : enhancement
2446 Description: Improve multi-block allocation algorithm to avoid fragmentation
2447 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2448 allocation locality and avoid filesystem fragmentation.
2450 ------------------------------------------------------------------------------
2452 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2454 * Support for kernels:
2455 2.6.9-42.0.3.EL (RHEL 4)
2456 2.6.5-7.276 (SLES 9)
2457 2.4.21-47.0.1.EL (RHEL 3)
2458 2.6.12.6 vanilla (kernel.org)
2459 2.6.16.21-0.8 (SLES10)
2460 * Recommended e2fsprogs version: 1.39.cfs2-0
2462 * The backwards-compatible /proc/sys/portals symlink has been removed
2463 in this release. Before upgrading, please ensure that you change
2464 any configuration scripts or /etc/sysctl.conf files that access
2465 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2466 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2467 in advance of the upgrade on any system running Lustre 1.4.6 or
2468 newer, since /proc/sys/lnet was added in that version.
2469 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2473 Frequency : only when quota is used
2475 Description: avoid scanning export list for quota master
2476 Details : Change the algorithms to avoid scanning export list in order
2477 to improve the efficiency.
2480 Frequency : MDS failover only, very rarely
2482 Description: "went back in time" messages on mds failover
2483 Details : The greatest transno may be lost when the current operation
2484 finishes with an error (transno==0) and the client's last_rcvd
2485 record is over-written. Save the greatest transno in the
2486 mds_last_transno for this case.
2489 Frequency : always for specific kernels and striping counts
2491 Description: client may get "Matching packet too big" without ACL support
2492 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2493 when trying to access files in certain configurations. The
2494 clients should in fact be denied when mounting because they do
2495 not understand ACLs.
2498 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2500 Description: many clients connecting with IO in progress causes connect timeouts
2501 Details : Avoid synchronous journal commits to avoid delays caused by many
2502 clients connecting/disconnecting when bulk IO is in progress.
2503 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2504 OST_IO_PORTAL to avoid delays behind potentially many pending
2508 Frequency : occasionally with multiple writers to a single file
2510 Description: shared writes to file may result in wrong size reported by stat()
2511 Details : Allow growing of kms when extent lock is cancelled
2514 Frequency : always with random mmap IO to multi-striped file
2516 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2517 Details : Only if the hole is at the end of OST object so that kms is too
2518 small. Fix is to increase kms accordingly in ll_nopage.
2521 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2523 Description: writes to a missing object would leak memory on the OST
2524 Details : If there is an inconsistency between the MDS and OST filesystems,
2525 such that the MDS references an object that doesn't exist, writes
2526 to that object will leak memory due to incorrect cleanup in the
2527 error handling path, eventually running out of memory on the OST.
2532 Description: Creating too long symlink causes lustre errors
2533 Details : Check symlink and name lengths before sending requests to MDS.
2536 Frequency : only if flock is enabled (not on by default)
2538 Description: posix locks not released on fd closure on 2.6.9+
2539 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2540 kernels, this caused such locks not to be released on fd close and
2541 then assertions on fs unmount about still used locks.
2544 Frequency : MDS failover only, very rarely
2546 Description: clients may get ASSERTION(granted_lock != NULL)
2547 Details : When request was taking a long time, and a client was resending
2548 a getattr by name lock request. The were multiple lock requests
2549 with the same client lock handle and
2550 mds_getattr_name->fixup_handle_for_resent_request found one of the
2551 lock handles but later failed with ASSERTION(granted_lock != NULL).
2556 Description: handle->h_buffer_credits > 0, assertion failure
2557 Details : h_buffer_credits is zero after truncate, causing assertion
2558 failure. This patch extends the transaction or creates a new
2562 Frequency : NFS re-export or patchless client
2563 Bugzilla : 11179, 10796
2564 Description: Crash on NFS re-export node (__d_move)
2565 Details : We do not want to hash the dentry if we don't have a lock.
2566 But if this dentry is later used in d_move, we'd hit uninitialised
2567 list head d_hash, so we just do this to init d_hash field but
2568 leave dentry unhashed.
2571 Frequency : NFS re-export or patchless client
2573 Description: NFS exports has problem with symbolic link
2574 Details : lustre client didn't properly install dentry when re-exported
2575 to NFS or running patchless client.
2578 Frequency : NFS re-export or patchless client
2580 Description: Various nfs/patchless fixes.
2581 Details : fixes reuse disconected alias for lookup process - this fixes
2582 warning "find_exported_dentry: npd != pd",
2583 fix permission error with open files at nfs.
2584 fix apply umask when do revalidate.
2587 Frequency : occasional
2589 Description: Crash on NFS re-export node
2590 Details : calling clear_page() on the wrong pointer triggered oops in
2591 generic_mapping_read().
2594 Frequency : rarely, using O_DIRECT IO
2596 Description: unaligned directio crashes client with LASSERT
2597 Details : check for unaligned buffers before trying any requests.
2600 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2602 Description: stale data returned from RAID cache
2603 Details : If only a small amount of IO is done to the RAID device before
2604 reading it again it is possible to get stale data from the RAID
2605 cache instead of reading it from disk.
2608 Frequency : always for sles10 kernel
2610 Description: sles10 support
2611 Details : ll_follow_link: compile fixes and using of nd_set_link
2612 under newer kernels.
2615 Frequency : depends on arch, kernel and compiler version, always on sles10
2618 Description: recursive or deep enough symlinks cause stack overflow
2619 Details : getting rid of large stack-allocated variable in
2623 Frequency : depends on hardware
2625 Description: lustre write performance loss in the SLES10 kernel
2626 Details : the performance loss is caused by using of write barriers in the
2627 ext3 code. The SLES10 kernel turns barrier support on by
2628 default. The fix is to undo that change for ldiskfs.
2630 ------------------------------------------------------------------------------
2632 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2634 * Support for kernels:
2635 2.6.9-42.0.3EL (RHEL 4)
2636 2.6.5-7.276 (SLES 9)
2637 2.4.21-47.0.1.EL (RHEL 3)
2638 2.6.12.6 vanilla (kernel.org)
2642 Frequency : quota enabled and large files being deleted
2644 Description: releasing more than 4GB of quota at once hangs OST
2645 Details : If a user deletes more than 4GB of files on a single OST it
2646 will cause the OST to spin in an infinite loop. Release
2647 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2652 Description: statfs data retrieved from /proc may be stale or zero
2653 Details : When reading per-device statfs data from /proc, in the
2654 {kbytes,files}_{total,free,avail} files, it may appear
2655 as zero or be out of date.
2658 Frequency : systems with MD RAID1 external journal devices
2660 Description: lconf's call to blkid is confused by RAID1 journal devices
2661 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2662 of returning all block devices that match the journal UUID.
2665 Frequency : always, for aggregate stripe size over 4GB
2667 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2668 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2669 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2672 Frequency : always if "lfs find" used on a local file/directory
2674 Description: "lfs find" segfaults if used on a local file/directory
2675 Details : The case where a directory component was not specified wasn't
2676 handled correctly. Handle this properly.
2679 Frequency : always on ppc64
2681 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2682 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2683 perform endianness conversion.
2686 Frequency : rarely (truncate to non-zero file size after write under load)
2687 Bugzilla : 10730, 10687
2688 Description: Files padded with zeros to next 4K multiple
2689 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2690 it is possible that files that are truncated to a non-zero size
2691 immediately after being written are filled with zero bytes beyond
2692 the truncated size. No file data is lost.
2694 Severity : enhancement
2696 Description: Allow recovery/failover for liblustre clients.
2697 Details : liblustre clients were unaware of failover configurations until
2700 Severity : enhancement
2702 Description: user file locks should fail when not mounting with flock option
2703 Details : Set up an error-returning stub in ll_file_operations.lock field
2704 to prevent incorrect behaviour when client is mounted without
2705 flock option. Also, set up properly f_op->flock field for
2709 Frequency : always on ia64
2711 Description: "lfs df" loops on printing out MDS statfs information
2712 Details : The obd_ioctl_data was not initialized and in some systems
2713 this caused a failure during the ioctl that did not return
2714 an error. Initialize the struct and return an error on failure.
2717 Frequency : SLES 9 only
2719 Description: Error of copying files with lustre special EAs as root
2720 Details : Client side always return success for setxattr call for lustre
2721 special xattr (currently only "trusted.lov").
2724 Frequency : rarely on clusters with both ia64+i386 clients
2726 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2727 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2728 clients are doing concurrent writes to the same file it is
2729 possible that the ia64 clients may LASSERT because the OST
2730 extent locks are not PAGE_SIZE aligned. Ensure that grown
2731 locks are always aligned on the request boundary.
2734 Frequency : specific use, occasional
2736 Description: Overwriting in use executable truncates on-disk binary image
2737 Details : If one node attempts to overwrite an executable in use by
2738 another node, we now correctly return ETXTBSY instead of
2739 truncating the file.
2741 Severity : enhancement
2743 Description: Async OSC create to avoid the blocking unnecessarily.
2744 Details : If a OST has no remain object, system will block on the creating
2745 when need to create a new object on this OST. Now, ways use
2746 pre-created objects when available, instead of blocking on an
2747 empty osc while others are not empty. If we must block, we block
2748 for the shortest possible period of time.
2753 Description: chmod on Lustre root is propagated to other clients
2754 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2755 invalid by the follow_mount time.
2758 Frequency : liblustre clients only
2760 Description: Race in 'instant cancel' lock handling could lead to such locks
2761 never to be granted in case of SMP MDS
2762 Details : Do not destroy not yet granted but cbpending locks in
2766 Frequency : replay/resend of open
2768 Description: non null lock assetion failure in mds_intent_policy
2769 Details : Trying to replay/resend lockless open requests resulted in
2770 mds_open() returning 0 with no lock. Now it sets a flag if
2771 a lock is going to be returned.
2773 Severity : enhancement
2775 Description: Checksum enhancements
2776 Details : New checksum enhancements allow for resending RPCs that failed
2779 Severity : enhancement
2781 Description: Tunables on number of dirty pages in cacche
2782 Details : Allow to set limit on number of dirty pages cached.
2787 Description: client crash on unmount - lock still has references
2788 Details : In some error handling cases it was possible to leak a lock
2789 reference on a client while accessing a file. This was not
2790 harmful to the client during operation, but would cause the
2791 client to crash when the filesystem is unmounted.
2794 Frequency : specific case, rare
2796 Description: ETXTBSY on mds though file not in use by client
2797 Details : ETXTBSY is no longer incorrectly returned when attempting to
2798 chmod or chown a directory that the user previously tried to
2799 execute or a currently-executing binary.
2802 Frequency : extremely rare except on liblustre-based clients
2804 Description: Lustre space not freed when files are deleted
2805 Details : Clean up open-unlinked files after client eviction. Previously
2806 the unlink was skipped and the files remained as orphans.
2811 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2812 Details : In some cases it was possible to send a blocking callback to
2813 a client doing a glimpse, even though that client didn't get
2814 a lock granted. When the glimpse lock is cancelled on the OST
2815 the freed lock is left on the waiting list and corrupted the list.
2818 Frequency : all core dumps
2820 Description: Broke core dumps to lustre
2821 Details : Negative dentry may be unhashed if parent does not have UPDATE
2822 lock, but some callers, e.g. do_coredump, expect dentry to be
2823 hashed after successful create, hash it in ll_create_it.
2825 ------------------------------------------------------------------------------
2827 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2829 * Support for kernels:
2830 2.6.9-42.0.2.EL (RHEL 4)
2831 2.6.5-7.276 (SLES 9)
2832 2.4.21-40.EL (RHEL 3)
2833 2.6.12.6 vanilla (kernel.org)
2837 Frequency : always on RHEL 3
2839 Description: Number of open files grows over time
2840 Details : The number of open files grows over time, whether or not
2841 Lustre is started. This was due to a filp leak introduced
2842 by one of our kernel patches.
2844 ------------------------------------------------------------------------------
2846 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
2848 * Support for kernels:
2849 2.6.9-42.EL (RHEL 4)
2850 2.6.5-7.267 (SLES 9)
2851 2.4.21-40.EL (RHEL 3)
2852 2.6.12.6 vanilla (kernel.org)
2857 Bugzilla : 5719, 9635, 9792, 9684
2858 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2859 Details : If a server is under heavy load and cannot reply to new
2860 connection requests before the client resends the (re)connect,
2861 the connection handling code can behave badly if two service
2862 threads are concurrently handing separate (re)connections from
2863 the same client. Add better locking to the connection handling
2864 code, and ensure that only a single connection will be processed
2865 for a given client UUID, even if the lock is dropped.
2867 Severity : enhancement
2869 Description: add TCP zero-copy support to kernel
2870 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2871 sends if the hardware supports scatter-gather and checksumming.
2872 This allows socklnd to do client-write and server-read more
2873 efficiently and reduce CPU utilization from skbuf copying.
2876 Frequency : only if NFS exporting from client
2878 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2879 Details : If setting attributes on a file created under NFS that had
2880 never been opened it would be possible to oops the client
2881 if the file had no objects.
2885 Bugzilla : 9326, 10402, 10897
2886 Description: client crash in ptlrpcd_wake() thread when sending async RPC
2887 Details : It is possible that ptlrpcd_wake() dereferences a freed async
2888 RPC. In rare cases the ptlrpcd thread alread processed the RPC
2889 before ptlrpcd_wake() was called and the request was freed.
2892 Frequency : always for liblustre
2894 Description: liblustre client does MDS+OSTs setattr RPC for each write
2895 Details : When doing a write from a liblustre client, the client
2896 incorrectly issued an RPC to the MDS and each OST the file was
2897 striped over in order to update the timestamps. When writing
2898 with small chunks and many clients this could overwhelm the MDS
2899 with RPCs. In all cases it would slow down the write because
2900 these RPCs are unnecessary.
2902 Severity : enhancement
2904 Description: allow number of MDS service threads to be changed at module load
2905 Details : It is now possible to change the number of MDS service threads
2906 running. Adding "options mds mds_num_threads={N}" to the MDS's
2907 /etc/modprobe.conf will set the number of threads for the next
2908 time Lustre is restarted (assuming the "mds" module is also
2909 reloaded at that time). The default number of threads will
2910 stay the same, 32 for most systems.
2915 Description: OST crash if filesystem is unformatted or corrupt
2916 Details : If an OST is started on a device that has never been formatted
2917 or if the filesystem is corrupt and cannot even mount then the
2918 error handling cleanup routines would dereference a NULL pointer.
2923 Description: NULL pointer deref in llap_from_page.
2924 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2925 incorrect) mapping if the page was truncated/reclaimed while it was
2926 searched for. Check for this condition and skip such pages when
2927 doing readahead. Introduce extra check to llap_from_page() to
2928 verify page->mapping->host is non-NULL (so page is not anonymous).
2931 Frequency : Sometimes when using sys_sendfile
2933 Description: "page not covered by a lock" warnings from ll_readpage
2934 Details : sendfile called ll_readpage without right page locks present.
2935 Now we introduced ll_file_sendfile that does necessary locking
2936 around call to generic_file_sendfile() much like we do in
2940 Frequency : with certain MDS communication failures at client mount time
2942 Description: NULL pointer deref after failed client mount
2943 Details : a client connection request may delayed by the network layer
2944 and not be sent until after the PTLRPC layer has timed out the
2945 request. If the client fails the mount immediately it will try
2946 to clean up before the network times out the request. Add a
2947 reference from the request import to the obd device and delay
2948 the cleanup until the network drops the request.
2951 Frequency : occasionally during client (re)connect
2953 Description: assertion failure during client (re)connect
2954 Details : processing a client connection request may be delayed by the
2955 client or server longer than the client connect timeout. This
2956 causes the client to resend the connection request. If the
2957 original connection request is replied in this interval, the
2958 client may trip an assertion failure in ptlrpc_connect_interpret()
2959 which thought it would be the only running connect process.
2962 Frequency : only with obd_echo servers and clients that are rebooted
2964 Description: kernel BUG accessing uninitialized data structure
2965 Details : When running an obd_echo server it did not start the ping_evictor
2966 thread, and when a client was evicted an uninitialized data
2967 structure was accessed. Start the ping_evictor in the RPC
2968 service startup instead of the OBD startup.
2970 Severity : enhancement
2971 Bugzilla : 10193 (patchless)
2972 Description: Remove dependency on various unexported kernel interfaces.
2973 Details : No longer need reparent_to_init, exit_mm, exit_files,
2974 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2977 Frequency : rare (only users of deprecated and unsupported LDAP config)
2979 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2980 Details : LDAP apparently contains 'lustreName' attributes instead of
2981 'name'. A simple remapping of the name is sufficient.
2984 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2986 Description: waiting_locks_callback trips kernel BUG if client is evicted
2987 Details : Running with the dump_on_timeout debug flag turned on makes
2988 it possible that the waiting_locks_callback() can try to dump
2989 the Lustre kernel debug logs from an interrupt handler. Defer
2990 this log dumping to the expired_lock_main() thread.
2992 Severity : enhancement
2994 Description: Support NFS exporting on 2.6 kernels.
2995 Details : Implement non-rawops metadata methods for NFS server to use without
2996 changing NFS server code.
2999 Frequency : very rare (synthetic metadata workload only)
3001 Description: two racing renames might cause an MDS thread to deadlock
3002 Details : Running the "racer" program may cause one MDS thread to rename
3003 a file from being the source of a rename to being the target of
3004 a rename at exactly the same time that another thread is doing
3005 so, and the second thread has already enqueued these locks after
3006 doing a lookup of the target and is trying to relock them in
3007 order. Ensure that we don't try to re-lock the same resource.
3010 Frequency : only very large systems with liblustre clients
3012 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3013 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3014 all OSTs in parallel. This allows the eviction of stale liblustre
3015 clients to proceed much faster than if they were done in series,
3016 and also offers similar improvements for other set_info RPCs.
3021 Description: excessive CPU usage during initial read phase on client
3022 Details : During the initial read phase on a client, it would agressively
3023 retry readahead on the file, consuming too much CPU and impacting
3024 performance (since 1.4.5.8). Improve the readahead algorithm
3025 to avoid this, and also improve some other common cases (read
3026 of small files in particular, where "small" is files smaller than
3027 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3032 Description: MDS crash when receiving packet with unknown intent.
3033 Details : Do not LBUG in unknown intent case, just return -EFAULT
3035 Severity : enhancement
3036 Bugzilla : 9293, 9385
3037 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3038 Details : Do not serialize getattr (non-intent version) and statfs.
3041 Frequency : occasional, when OST network is overloaded/intermittent
3043 Description: client evicted by OST after bulk IO timeout
3044 Details : If a client sends a bulk IO request (read or write) the OST
3045 may evict the client if it is unresposive to its data GET/PUT
3046 request. This is incorrect if the network is overloaded (takes
3047 too long to transfer the RPC data) or dropped the OST GET/PUT
3048 request. There is no need to evict the client at all, since
3049 the pinger and/or lock callbacks will handle this, and the
3050 client can restart the bulk request.
3053 Frequency : Always when mmapping file with no objects
3055 Description: client crashes when mmapping file with no objects
3056 Details : Check that we actually have objects in a file before doing any
3057 operations on objects in ll_vm_open, ll_vm_close and
3063 Description: Request leak when working with deleted CWD
3064 Details : Introduce advanced request refcount tracking for requests
3065 referenced from lustre intent.
3067 Severity : Enhancement
3069 Description: Cache open file handles on client.
3070 Details : MDS now will return special lock along with openhandle, if
3071 requested and client is allowed to hold openhandle, even if unused,
3072 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3073 closing files for every read/write openration.
3075 Severity : Enhancement
3077 Description: Cache open negative dentries on client when possible.
3078 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3079 negative dentries on lock revocation.
3084 Description: Remounting a client read-only wasn't possible with a zconf mount
3085 Details : It wasn't possible to remount a client read-only with llmount.
3087 Severity : enhancement
3088 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3089 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3090 MPICH is included to add Lustre-specific ADIO interfaces.
3091 This is based closely on the UFS ADIO layer and only differs
3092 in file creation, in order to allow the OST striping to be set.
3093 This is user-contributed code and not supported by CFS.
3098 Description: extended inode attributes (immutable, append-only) work improperly
3099 when 2.4 and 2.6 kernels are used on client/server or vice versa
3100 Details : Introduce kernel-independent values for these flags.
3102 Severity : enhancement
3105 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3106 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3107 granularity. Now, Fractional MB granularity is be supported,
3108 this is very useful for low memory system.
3110 Severity : enhancement
3112 Description: Getattr by fid
3113 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3114 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3115 have locks and avoids getattr rpc requests for it on every stat.
3118 Frequency : Always, for filesystems larger than 2TB
3120 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3121 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3122 filesystems larger than 2TB to be mounted with mballoc enabled.
3125 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3127 Description: filesystem corruption for non-standard kernels and very large OSTs
3128 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3129 filesystems larger than 2TB is mounted then the kernel will
3130 silently corrupt the start of the filesystem. CONFIG_LBD is
3131 enabled for all CFS-supported kernels, but the possibility of
3132 this happening with a modified kernel config exists.
3134 Severity : enhancement
3136 Description: add client O_DIRECT support for 2.6 kernels
3137 Details : It is now possible to do O_DIRECT reads and writes to files
3138 in the Lustre client mountpoint on 2.6 kernel clients.
3140 Severity : enhancement
3142 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3143 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3144 parallel, not waiting for response from every OST before sending
3145 a rpc to the next OST.
3150 Description: setattr vs write race when updating file timestamps
3151 Details : Client processes that update a file timestamp into the past
3152 right after writing to the file (e.g. tar) it is possible that
3153 the updated file modification time can be reset to the current
3154 time due to a race between processing the setattr and write RPC.
3156 Severity : enhancement
3158 Description: Bring 'lfs find' closer in line with regular Linux find.
3159 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3160 -print0 options and obtains all the needed info through the lustre
3163 Severity : enhancement
3165 Description: support up to 1024 configured devices on one node
3166 Details : change obd_dev array from statically allocated to dynamically
3167 allocated structs as they are first used to reduce memory usage
3172 Description: Flush dirty partially truncated pages during truncate
3173 Details : Immediatelly flush partially truncated pages in filter_setattr,
3174 this way we completely avoid having any pages in page cache on OST
3175 and can retire ugly workarounds during writes to flush such pages.
3180 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3181 Details : protect inode from truncation within vfs_unlink() context
3182 just take a reference before calling vfs_unlink() and release it
3183 when parent's i_sem is free.
3186 Frequency : always, if extents are used on OSTs
3188 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3189 Details : OSTs using the extents format would not zero the high 16 bits of
3190 the index physical block number. This is not a problem for any
3191 OST filesystems smaller than 16TB, and no kernels support ext3
3192 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3193 new/modified files) and can be fixed for existing filesystems
3194 with e2fsprogs-1.39-cfs1.
3199 Description: import connection selection may be incorrect if timer wraps
3200 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3201 import connections to be ignored if the 32-bit jiffies counter
3202 wraps. Use a 64-bit jiffies counter.
3205 Frequency : during server recovery
3207 Description: crash after server is denying duplicate export
3208 Details : If clients are resending connect requests to the server, the
3209 server refuses to allow a client to connect multiple times.
3210 Fixed a bug in the handling of this case.
3213 Frequency : very large clusters immediately after boot
3215 Description: LNET request buffers exhausted under heavy short-term load
3216 Details : If a large number of client requests are generated on a service
3217 that has previously never seen so many requests it is possible
3218 that the request buffer growth cannot keep up with the spike in
3219 demand. Instead of dropping incoming requests, they are held in
3220 the LND until the RPC service can accept more requests.
3223 Frequency : Sometimes during replay
3225 Description: Assertion failure in ll_local_open after replay.
3226 Details : If replay happened on an open request reply before we were able
3227 to set replay handler, reply will become not swabbed tripping the
3228 assertion in ll_local_open. Now we set the handler right after
3229 recognising of open request
3232 Frequency : very rare
3234 Description: kernel reports "badness in vsnprintf"
3235 Details : Reading from the "recovery_status" /proc file in small chunks
3236 may cause a negative length in lprocfs_obd_rd_recovery_status()
3237 call to vsnprintf() (which is otherwise harmless). Exit early
3238 if there is no more space in the output buffer.
3240 Severity : enhancement
3242 Description: clear OBD RPC statistics by writing to them
3243 Details : It is now possible to clear the OBD RPC statistics by writing
3244 to the "stats" file.
3249 Description: Client mtime is not the same on different clients after utimes
3250 Details : In some cases, the client was using the utimes() syscall on
3251 a file cached on another node. The clients now validate the
3252 ctime from the MDS + OSTs to determine which one is right.
3257 Description: Inability to activate failout mode
3258 Details : lconf script incorrectly assumed that in python string's numeric
3259 value is used in comparisons.
3262 Frequency : always with multiple stripes per file
3264 Description: Inefficient object allocation for mutli-stripe files
3265 Details : When selecting which OSTs to stripe files over, for files with
3266 a stripe count that divides evenly into the number of OSTs,
3267 the MDS is always picking the same starting OST for each file.
3268 Return the OST selection heuristic to the original design.
3273 Description: mount failures may take full timeout to return an error
3274 Details : Under some heavy load conditions it is possible that a
3275 failed mount can wait for the full obd_timeout interval,
3276 possibly several minutes, before reporting an error.
3277 Instead return an error as soon as the status is known.
3279 ------------------------------------------------------------------------------
3281 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
3283 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3284 INTEROPERATE with older versions automatically. Please read the
3285 user documentation before upgrading any part of a live system.
3286 * WARNING: Lustre networking configuration changes are required with
3287 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3290 * Support for kernels:
3291 2.6.9-22.0.2.EL (RHEL 4)
3292 2.6.5-7.244 (SLES 9)
3293 2.6.12.6 vanilla (kernel.org)
3296 Severity : enhancement
3297 Bugzilla : 7981/8208
3298 Description: Introduced Lustre Networking (LNET)
3299 Details : LNET is new networking infrastructure for Lustre, it includes
3300 a reorganized network configuration mode (see the user
3301 documentation for full details) as well as support for routing
3302 between different network fabrics. Lustre Networking Devices
3303 (LNDs) for the supported network fabrics have also been
3304 created for this new infrastructure.
3306 Severity : enhancement
3307 Description: Introduced Access control lists
3308 Details : clients can set ACLs on files and directories in order to have
3309 more fine-grained permissions than the standard Unix UGO+RWX.
3310 The MDS must be started with the "-o acl" mount option.
3312 Severity : enhancement
3313 Description: Introduced filesystem quotas
3314 Details : Administrators may now establish per-user quotas on the
3317 Severity : enhancement
3319 Description: Configuration change for the XT3
3320 The PTLLND is now used to run Lustre over Portals on the XT3
3321 The configure option(s) --with-cray-portals are no longer used.
3322 Rather --with-portals=<path-to-portals-includes> is used to
3323 enable building on the XT3. In addition to enable XT3 specific
3324 features the option --enable-cray-xt3 must be used.
3329 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3330 Details : A race between allocating a new llcd and re-getting the llcd_lock
3331 allowed another thread to grab newly-allocated llcd.
3333 Severity : enhancement
3335 Description: 2.6 OST async journal commit and locking fix to improve performance
3336 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3337 2.6 kernels are now async as they already were in 2.4 kernels so
3338 that they can commit concurrently with the network bulk transfer.
3339 For block-allocated files the filter allocation semaphore is held
3340 to avoid filesystem fragmentation during allocation. BKL lock
3341 removed for 2.6 xattr operations where it is no longer needed.
3346 Description: lconf incorrectly determined whether two IP networks could talk
3347 Details : In some more complicated routing and multiple-network
3348 configurations, lconf will avoid trying to make a network
3349 connection to a disjoint part of the IP space. It was doing the
3350 math incorrectly for one set of cases.
3355 Description: Fix for potential infinite loop processing records in an llog.
3356 Details : If an llog record is corrupted/zeroed, it is possible to loop
3357 forever in llog_process(). Validate the llog record length
3358 and skip the remainder of the block on error.
3361 Frequency : occasional (liblustre only)
3363 Description: liblustre could not open files whose last component is a symlink
3364 Details : sysio_path_walk() would incorrectly pass the open intent to
3365 intermediate path components.
3368 Frequency : rare (liblustre only with non-standard tuning)
3369 Bugzilla : 7201 (7350)
3370 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3371 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3372 after it has been released, only do this for referenced locks
3374 Severity : enhancement
3376 Description: specifying an (invalid) directory default stripe_size of -1
3377 would reset the directory default striping
3378 Details : stripe_size -1 was used internally to signal directory stripe
3379 removal, now use "all default" to signal dir stripe removal
3380 as a directory striping of "all default" is not useful
3383 Frequency : common for large clusters running liblustre clients
3385 Description: doing an ls when liblustre clients are running is slow
3386 Details : sending a glimpse AST to a liblustre client waits for every AST
3387 to time out, as liblustre clients will not respond. Since they
3388 cannot cache data we refresh the OST lock LVB from disk instead.
3390 Severity : enhancement
3392 Description: doing an ls at the same time as file IO can be slow
3393 Details : enqueue and other "small" requests can be blocked behind many
3394 large IO requests. Create a new OST IO portal for non-IO
3395 requests so they can be processed faster.
3398 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3400 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3401 created file with mode 000
3402 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3403 allow an ftruncate() as a non-root user to a file with mode 000.
3404 Lustre now respects this flag to disable mode checks when
3405 truncating a file owned by the user
3408 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3411 Description: Revoking locks from clients that went dead or catatonic might take
3413 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3414 cancellation of such locks instant on servers without waiting for
3415 any reply from clients. Clients drops these locks when cancel
3416 notification from server is received without replying.
3419 Frequency : liblustre-only, when liblustre client dies or becomes busy
3421 Description: Doing ls on Linux clients can take a long time with active
3423 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3424 granting such locks to it in the first place if possible. Locks
3425 are taken by proxy on the OST during the read or write and
3426 dropped immediately afterward. Add connect flags handling, do
3427 not grant locks to liblustre clients for glimpse ASTs.
3429 Severity : enhancement
3431 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3432 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3433 sequential reads and behaves badly with applications doing random
3434 reads. Improve it by reading ahead at least read region, and
3435 avoiding excessive large RPC for small reads.
3437 Severity : enhancement
3439 Description: Creating more than 1000 files for a single job may cause a load
3440 imbalance on the OSTs if there are also a large number of OSTs.
3441 Details : qos_prep_create() uses an OST index reseed value that is an
3442 even multiple of the number of available OSTs so that if the
3443 reseed happens in the middle of the object allocation it will
3444 still utilize the OSTs as uniformly as possible.
3449 Description: OST or MDS may oops in ping_evictor_main()
3450 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3451 but doesn't restart at beginning of obd_exports_timed list
3454 Severity : enhancement
3456 Description: improve by-nid export eviction on the MDS and OST
3457 Details : allow multiple exports with the same NID to be evicted at one
3458 time without re-searching the exports list.
3461 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3463 Description: MDS may oops in groups_free()
3464 Details : in rare race conditions a newly allocated group_info struct is
3465 freed again, and this can be NULL. The 2.4 compatibility code
3466 for groups_free() checked for a NULL pointer, but 2.6 did not.
3469 Frequency : common for liblustre clients doing little filesystem IO
3470 Bugzilla : 9352, 7313
3471 Description: server may evict liblustre clients accessing contended locks
3472 Details : if a client is granted a lock or receives a completion AST
3473 with a blocking AST already set it would not reply to the AST
3474 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3477 Frequency : lfs setstripe, only systems with more than 160 OSTs
3479 Description: unable to set striping with a starting offset beyond OST 160
3480 Details : llapi_create_file() incorrectly limited the starting stripe
3481 index to the maximum single-file stripe count.
3484 Frequency : LDAP users only
3486 Description: lconf did not handle in-kernel recovery with LDAP properly
3487 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3489 Severity : enhancement
3491 Description: bind OST threads to NUMA nodes to improve performance
3492 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3493 node and do their allocations there to localize memory access
3495 Severity : enhancement
3497 Description: llmount can determine client NID directly from Myrinet (GM)
3498 Details : the client NID code from gmnalnid was moved directly into
3499 llmount, removing the need to use this or specifying the
3500 client NID explicitly when mounting GM clients with zeroconf
3503 Frequency : if client is started with down MDS
3505 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3506 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3507 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3508 ptlrpc_import_delayed_req() because we didn't check if the
3509 request was interrupted, and we also didn't break out of the
3510 event loop if there was no timeout
3515 Description: data loss during non-page-aligned writes to a single file from
3516 both multiple nodes and multiple threads on one node at same time
3517 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3518 inconsistency led to false short-reads, that were cached and later
3519 used by ->prepare_write() to fill in partially written page,
3520 leading to data loss.
3523 Frequency : always, if lconf --abort_recovery used
3525 Description: lconf --abort_recovery fails with 'Operation not supported'
3526 Details : lconf was attempting to abort recovery on the MDT device and not
3529 Severity : enhancement
3531 Description: remove cleanup logs
3532 Details : replace lconf-generated cleanup logs with lustre internal
3533 cleanup routines. Eliminates the need for client-cleanup and
3536 Severity : enhancement
3538 Description: add support for EAs (user and system) on lustre filesystems
3539 Details : it is now possible to store extended attributes in the Lustre
3540 client filesystem, and with the user_xattr mount option it
3541 is possible to allow users to store EAs on their files also
3543 Severity : enhancement
3545 Description: Add possibility (config option) to show minimal available OST free
3547 Details : When compiled with --enable-mindf configure option, statfs(2)
3548 (and so, df) will return least minimal free space available from
3549 all OSTs as amount of free space on FS, instead of summary of
3550 free spaces of all OSTs.
3552 Severity : enhancement
3554 Description: do not expand extent locks acquired on OST-side
3555 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3556 by server: they are not cached anyway.
3559 Frequency : when mmap is used/binaries executed from Lustre
3561 Description: Unmmap pages before throwing them away from read cache.
3562 Details : llap_shrink cache now attempts to unmap pages before discarding
3563 them (if unmapping failed - do not discard). SLES9 kernel has
3564 extra checks that trigger if this unmapping is not done first.
3569 Description: lconf didn't resolve symlinks before checking to see whether a
3570 given mountpoint was already in use
3573 Frequency : when migrating failover services
3574 Bugzilla : 6395, 9514
3575 Description: When migrating a subset of services from a node (e.g. failback
3576 from a failover service node) the remaining services would
3577 time out and evict clients.
3578 Details : lconf --force (implied by --failover) sets the global obd_timeout
3579 to 5 seconds in order to quickly disconnect, but this caused
3580 other RPCs to time out too quickly. Do not change the global
3581 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3583 Severity : enhancement
3584 Frequency : if MDS is started with down OST
3585 Bugzilla : 9439,5706
3586 Description: Allow startup/shutdown of an MDS without depending on the
3587 availability of the OSTs.
3588 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3589 Add appropriate locking and lov-osc refcounts for safe
3590 cleaning. Add osc abort_inflight calls in case the
3591 synchronize never started.
3594 Frequency : occasional (Cray XT3 only)
3596 Description: root not authorized to access files in CRAY_PORTALS environment
3597 Details : The client process capabilities were not honoured on the MDS in
3598 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3599 been accessed by an authorized user then root was able to access
3600 the file on the local client also. The root user capabilities
3601 are now allowed on the MDS, as this environment has secure UID.
3604 Frequency : occasional
3606 Description: ldiskfs "too long searching" message happens too often
3607 Details : A debugging message (otherwise harmless) prints too often on
3608 the OST console. This has been reduced to only happen when
3609 there are fragmentation problems on the filesystem.
3614 Description: Division by zero in statfs when all OSCs are inactive
3615 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3616 lov_statfs divides by value returned by lov_get_stripecnt().
3620 Bugzilla : 9489, 3273
3621 Description: First write from each client to each OST was only 4kB in size,
3622 to initialize client writeback cache, which caused sub-optimal
3623 RPCs and poor layout on disk for the first writen file.
3624 Details : Clients now request an initial cache grant at (re)connect time
3625 and so that they can start streaming writes to the cache right
3626 away and always do full-sized RPCs if there is enough data.
3627 If the OST is rebooted the client also re-establishes its grant
3628 so that client cached writes will be honoured under the grant.
3633 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3634 Details : Now I/O RPCs go to different portal number and (presumably) fast
3635 lock requests (and glimses) and other RPCs get their own service
3636 threads pool that should be able to service those RPCs
3639 Severity : enhancement
3641 Description: Ability to exchange lustre version between client and servers and
3642 issue warnings at client side if client is too old. Also for
3643 liblustre clients there is ability to refuse connection of too old
3645 Details : New 'version' field is added to connect data structure that is
3646 filled with version info. That info is later checked by server and
3650 Frequency : rare, liblustre only.
3651 Bugzilla : 9296, 9581
3652 Description: Two simultaneous writes from liblustre at offset within same page
3653 might proceed at the same time overwriting eachother with stale
3655 Details : I/O lock withing llu_file_prwv was released too early, before data
3656 actually was hitting the wire. Extended lock-holding time until
3657 server acknowledges receiving data.
3660 Frequency : extremely rare. Never observed in practice.
3662 Description: avoid generating lustre_handle cookie of 0.
3663 Details : class_handle_hash() generates handle cookies by incrementing
3664 global counter, and can hit 0 occasionaly (this is unlikely, but
3665 not impossible, because initial value of cookie counter is
3666 selected randonly). Value of 0 is used as a sentinel meaning
3667 "unassigned handle" --- avoid it. Also coalesce two critical
3668 sections in this function into one.
3670 Severity : enhancement
3672 Description: allow liblustre clients to delegate truncate locking to OST
3673 Details : To avoid overhead of locking, liblustre client instructs OST to
3674 take extent lock in ost_punch() on client's behalf. New connection
3675 flag is added to handle backward compatibility.
3677 Severity : enhancement
3678 Bugzilla : 4928, 7341, 9758
3679 Description: allow number of OST service threads to be specified
3680 Details : a module parameter allows the number of OST service threads
3681 to be specified via "options ost ost_num_threads={N}" in the
3682 OSS's /etc/modules.conf or /etc/modprobe.conf.
3686 Bugzilla : 6146, 9635, 9895
3687 Description: servers crash with bad pointer in target_handle_connect()
3688 Details : In rare cases when a client is reconnecting it was possible that
3689 the connection request was the last reference for that export.
3690 We would temporarily drop the export reference and get a new
3691 one, but this may have been the last reference and the export
3692 was just destroyed. Get new reference before dropping old one.
3694 Severity : enhancement
3695 Frequency : if client is started with failover MDS
3697 Description: Allow multiple MDS hostnames in the mount command
3698 Details : Try to read the configuration from all specified MDS
3699 hostnames during a client mount in case the "primary"
3702 Severity : enhancement
3704 Description: Stop sending data to evicted clients as soon as possible.
3705 Details : Check if the client we are about to send or are sending data to
3706 was evicted already. (Check is done every second of waiting,
3707 for which l_wait_event interface was extended to allow checking
3708 of exit condition at specified intervals).
3711 Frequency : rare, normally only when NFS exporting is done from client
3713 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3715 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3718 Frequency : always, if srand() is called before liblustre initialization
3720 Description: Liblustre uses system PRNG disturbing its usage by user application
3721 Details : Introduce internal to lustre fast and high-quality PRNG for
3722 lustre usage and make liblustre and some other places in generic
3723 lustre code to use it.
3725 Severity : enhancement
3726 Bugzilla : 9477, 9557, 9870
3727 Description: Verify that the MDS configuration logs are updated when xml is
3728 Details : Check if the .xml configuration logs are newer than the config
3729 logs stored on the MDS and report an error if this is the case.
3730 Request --write-conf, or allow starting with --old_conf.
3732 Severity : enhancement
3734 Description: Handle symlinks in the path when checking if Lustre is mounted.
3735 Details : Resolve intermediate symlinks when checking if a client has
3736 mounted a filesystem to avoid duplicate client mounts.
3741 Description: lconf can hit an error exception but still return success.
3742 Details : The lconf command catches the Command error exception at the top
3743 level script context and will exit with the associated exit
3744 status, but doesn't ensure that this exit status is non-zero.
3749 Description: failure of ptlrpc thread startup can cause oops
3750 Details : Starting a ptlrpc service thread can fail if there are a large
3751 number of threads or the server memory is very fragmented.
3752 Handle this without oopsing.
3755 Frequency : always, only if liblustre and non-default acceptor port was used
3757 Description: liblustre cannot connect to servers with non-default acceptor port
3758 Details : tcpnal_set_default_params() was not called and was therefore
3759 ignoring the environment varaible TCPNAL_PORT, as well as other
3760 TCPNAL_ environment variables
3765 Description: two objects could be created on the same OST for a single file
3766 Details : If an OST is down, in some cases it was possible to create two
3767 objects on a single OST for a single file. No problems other
3768 than potential performance impact and spurious error messages.
3772 Bugzilla : 5681, 9562
3773 Description: Client may oops in ll_unhash_aliases
3774 Details : Client dcache may become inconsistent in race condition.
3775 In some cases "getcwd" can fail if the current directory is
3781 Description: Inode refcounting problems in NFS export code
3782 Details : link_raw functions used to call d_instantiate without obtaining
3783 extra inode reference first.
3787 Bugzilla : 9942, 9903
3788 Description: Referencing freed requests leading to crash, memleaks with NFS.
3789 Details : We used to require that call to ll_revalidate_it was always
3790 followed by ll_lookup_it. Also with revalidate_special() it is
3791 possible to call ll_revalidate_it() twice for the same dentry
3792 even if first occurence returned success. This fix changes semantic
3793 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3794 reference on a request referred from the intent.
3795 ll_intent_release() then releases such a request.
3798 Frequency : rare, normally benchmark loads only
3800 Description: unlinked inodes were kept in memory on the client
3801 Details : If a client is repeatedly creating and unlinking files it
3802 can accumulate a lot of stale inodes in the inode slab cache.
3803 If there is no other client load running this can cause the
3804 client node to run out of memory. Instead flush old inodes
3805 from client cache that have the same inode number as a new inode.
3808 Frequency : SLES9 2.6.5 kernel and long filenames only
3809 Bugzilla : 9969, 10379
3810 Description: utime reports stale NFS file handle
3811 Details : SLES9 uses out-of-dentry names in some cases, which confused
3812 the lustre dentry revalidation. Change it to always use the
3816 Frequency : rare, unless heavy write-truncate concurrency is continuous
3817 Bugzilla : 4180, 6984, 7171, 9963, 9331
3818 Description: OST becomes very slow and/or deadlocked during object unlink
3819 Details : filter_destroy() was holding onto the parent directory lock
3820 while truncating+unlinking objects. For very large objects this
3821 may block other threads for a long time and slow overall OST
3822 responsiveness. It may also be possible to get a lock ordering
3823 deadlock in this case, or run out of journal credits because of
3824 the combined truncate+unlink. Solution is to do object truncate
3825 first in one transaction without parent lock, and then do the
3826 final unlink in a new transaction with the parent lock. This
3827 reduces the lock hold time dramatically.
3830 Frequency : rare, 2.4 kernels only
3832 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3833 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3834 before unmounting their filesystem in order to avoid deadlock.
3835 The kernel_locked() function in 2.4 kernels only checks whether
3836 the kernel lock is held, not whether it is this process that is
3837 holding it as 2.6 kernels do.
3842 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3843 Details : The client ptlrpc code may be trying to reconnect to a down
3844 server before a previous connection attempt has timed out.
3845 Increase the reconnect interval to be longer than the connection
3846 timeout interval to avoid sending duplicate connections to
3850 Frequency : echo_client brw_test command
3852 Description: fix echo_client to work with OST preallocated code
3853 Details : OST preallocation code (5137) didn't take echo_client IO path
3854 into account: echo_client calls filter methods outside of any
3855 OST thread and, hence, there is no per-thread preallocated
3856 pages and buffers to use. Solution: hijack pga pages for IO. As
3857 a byproduct, this avoids unnecessary data copying.
3861 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3862 Description: Client can oops in mdc_commit_close() after open replay
3863 Details : It was possible for the MDS to return an open request with no
3864 transaction number in mds_finish_transno() if the client was
3865 evicted, but without actually returning an error. Clients
3866 would later try to replay that open and may trip an assertion
3867 Simplify the client close codepath, and always return an error
3868 from the MDS in case the open is not successful.
3871 Frequency : rare, 2.6 OSTs only
3873 Description: OST may deadlock under high load on fragmented files
3874 Details : If there was a heavy load and highly-fragmented OST filesystems
3875 it was possible to have all the OST threads deadlock waiting on
3876 allocation of biovecs, because the biovecs were not released
3877 until the entire RPC IO was completed. Instead, release biovecs
3878 as soon as they are complete to ensure forward IO progress.
3880 Severity : enhancement
3882 Description: Support for specifying external journal device at mount
3883 Details : If an OST or MDS device is formatted with an external journal
3884 device, this device major/minor is stored in the ext3 superblock
3885 and may not be valid for failover. Allow detecting and
3886 specifying the external journal at mount time.
3891 Description: Mounting an MDS with pending unlinked files may cause oops
3892 Details : target_finish_recovery() calls mds_postrecov() which returned
3893 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3894 considers this an error and immediately begins cleaning up the
3895 lov, just after starting the mds_lov process
3897 Severity : enhancement
3899 Description: Implement 'lfs df' to report actual free space on per-OST basis
3900 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3901 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3902 usage of objects; '-h' to report in human readable format.
3904 ------------------------------------------------------------------------------
3906 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
3913 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3914 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3915 at startup during orphan unlinking. Instead, skip buddy update
3916 before setup, as it will be regenerated after recovery is complete.
3921 Description: If an OST is inactive, its locks might reference stale inodes.
3922 Details : lov_change_cbdata() must iterate over all namespaces, even if
3923 they are inactive to clear inode references from the lock.
3925 Severity : enhancement
3926 Frequency : occasional, if non-standard max_dirty_mb used
3928 Description: Client will block write RPCs if not enough grant
3929 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3930 then the client will block writes while waiting for another RPC
3931 to complete instead of consuming its dirty limit. With change
3932 we get improved performance when max_dirty_mb is small.
3934 Severity : enhancement
3935 Bugzilla : 3389, 6253
3936 Description: Add support for supplementary groups on the MDS.
3937 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3938 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3939 lookups for user supplementary groups into a cache.
3943 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3944 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3945 for liblustre O_EXCL is enforced by MDS.
3949 Description: readdir never returns NULL in liblustre.
3950 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3951 dentry in struct dirent.
3955 Frequency : liblustre only, depends on application IO pattern
3956 Description: liblustre clients evicted if not contacting servers
3957 Details : Don't put liblustre clients into the ping_evictor list, so
3958 they will not be evicted by the pinger ever.
3960 Severity : enhancement
3962 Description: Add ability to evict clients by NID from MDS.
3963 Details : By echoing "nid:$NID" string into
3964 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3965 $NID would be instantly evicted from this MDS and from all active
3966 OSTs connected to it.
3970 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3971 Details : lookup_it_finish() used to query file size from OSTs that was not
3976 Description: service threads change working directory to that of init
3977 Details : Starting lustre service threads may pin the working directory
3978 of the parent thread, making that filesystem busy. Threads
3979 now change to the working directory of init to avoid this.
3983 Frequency : during shutdown only
3984 Description: shutdown with a failed MDS or OST can cause unmount to hang
3985 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3990 Frequency : 2.6 only
3991 Description: chmod/chown may include an extra supplementary group
3992 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3993 supplementary group and if none is specified this is used.
3996 Bugzilla : 5479 (6816)
3998 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3999 Details : If lookup is in progress on a file that is unlinked we might try
4000 to revalidate the inode and fail in revalidate after lookup is
4001 complete and ll_file_open() enqueues the open again but
4002 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4005 Frequency : always, if lconf --abort_recovery used
4007 Description: lconf --abort_recovery fails with 'Operation not supported'
4008 Details : lconf was attempting to abort recovery on the MDT device and not
4011 ------------------------------------------------------------------------------
4013 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4018 Frequency : rare (only unsupported configurations with a node running as an
4020 Bugzilla : 6514, 5137
4021 Description: Mounting a Lustre file system on a node running as an OST could
4023 Details : OSTs now preallocates memory needed to write out data at
4024 startup, instead of when needed, to avoid having to
4025 allocate memory in possibly low memory situations.
4026 Specifically, if the file system is mounted on on OST,
4027 memory pressure could force it to try to write out data,
4028 which it needed to allocate memory to do. Due to the low
4029 memory, it would be unable to do so and the node would
4030 become unresponsive.
4032 Severity : enhancement
4034 Description: Addition of lconf --service command line option
4035 Details : lconf now accepts a '--service <arg>' option, which is
4036 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4038 Severity : enhancement
4040 Description: Failover mode is now the default for OSTs.
4041 Details : By default, OSTs will now run in failover mode. To return to
4042 the old behaviour, add '--failout' to the lmc line for OSTs.
4044 Severity : enhancement
4046 Description: Health checks are now provided for MDS and OSTs
4047 Details : Additional detailed health check information on MSD and OSTs
4048 is now provided through the procfs health_check value.
4051 Frequency : occasional, depends on IO load
4053 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4054 after numerous create/delete cycles
4055 Details : The ext3 inode allocation policy would not allocate new inodes
4056 very well on the OSTs because there are no new directories
4057 being created. Instead we look for groups with free space if
4058 the parent directories are nearly full.
4063 Description: Network or server problems during mount may cause partially
4064 mounted clients instead of returning an error.
4065 Details : The config llog parsing code may overwrite the error return
4066 code during mount error handling, returning success instead
4072 Description: MDS can fail to allocate large reply buffers
4073 Details : After long uptimes the MDS can fail to allocate large reply
4074 buffers (e.g. zconf client mount config records) due to memory
4075 fragmentation or consumption by the buffer cache. Preallocate
4076 some large reply buffers so that these replies can be sent even
4077 under memory pressure.
4081 Frequency : rare (liblustre)
4082 Description: fsx running with liblustre complained that using truncate() to
4083 extend the file doesn't work. This patch corrects that issue.
4084 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4085 ATTR_SIZE and lsm use in llu_setattr_raw.
4089 Frequency : rare, only 2.6 kernels
4090 Description: Unusual file access patterns on the MDS may result in inode
4091 data being lost in very rare circumstances.
4092 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4093 code in the 2.6 kernel caused the inode and/or EA data not to
4094 be read from disk, causing single-file corruption.
4098 Frequency : rare, only 2.6 filesystems using extents
4099 Description: Heavy concurrent write and delete load may cause data corruption.
4100 Details : It was possible under high-load situations to have an extent
4101 metadata block in the block device cache from a just-unlinked
4102 file overwrite a newly-allocated data block. We now unmap any
4103 metadata buffers that alias just-allocated data blocks.
4107 Frequency : filesystems with default stripe_count larger than 77
4108 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4109 Details : lconf specifies an inode size of 4096 bytes when the default
4110 stripe_count is larger than 77. This conflicts with the default
4111 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4112 this case to avoid pinning too much memory for large EAs.
4114 ------------------------------------------------------------------------------
4116 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4121 Frequency : rare (extremely heavy IO load with hundreds of clients)
4123 Description: Client is evicted, gets IO error writing to file
4124 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4125 caused clients to be evicted for AST timeouts. The fixes in
4126 bug 5192 mean we no longer need to have such short AST timeouts
4127 so ldlm_timeout has been increased.
4130 Frequency : occasional during --force or --failover shutdown under load
4131 Bugzilla : 5949, 4834
4132 Description: Server oops/LBUG if stopped with --force or --failover under load
4133 Details : a collection of import/export refcount and cleanup ordering
4134 issues fixed for safer force cleanup
4137 Frequency : only filesystems larger than 120 OSTs
4138 Bugzilla : 5990, 6223
4139 Description: lfs getstripe would oops on a very large filesystem
4140 Details : lov_getconfig used kfree on vmalloc'd memory
4143 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4144 Bugzilla : 6242, 6243
4145 Description: reading from files that had been truncated to a non-zero size
4146 but never opened returned no data
4147 Details : ll_file_read() reads zeros from no-object files to EOF
4152 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4153 Details : There's little harm in aborting MDS/OSS recovery and letting it
4154 try again, so I removed the LASSERT and return an error instead.
4156 Severity : enhancement
4158 Description: New debugging infrastructure for tracking down data corruption
4159 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4160 (b) cover more of the client-side code path, and (c) try to narrow
4161 down where problems occurred
4165 Bugzilla : 3819, 4364, 4397, 6313
4166 Description: Racing close and eviction MDS could cause assertion in mds_close
4167 Details : It was possible to get multiple mfd references during close and
4168 client eviction, leading to one thread referencing a freed mfd.
4170 Severity: : enhancement
4171 Bugzilla : 3262, 6359
4172 Description: Attempts to reconnect to servers are now more aggressive.
4173 Details : This builds on the enhanced upcall-less recovery that was added
4174 in 1.4.2. When trying to reconnect to servers, clients will
4175 now try each server in the failover group every 10 seconds. By
4176 default, clients would previously try one server every 25 seconds.
4181 Description: After recovery, certain operations trigger a failed
4182 assertion on a client.
4183 Details : Failing over an mds, using lconf -d --failover, while a
4184 client was doing a readdir() call would cause the client to
4185 LBUG after recovery completed and the readdir() was resent.
4187 Severity : enhancement
4189 Description: Default groups are now added by lconf
4190 Details : You can now run lconf --group <servicename> without having to
4191 manually add groups with lmc.
4194 Frequency : occasional
4196 Description: Nodes with an elan id of 0 trigger a failed assertion
4199 Frequency : always when accessing e.g. tty/console device nodes
4201 Description: tty and some other devices nodes cannot be used on lustre
4202 Details : file's private_data field is used by device data and lustre
4203 values in there got lost. New field was added to struct file to
4204 store fs-specific private data.
4207 Frequency : when exporting Lustre via NFS
4209 Description: NFSD failed occasionally when looking up a path component
4210 Details : NFSD is looking up ".." which was broken in ext3 directories
4211 that had grown large enough to become hashed.
4214 Frequency : Clusters with multiple interfaces not on the same subnet
4216 Description: Nodes will repeatedly try to reconnect to an interface which it
4217 cannot reach and report an error to the log.
4218 Details : Extra peer list entries will be created by lconf with some peers
4219 unreachable. lconf now validates the peer before adding it.
4222 Frequency : Only if a default stripe is set on the filesystem root.
4224 Description: Setting a default stripe on the filesystem root prevented the
4225 filesystem from being remounted.
4226 Details : The client was sending extra request flags in the root getattr
4227 request and did not allocate a reply buffer for the dir EA.
4230 Frequency : occasional, higher if lots of files are accessed by one client
4231 Bugzilla : 6159, 6097
4232 Description: Client trips assertion regarding lsm mismatch/magic
4233 Details : While revalidating inodes the VFS looks up inodes with ifind()
4234 and in rare cases can find an inode that is being freed.
4235 The ll_test_inode() code will free the lsm during ifind()
4236 when it finds an existing inode and then the VFS later attaches
4237 this free lsm to a new inode.
4241 Bugzilla : 6422, 7030
4242 Description: MDS deadlock between mkdir and client eviction
4243 Details : Creating a new file via mkdir or mknod (starting a transaction
4244 and getting the ns lock) can deadlock with client eviction
4245 (gets ns lock and trying to finish a synchronous transaction).
4248 Frequency : occasional
4249 Description: While starting a server, the fsfilt_ext3 module could not be
4251 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4252 kernels. Previously, lconf would still use the ext3 name
4253 when trying to load modules. Now, it will correctly use
4254 ext3 on 2.4 and ldiskfs on 2.6.
4256 Severity : enhancement
4257 Description: The default stripe count has been changed to 1
4258 Details : The interpretation of the default stripe count (0, to lfs
4259 or lmc) has been changed to mean striping across a single
4260 OST, rather than all available. For general usage we have
4261 found a stripe count of 1 or 2 works best.
4263 Severity : enhancement
4264 Description: Add support for compiling against Cray portals.
4265 Details : Conditional compiling for some areas that are different
4269 Frequency : occasional
4270 Bugzilla : 6409, 6834
4271 Description: Creating files with an explicit stripe count may lead to
4272 a failed assertion on the MDS
4273 Details : If some OSTs are full or unavailable, creating files may
4274 trigger a failed assertion on the MDS. Now, Lustre will
4275 try to use other servers or return an error to the
4279 Frequency : occasional
4281 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4282 caused lock timeout during readahead (since 1.4.2).
4283 Details : Processes doing readahead might match a lock that hasn't been
4284 granted yet if there are overlapping and conflicting lock
4285 requests. The readahead process waits on ungranted lock
4286 (original lock is CBPENDING), while OST waits for that process
4287 to cancel CBPENDING read lock and eventually evicts client.
4289 Severity : enhancement
4291 Description: Initial enabling of flock support for clients
4292 Details : Implements fcntl advisory locking and file status functions.
4293 This feature is provided as an optional mount flag (default
4294 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4295 locking are implemented yet, and those that are are not guaranteed
4296 to be completely correct in production environments.
4297 mount -t lustre -o [flock|noflock] ...
4300 Frequency : occasional
4302 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4303 assertion in the ext3 JBD (journaling) layer.
4304 Details : The b_committed_data struct is protected by the big kernel lock
4305 in 2.4 kernels, serializing journal_commit_transaction() and
4306 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4307 access to this struct. In 2.6 kernels there is finer grained
4308 locking to improve SMP performance of the JBD layer.
4312 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4314 -----------------------------------------------------------------------------
4316 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4318 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4319 versions of Lustre. Please update all servers and clients to
4320 version 1.4.2 or later at the same time. You must also run
4321 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4322 to update the configuration logs.
4324 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4325 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4326 - swab llog records of type '0' so we get proper header size/idx (5861)
4327 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4328 - fix rename of one directory over another leaking an inode (5953)
4329 - avoid SetPageDirty on 2.6 (5981)
4330 - don't re-add just-being-destroyed locks to the waiting list (5653)
4331 - when creating new directories, inherit the parent's custom
4332 striping settings if present parent (3048)
4333 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4334 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4335 - don't hold client locks on temporary worklist from l_lru (5666)
4336 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4337 - automatically evict dead clients (5921)
4338 - Update file size properly in create+truncate+fstat case (6196)
4339 - Do not unhash mountpoint dentries, do not allow removal of
4341 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4342 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4343 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4344 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4345 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4346 - Correctly handle reads of files with no objects (6243)
4347 - lctl recover will also mark a device active if deactivate used (5933)
4349 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4350 - allow --write-conf on an MDS with different nettype than client (5619)
4351 - don't write config llogs to MDS for mounts not from that MDS (5617)
4352 - lconf should create multiple TCP connections from a client (5201)
4353 - init scripts are now turned off by default; run chkconfig --on
4354 lustre and chkconfig --on lustrefs to use them
4355 - upcalls are no longer needed for clients to recover to failover
4357 - add --abort-recovery option to lconf to abort recovery on device
4359 - add support for an arbitrary number of OSTs (3026)
4360 - Quota support protocol changes.
4361 - forward compatibility changes to wire structs (6007)
4362 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4363 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4364 - improved "lustre status" script
4365 - initialize blocksize for non-regular files (6062)
4366 - added --disable-server and --disable-client configure options (5782)
4367 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4368 - Vanilla 2.4.29 support
4369 - increase maximum number of obd devices to 520 (6242)
4370 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4371 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4372 SLES 9 2.6.5 kernels
4373 - limit stripes per file to 160 (the maximum EA size) (6093)
4375 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4378 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4379 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4380 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4381 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4382 - small ext3 extents cleanups and fixes (5733)
4383 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4384 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4385 - avoid needless client->OST connect, fix handle mismatch (5317)
4386 - fix DLM error path that led to out-of-sync client, long delays (5779)
4387 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4388 - fix several locking issues related to i_size (5492,5624,5654,5672)
4389 - don't move pending lock onto export if it is already evicted (5683)
4390 - fix kernel oops when creating .foo in unlinked directory (5548)
4391 - fix deadlock in obdfilter statistics vs. object create (5811)
4392 - use time_{before,after} to avoid timer jiffies wrap (5882)
4393 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4394 - Do not leak request if server was not able to process it (5154)
4395 - If mds_open unable to find parent dir, make that negative lookup(5154)
4396 - don't create new directories with extent-mapping (5909, 5936)
4398 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4399 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4400 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4401 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4402 - support for RHEL4 kernel on i686 (5773)
4403 - provide error messages when incompatible logs are encountered (5898)
4405 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4406 * version 1.4.0.10 (1.4.1 release candidate 1)
4408 - don't keep a lock reference when lock is not granted (4238)
4409 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4410 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4411 - drop import inflight refcount on signal_completed_replay error (5255)
4412 - unlock page after async write error during send (3677)
4413 - handle missing objects in filter_preprw_read properly (5265)
4414 - no transno return for symlink open, don't save no-trasno open (3440)
4415 - don't try to complete elan receive that already failed (4012)
4416 - free RPC server reply state on error (5406)
4417 - clean up thread from ptlrpc_start_thread() on error (5160)
4418 - readahead could read extra page into cache that wasn't ejected (5388)
4419 - prevent races in class_attach/setup/cleanup/detach (5260)
4420 - don't dereference de->d_inode after l_dput of de (5458)
4421 - use "int" for stripe value returned from lock_to_stripe (5544)
4422 - mballoc allocation and error-checking fixes in 2.6 (5504)
4423 - block device patches to fix I/O request sizes in 2.6 (5482)
4424 - look up hostnames for IB nals (5602)
4425 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4426 - don't start multiple acceptors for the same port (5277)
4427 - fix incorrect LASSERT in mds_getattr_name (5635)
4428 - export a proc file for general "ping" checking (5628)
4429 - fix "lfs check" to not block when the MDS is down (5628)
4431 - service request history (4965)
4432 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4433 - create an "evict_client" /proc entry on OSTs, like the MDS has
4434 - fix mount usage message, return errors per mount(8) (5168)
4435 - change grep [] to grep "[]" in tests so they work in more UMLs
4436 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4437 - remove ancient LOV_MAGIC_V0 EA support (5047)
4438 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4439 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4440 - allow readahead allocations to fail when low on memory (5383)
4441 - mmap locking landed again, after considerable improvement (2828)
4442 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4444 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4447 - send OST transaction number in read/write reply to free req (4966)
4448 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4449 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4450 - the watchdog thread now runs as interruptible (5246)
4451 - flock/lockf fixes (but it's still disabled, pending 5135)
4452 - don't use EXT3 constants in llite code (5094)
4453 - memory shortage at startup could cause assertion (5176)
4455 - reorganization of lov code
4456 - single portals codebase
4458 - add extents/mballoc support (5025)
4459 - direct I/O reads in the obdfilter (4048)
4460 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4462 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4465 - send OST transaction number in read/write reply to free req (4966)
4466 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4467 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4468 - the watchdog thread now runs as interruptible (5246)
4469 - handle missing objects in filter_preprw_read properly (5265)
4470 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4471 - drop import inflight refcount on signal_completed_replay error (5255)
4472 - unlock page after async write error during send (3677)
4473 - return original error code on reconstructed replies (3761)
4474 - no transno return for symlink open, don't save no-trasno open (3440)
4476 - add pid to ldlm debugging output (4922)
4477 - bump the watchdog timeouts -- we can't handle 30sec yet
4478 - extra debugging for orphan dentry/inode bug (5259)
4480 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4483 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4484 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4485 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4486 - fix NULL dereference in /proc/sys/portals/routes (4827)
4487 - allow failed mdc_close() operations to be interrupted (4561)
4488 - stop precreate on OST before MDS would time out on it (4778)
4489 - don't send partial-page writes before EOF from client (4410)
4490 - discard client grant for sub-page writes on large-page clients (4520)
4491 - don't free dentries not owned by NFS code, check generation (4806)
4492 - fix lsm leak if mds_create_objects() fails (4801)
4493 - limit debug_daemon file size, always print CERROR messages (4789)
4494 - use transno after validating reply (3892)
4495 - process timed out requests if import state changes (3754)
4496 - update mtime on OST during writes, return in glimpse (4829)
4497 - add mkfsoptions to LDAP (4679)
4498 - use ->max_readahead method instead of zapping global ra (5039)
4499 - don't interrupt __l_wait_event() during strace
4501 - add software watchdogs to catch hung threads quickly (4941)
4502 - make lustrefs init script start after nfs is mounted
4503 - fix CWARN/ERROR duplication (4930)
4504 - return async write errors to application if possible (2248)
4505 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4506 - print NAL number in %x format (4645)
4507 - update barely-supported suse-2.4.21-171 series (4842)
4508 - support for sles 9 %post scripts
4509 - support for building 2.6 kernel-source packages
4510 - support for sles km_* packages
4512 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4515 - ignore -ENOENT errors in osc_destroy (3639)
4516 - notify osc create thread that OSC is being cleaned up (4600)
4517 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4518 - reconstruct ost_handle() like mds_handle() (4657)
4519 - create a new thread to do import eviction to avoid deadlock (3969)
4520 - let lconf resolve symlinked-to devices (4629)
4521 - don't unlink "objects" from directory with default EA (4554)
4522 - hold socknal file ref over connect in case target is down (4394)
4523 - allow more than 32000 subdirectories in a single directory (3244)
4524 - fix blocks count for O_DIRECT writes (3751)
4525 - OST returns ENOSPC from object create when no space left (4539)
4526 - don't send truncate RPC if file size isn't changing (4410)
4527 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4528 - bind to privileged port in socknal and tcpnal (3689)
4530 - rate limit CERROR/CWARN console message to avoid overload (4519)
4531 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4532 - basic mmap support (3918)
4533 - kernel patch series update from b1_4 (4711)
4535 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4538 - avoid crash during MDS cleanup with OST shut down (2775)
4539 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4540 - don't use bad inodes on the MDS (3744)
4541 - dynamic object preallocation to improve recovery speed (4236)
4542 - don't hold spinlock over lock dumping or change debug flags (4401)
4543 - don't zero obd_dev when it is force cleaned (3651)
4544 - print grants to console if they go negative (4431)
4545 - "lctl deactivate" will stop automatic recovery attempts (3406)
4546 - look for existing locks in ldlm_handle_enqueue() (3764)
4547 - don't resolve lock handle twice in recovery avoiding race (4401)
4548 - revalidate should check working dir is a directory (4134)
4550 - don't always mark "slow" obdfilter messages as errors (4418)
4552 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4555 - don't close LustreDB during write_conf until it is done (3860)
4556 - fix typo in lconf for_each_profile (3821)
4557 - allow dumping logs from multiple threads at one time (3820)
4558 - don't allow multiple threads in OSC recovery (3812)
4559 - fix debug_size parameters (3864)
4560 - fix mds_postrecov to initialize import for llog ctxt (3121)
4561 - replace config semaphore with spinlock (3306)
4562 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4563 - don't allow enqueue to complete on a destroyed export (3822)
4564 - down write_lock before checking llog header bitmap (3825)
4565 - recover from lock replay timeout (3764)
4566 - up llog sem before sending rpc (3652)
4567 - reduce ns lock hold times when setting kms (3267)
4568 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4569 - fix NULL deref and obd_dev leak on setup error (3312)
4570 - replace some LBUG about llog ops with error handling (3841)
4571 - don't match INVALID dentries from d_lookup and spin (3784)
4572 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4573 - fix invalid assertion in ptlrpc_set_wait (3880)
4575 - add libwrap support for the TCP acceptor (3996)
4576 - add /proc/sys/portals/routes for non-root route listing (3994)
4577 - allow setting MDS UUID in .xml (2580)
4578 - print the stack of a process that LBUGs (4228)
4580 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4583 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4584 - make sure to unset replay flag from failed open requests (3440)
4585 - if default stripe count is 0, use OST count for inode size (3636)
4586 - update parent mtime/ctime on client for create/unlink (2611)
4587 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4588 - free recovery state on server during a forced cleanup (3571)
4589 - unregister_reply for resent reqs (3063)
4590 - loop back devices mounting and status check on 2.6 (3563)
4591 - fix resource-creation race that can provoke i_size == 0 (3513)
4592 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4593 - more debugging for page-accounting assertion (3746)
4594 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4595 - avoid deadlock after precreation failure (3758)
4596 - fix race and lock order deadlock in orphan handling (3450, 3750)
4597 - add validity checks when grabbing inodes from l_ast_data (3599)
4599 - add /proc/.../recovery_status to obdfilter (3428)
4600 - lightweight CDEBUG infrastructure, debug daemon (3668)
4601 - change default OSC RPC parameters to be better on small clusters
4602 - turn off OST read cache for files smaller than 32MB
4603 - install man pages and include them in rpms (3100)
4604 - add new init script for (un)mounting lustre filesystems (2593)
4605 - run chkconfig in %post for init scripts (3701)
4606 - drop scimac NAL (unmaintained)
4608 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4611 - clean kiobufs before and after use (3485)
4612 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4613 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4614 - change init script to fail more clearly if not run as root (1528)
4615 - allow clients to reconnect during replay (1742)
4616 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4617 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4618 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4619 - fix cond_resched (3554)
4620 - only evict unfinished clients after recovery (3515)
4621 - allow bulk resend, prevent data loss (3570)
4622 - dynamic ptlrpc request buffer allocation (2102)
4623 - don't allow unlinking open directory if it isn't empty (2904)
4624 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4625 - remove extraneous obd dereference causing LASSERT failure (3334)
4626 - don't use get_cycles() when creating temp. files on the mds (3156)
4627 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4628 - handle EEXIST for set-stripe, set proper directory name (3336)
4630 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4631 - fix ksocknal_fmb_callback() error messages (2918)
4633 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4636 - don't copy lvb into (possibly NULL) reply on error (2983)
4637 - don't deref dentry after dput, don't free lvb on error (2922)
4638 - use the kms to determine writeback rpc length (2947)
4639 - increment oti_logcookies when osc is inactive (2948)
4640 - update client's i_blocks count via lvb messages (2543)
4641 - handle intent open/close of special files properly (1557)
4642 - mount MDS with errors=remount-ro, like obdfilter (2009)
4643 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4644 - don't use cancelling-locks' kms values (2947)
4645 - use highest lock extent for kms, not last one (2925)
4646 - don't dereference ERR_PTR() dentry in error handling path (3107)
4647 - fix thread race in portals_debug_dumplog() (3122)
4648 - create lprocfs device entries at setup instead of at attach (1519)
4649 - common AST error handler, don't evict client on completion race (3145)
4650 - zero nameidata in detach_mnt in 2.6 (3118)
4651 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4652 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4653 - initialize RPC timeout timer earlier for 2.6 (3219)
4654 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4655 - print nal/nid for unknown nid (3258)
4656 - additional checks for oscc recovery before doing precreate (3284)
4657 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4658 - don't crash in mdc_close for bad permissions on open (3285)
4659 - zero i_rdev for non-device files (3147)
4660 - clear page->private before handing to FS, better assertion (3119)
4661 - tune the read pipeline (3236)
4662 - fix incorrect decref of invalidated dentry (2350)
4663 - provide read-ahead stats and refine rpc in flight stats (3328)
4664 - don't hold journal transaction open across create RPC (3313)
4665 - update atime on MDS at close time (3265)
4666 - close LDAP connection when recovering to avoid server load (3315)
4667 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4668 - don't leak open file on MDS after open resend (3325)
4669 - serialize filter_precreate and filter_destroy_precreated (3329)
4670 - loop device shouldn't call sync_dev() for nul device (3092)
4671 - clear page cache after eviction (2766)
4672 - resynchronize MDS->OST in background (2824)
4673 - refuse to mount the same filesystem twice on same mountpoint (3394)
4674 - allow llmount to create routes for mounting behind routers (3320)
4675 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4676 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4677 - fix rare problem with rename on htree directories (3417)
4679 - allow default OST striping configuration per directory (1414)
4680 - fix compilation for qswnal for 2.6 kernels (3125)
4681 - increase maximum number of MDS request buffers for large systems
4682 - change liblustreapi to be useful for external progs like lfsck (3098)
4683 - increase local configuration timeout for slow disks (3353)
4684 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4686 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4689 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4690 - don't overwrite extent policy data in reply if lock was blocked (2901)
4691 - drop filter export grants atomically with removal from device (2663)
4692 - del obd_self_export from work_list in class_disconnect_exports (2908)
4693 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4694 - swab reply message in mdc_close, other PPC fixes (2464)
4695 - fix destroying of named logs (2325)
4696 - overwrite old logs when running lconf --write_conf (2264)
4697 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4698 - fix race in target_handle_connect (2898)
4699 - mds_reint_create() should take same inode create lock (2926)
4700 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4701 - don't close files for self_export to avoid uninitialized obd (2936)
4702 - allow MDS with the same name as client node (2939)
4703 - hold dentry reference for closed log files for unlink (2325)
4704 - reserve space for all logs during transactions (2059)
4705 - don't evict page beyond end of stripe extent (2925)
4706 - don't oops on a deleted current working directory (2399)
4707 - handle hard links to targets without a parent properly (2517)
4708 - don't dereference NULL lock when racing during eviction (2867)
4709 - don't grow lock extents when lots of conflicting locks (2919)
4711 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4714 - account for cache space usage on clients to avoid data loss (974)
4715 - lfsck support in lustre kernel code (2349)
4716 - reduce journal credits needed for BRW writes (2370)
4717 - orphan handling to avoid losing space on client/server crashes
4718 - ptlrpcd can be blocked, stopping ALL progress (2477)
4719 - use lock value blocks to assist in proper KMS, faster stat (1021)
4720 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4721 - recovery for initial connections (2355)
4722 - fixes for mds_cleanup_orphans (1934)
4723 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4724 - block all file creations until orphan recovery completes (1901)
4725 - client remove rq_connection from request struct (2423)
4726 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4727 - recovery timer race (2670)
4728 - mdc_close recovey bug (2532)
4729 - ptlrpc cleanup bug (2710)
4730 - mds timeout on local locks (2588)
4731 - namespace lock held during RPCs (2431)
4732 - handle interrupted sync write properly (2503)
4733 - don't try to handle a message that hasn't been replied to (2699)
4734 - client assert failure during cleanup after abort recovery (2701)
4735 - leak mdc device after failed mount (2712)
4736 - ptlrpc_check_set allows timedout requests to complete (2714)
4737 - wait for inflight reqs when ptlrpcd finishes (2710)
4738 - make sure unregistered services are removed from the srv_list
4739 - reset bulk XID's when resending them (caught by 1138 test)
4740 - unregister_bulk after timeout
4741 - fix lconf error (2694)
4742 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4743 - readahead locks pages, leaves pending causing memory pressure (2673)
4744 - increase OST request buffers to 4096 on large machines (2729)
4745 - fix up permission of existing directories in simple_mkdir (2661)
4746 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4747 - don't assign transno to errored transactions (2742)
4748 - don't delete objects on OST if given a bogus objid from MDS (2751)
4749 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4750 - if rq_no_resend, then timeout request after recovery (2432)
4751 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4752 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4753 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4754 - MDS recovery shouldn't delete objects that it has given out (2730)
4755 - if enqueue arrives after completion, don't clobber LVB (2819)
4756 - don't unlock pages twice when trigger_group_io returns error (2814)
4757 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4758 - don't write pages to disk if there was an error (1450)
4759 - don't ping imports that have recovery disabled (2676)
4760 - take buffered bytes into account when balancing socknal conn (2817)
4761 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4762 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4763 - remove little-endian swabbing of llog records (1987)
4764 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4765 - retry reposting request buffers if they fail (1191)
4766 - grow extent at grant time to avoid granting a revoked lock (2809)
4767 - lock revoke doesn't evict page if covered by a second lock (2765)
4768 - disable VM readahead to avoid reading outside lock extents (2805)
4770 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4771 - updated kernel patches for hp-2.4.20 kernel (2681)
4773 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4776 - fix truncated write corruption (2366)
4777 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4779 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4780 - ensure len(uuid) < 37 in lmc (1171)
4781 - fix ia64 OOPS in llog_test (2255)
4782 - zero end of page at obdfilter for partial page writes (2648)
4783 - don't leave stale dentries around after renames (bug 2428)
4784 - fix timeouts when evicting a client with a single lock held (2642)
4785 - set deadline for the initial HELLO message to drain (2634)
4786 - print out dotted-quad IP addresses in the socknal (2302)
4788 - additional debugging for MDS client eviction problem (2443)
4789 - fix mkfsoptions support for osts (2603, 2604)
4791 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4794 - add series for the vanilla 2.6.0 kernel
4795 - add series for the vanilla 2.4.24 kernel
4796 - add series for a cray x86/64 UL kernel drop
4797 - fix xattr patches for the vanilla 2.4.19 series
4799 - generate true UUIDs in lmc (1171)
4800 - have portals stack dumping break in UML (2466)
4801 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4802 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4803 - wake callback waiting threads on client eviction (2460)
4804 - Add --ptldebug and --subsystem to lmc (1719)
4805 - update assertion to allow safe interrupt allocation
4806 - set rq_no_resend for cancel requests (2432)
4807 - recalculate ptlrpcd timeout after resend (2494)
4808 - call vfs_rmdir when removing pending directories (2368)
4809 - fix renaming a file to itself (2429)
4810 - lmc creates a default one-stripe lov (2454)
4811 - expand procfs space to handle large clusters (2326)
4812 - increase UML stack to avoid overflow
4813 - update lconf's list of debug and subsystem masks
4814 - fix lfs find --obd (2510)
4815 - /proc tunable for disabling filter read caching (2591)
4816 - stop rpm packages from altering slapd.conf (2301)
4817 - disable nagle in the socknal under 0conf (2578)
4818 - choose mds inode size based on stripe count (2572)
4819 - fix kernel-source rpm problems (2516)
4821 - add --disable-doc to avoid pdf generation (2421)
4822 - update documentation, tests, type-os, comments
4823 - avoid format warnings on ia64
4824 - remove the TOE NAL
4825 - tiny code cleanups by removing unused fields
4827 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4830 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4831 - remove the most blatant lies from BUILDING file (2371)
4832 - change default debug level to reasonable production setting
4833 - reduce client side cache size to reduce cache flush time
4834 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4835 - make TCP zerocopy and pinger support enabled by default (2476)
4836 - sync writes completed after process exits caused crashes (2319)
4837 - maintain correct mount count on the MDS (2356)
4838 - backout 1557, because 2316 wasn't really fixed
4839 - better file I/O statistics gathering in /proc
4840 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4841 - another kernel patch to fix zero-copy TCP function export
4842 - don't take duplicate lock when processing re-sent getattr (2420)
4843 - lctl uses obd_self_export instead of creating new conn (2353)
4844 - MDS/OST recovery case which requires object creation asserted (2425)
4845 - move lfs from /usr/sbin to /usr/bin in packages
4846 - fix race between mds_client_add and mds_client_free (2417)
4847 - use kmalloc instead of slabs in portals (2430)
4848 - don't create duplicate records when a failover MDS is present (2442)
4849 - remove unnecessary mount age check (2332)
4850 - don't remove directory inodes from locks prematurely (2451)
4851 - don't break if MDS service name is the same as hostname (2103)
4852 - fix races in client write RPC generation when cache full (2482)
4854 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4857 - remove now-unused request->rq_obd (278)
4858 - if an allocation fails, print out how much memory we've used (1933)
4859 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4860 - add the "configurable stack size" patch to most series files (1256)
4861 - ability to write large log records, for 100+ OST configs (2306)
4862 - fix NULL deref when filter_prep fails (2314)
4863 - fix operator precedence error in filter_sync
4864 - dynamic allocation of socknal TX descriptors (2315)
4865 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4866 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4867 - ensure that CWARN messages in Portals always get to the syslog
4868 - __init/__exit are not for prototype decls (ldlm_init/exit)
4869 - x86-64 compile warning fixes
4870 - fix gateway LMC keyword conflict (2318)
4871 - fix MDS lock inversions in getattr/reint paths (1844)
4872 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4873 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4874 - DLM race condition prevented some lock evictions (2328)
4875 - ENOMEM detection and retry on socknal sends (2230)
4876 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4877 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4878 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4879 - better error messages when a client is rejected during recovery (1505)
4880 - avoid cancelling locks which were never granted, after failure (2330)
4881 - fix i_sem/journal inversion in mds_client_add (2333)
4882 - fix truncate/getattr lock cycle deadlock (2334)
4883 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4884 - fix two rare exit paths which could leak an l_lock() ref (2321)
4885 - fencepost error in MDS/OST orphan recovery (2226)
4886 - make log record alignment 8 bytes (1988)
4887 - lstripe now fails when requested offset > ost_count (2237)
4888 - ensure that all kernel series have a complete list.h (1607)
4889 - fix crashes in special-file operations (2316)
4890 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4892 - allow configurable automake binary, for testing new versions
4893 - small update to the lfs documentation
4895 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4897 * fix negative export reference count in fsfilt_sync (2312)
4899 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4900 * release candidate 0.9.1
4902 - orphans are moved into the PENDING directory for possible recovery
4903 - replayed opens now open by fid for orphan/rename safety (1042)
4904 - last close of an orphan inode generates a transno (683)
4905 - chdir() and mount() now pin the directory entry (1020)
4906 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4907 - discard very old requests without processing them (1502)
4908 - remove some common, well-understood CERRORs (1505)
4909 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4910 - clear "grant" flags in OST replies until OST grant code lands (1644)
4911 - fix read performance by not clobbering i_blksize on client (1598)
4912 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4913 - make LRU size a /proc tunable, clears locks when reduced (707)
4914 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4915 - server locks take references on exports now (1558)
4916 - build fixes for 2.4.20-rh trees (1663)
4917 - return an error from lov_create if all OSCs are inactive (1751)
4918 - fix import levels when a reconnect happens without a timeout (1597)
4919 - exit early from mds_open if we get a lookup error (1749)
4920 - partial page read at EOF wouldn't wait for disk before sending (1642)
4921 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4922 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4923 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4924 - write/truncate lock inversion (1639)
4925 - Don't auto-load obdclass, portals modules during cleanup (1495)
4926 - fix timestamps from jumping to "now" (1763)
4927 - extra journal assertions (1648)
4928 - add an extra multiunlink test (1771)
4929 - fix read_record/write_record API (1776)
4930 - fix leak of offset_extent, possible incorrect i_size later (1772)
4931 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4932 - Debugging for the kqswnal_get_idle_tx problems (1820)
4933 - Allow recovery to be attempted multiple times (1536)
4934 - Write out MDS last_rcvd file after it is first created (1600)
4935 - Fix tx_descriptor leak in failed transmit situations (1827)
4936 - ext3 journaling fixes for assertion failure after IO error (1871)
4937 - class_export_put() on freed export after completion AST error (1896)
4938 - Fix revalidate looping in VFS (1322)
4939 - Don't access a freed export during MDS_REINT timeout (1521)
4940 - Add open-unlink recovery support on the MDS (1673,1764)
4941 - Return an error if no MDS data was read from last_rcvd (1946)
4942 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4943 - Don't setup a disk device that doesn't match exported UUID (317)
4944 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4945 - avoid committing NULL handle in force close
4946 - local.sh is now a one-stripe LOV configuration
4947 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4948 - fix ext3 htree duplicate directory entry corruption (1516)
4949 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4950 - update to new LOV EA format (2097)
4951 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4953 2003-06-15 Phil Schwan <phil@clusterfs.com>
4956 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4957 - per-import/export recovery handling (958, 931, 959)
4958 - multiple last-rcvd slots, for serving multiple FSes (949)
4959 - connections are again shared between multiple imp/exports (963, 964)
4960 - "umount -f" would hang if any requests needed to be sent (393, 978)
4961 - avoid pinning large req buffer by copying for queued messages (989)
4962 - add "uuid" to "lctl device" command to help upcalls (991)
4963 - "open" RPCs with transnos would confuse recovery counters (1037)
4964 - do proper endian conversion of all wire messages (288, 340, 891)
4965 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4966 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4967 - fix ptlrpc_connection leak in target_handle_connect (1174)
4968 - fix import refcounting bug in OST and MDS cleanup (1134)
4969 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4970 - fix very unlikely obd_types race condition (501)
4971 - remove osc_open hack for echo_client (1187)
4972 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4973 - a failure in read_inode2 leads to deadlock (1139)
4974 - cancel ack-locks as soon as transaction is committed (1072)
4975 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4976 - make sure to commitrw after any preprw to avoid deadlock (1162)
4977 - failing to execute a file in a lustre FS would lock inode (1203)
4978 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4979 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4980 - obd_uuid2tgt would walk off the end of the list (1255)
4981 - on IA64 the osc would give portals incorrect bulk size (1258)
4982 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4983 - fix lock inversion caused by new llite matching code (1282)
4984 - limit the number of dirty pages on a client to 10MB (1286)
4985 - timed out locks were not being corrected cancelled (1289)
4986 - fix O_DIRECT above 4GB on IA-32 (1292)
4987 * major user-visible changes
4988 - fail out/fail over policy now controlled by the upcall (993)
4990 - add OBD_PING to check server availability and failure (954)
4991 - lustre messages are now sent in sending host order (288, 340, 891)
4992 - add eadatalen to MDS getattr reply (340)
4993 - OST read replies may contain second buffer, with per-page status (593)
4995 2003-03-11 Phil Schwan <phil@clusterfs.com>
4998 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4999 - failed open()s could cause deadlock; fixed (867, 869)
5000 - stop cancelling OST locks when files are closed (481)
5001 - overlapping XID spaces caused network corruption (851, 853)
5002 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5003 - setattr_raw added, to do single-RPC, server-side setattrs
5004 - lmc/lconf syntax change for OST UUIDs
5005 - fix crashy race condition between ptlrpc_free_req and osc_close
5006 - don't use request in mdc_enqueue if we hit a timeout (889)
5007 - don't set the inode i_size for regular files from the MDS (896)
5008 - handle out of order completion AST (842)
5009 - don't LBUG if a lock request times out after receiving AST (913)
5010 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5011 - if a bad lock AST arrives, send an error instead of dropping entirely
5012 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5013 - fix leak in bulk IO when only partially completed (899, 900, 926)
5014 - fix O_DIRECT for ia64 (55)
5015 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5016 - C-z after timeout could hang a process forever; fixed (977)
5018 - client-side I/O cache (678, 924, 929, 941, 970)
5020 - READPAGE and SETATTRs which don't take server-side locks get
5023 2003-02-11 Phil Schwan <phil@clusterfs.com>
5026 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5027 - Fix fsfilt_extN_readpage() to read a full page of directory
5028 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5029 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5030 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5031 - Most intent operations were removed, in favour of a new RPC mode
5032 that does a single RPC to the server and bypasses most of the VFS
5033 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5034 - Aggressively cancel local locks on DLM servers
5035 - mds_reint_unlink sends EA to the client if it's the last nlink.
5036 client uses that EA to unlink OST objects.
5037 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5038 - recursive symlinks were fixed (439)
5039 - fixed NULL deref in DEBUG_REQ
5040 - filter_update_lastobjid no longer calls sync, which annoyed extN
5041 - fixed multi-client small-writes to a single file problem (445)
5042 - fixed mtime updates during file writes (607)
5043 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5044 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5045 - fixed Linux OST size reporting problem (444, 656)
5046 - OST now updates object mtime with writes or setattr (607, 619)
5047 - client verifies file size before zeroing page past EOF (445)
5048 - OST now writes last allocated objid to disk with allocation (108)
5049 - LOV on echo now works (409)
5051 - mds_reint_unlink sends a new buffer, with the EA included. this
5052 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5053 set if a regular file was being unlinked, and it was the last link
5054 - use PtlGet from the target for bulk writes (315)
5055 - OST now updates object mtime with writes or setattr (607, 619)
5056 - LDLM now has a grant-time callback to revalidate locked items, if
5058 - Many MDS operations were reorganized to combat race conditions
5060 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5061 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5063 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5066 - Fully reactivate OST imports after reconnection (512, others)
5067 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5068 - More graceful error handling for truncating on dead OST (515)
5069 - Don't error out unless we're actually accessing dead stripes (474)
5070 - Fix garbage sizes when stripes are missing (410)
5071 - LRU counters were broken, causing constant lock purge (433, 432)
5072 - garbage on read from stripes with failed OSTs (441)
5073 - mark OSCs as active before reconnecting during recovery (438)
5074 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5075 - lfind did not preserve OST order in output (443)
5076 - symlinks cause hung clients, incorrect data (439)
5077 - stop dereferencing request after dropping refcount (457)
5078 - don't LASSERT(spin_is_locked) on non-SMP (455)
5079 - fixes for many rename() bugs
5080 - fstat didn't correctly synchronize attributes (399)
5081 - server must handle lock cancellation during blocking AST prep (487)
5082 - bulk descriptors were free()d too soon (511)
5083 - fix paths in lconf, which would load incorrect modules (451, 507)
5084 - fix confusing lconf 'host not found' error message (386)
5085 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5086 - fix race condition in mdc_blocking_ast() for inode access (526)
5087 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5088 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5089 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5090 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5091 - fix lstripe in several ways (kernel side) (527)
5092 - fix request leak in ldlm_cli_enqueue (262)
5093 - incorrect OSC was marked inactive after OST failure
5094 - call mds_fs_cleanup before unmounting filesystem (524)
5095 - fix races between taking ns_lock and ldlm_lock_change_resource
5096 - fix races updating LOV export open file list
5097 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5098 - fix recovery NULL deref in ldlm_cli_cancel_unused
5099 - fix some DLM races by using new hash table for lock handles (419)
5100 - permit the client to specify desired inodes, at replay
5101 - duplicate requests when we queue them for replay reintegration
5102 - fix last_rcvd offset calculation
5103 - sync after each recovered transaction, so we always make progress
5104 - never, not always, ERESTART requests without transnos
5105 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5107 - skip replay if the MDS says that the client is already connected
5108 - don't check for a recovery-enabled export to match lctl's UUID
5109 - don't INC_USE_COUNT for phantom exports
5110 - don't crash when cleaning up phantom exports (567)
5111 - don't double-finish or set replay data for errored mdc_open requests
5112 - abort requests when they time out, so we don't get old replies
5113 - send/receive replies for AST messages again
5114 - if the client says that it doesn't have the lock, cancel it on the
5116 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5117 mark it as destroyed, it will all work out when decref is called
5118 - fix module use counts (22, 581)
5120 - ASTs now expect a reply (server cancels lock on error reply)
5122 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5125 - fix many simultaneous client startup (392)
5126 - fix dentry->d_it clobbering
5127 - credentials weren't being shipped for readdir/getattr operations
5128 - remove invalid assertions triggered during some concurrent MD
5130 - proper Lustre versions added (336, 389)
5131 - fix memory leak for create error case (398)
5132 - fix LOV locking bug that would get cli/srv out of sync
5133 - fix echo client over LOV (409)
5134 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5135 - fix double-O_EXCL intent crash (424)
5136 - avoid sending multiple lock CANCELs (352)
5138 - MDS can do multi-client recovery (modulo bugs in new code)
5140 - many updates, edits, cleanups
5142 2002-11-18 Phil Schwan <phil@clusterfs.com>
5145 - fix null d_it dereference (346)
5146 - fix full OST/dbench hang (333)
5147 - fix permission problem with file removal (286)
5148 - fix removal of OSCs from LOV when they fail
5149 - fix NULL deref during bulk timeout (214)
5150 - fix problems related to multiple filesystems on one MDS (241)
5151 - fixed serious subtle metadata locking bugs
5152 - free locks on clients when inodes are removed due to memory
5154 - fix inode pointer in lock data (285)
5155 - partial support for multiple MDS on a single host (241)
5156 - data locks weren't cancelled at clear_inode time (290, 311)
5157 - intent locks could lead to unbounded lock growth (205)
5158 - added a maximum lock count, an LRU list, and a flusher
5159 - fix multiple rename (365)
5160 - properly abstracted the echo client
5161 - OSC locked 1 byte too many; fixed
5162 - rewrote brw callback code:
5163 - fixed recovery bugs related to LOVs (306)
5164 - fixed too-many-pages-in-one-write crash (191)
5165 - fixed (again) crash in sync_io_timeout (214)
5166 - probably fixed callback-related race (385)
5168 - Add capability to MDS protocol
5169 - LDLM cancellations and callbacks on different portals
5171 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5174 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5175 - defer open object destruction to close time (601981, 138)
5176 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5177 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5178 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5179 - don't call obd_setattr in ll_file_release for destroyed objects
5180 * protocol change to lustre_msg: move |version| and add |flags|
5181 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5182 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5183 * added replay of create, unlink, link and rename operations during
5184 MDS failover; recovery should be much more robust now
5185 * remove failed OSCs from LOVs (only lov_create uses this so far)
5186 * the lustre-HOWTO was brought (more) up to date (582544)
5188 2002-10-23 Phil Schwan <phil@clusterfs.com>
5191 - in-use dentries weren't being reused properly (617851)
5192 - prevent multiple LDLM setup (599178)
5193 - fix LOV size calculations for truncate (617853)
5194 - fix client handling of MDS intent errors (POSIX)
5195 - fix permission bug in lovstripe.c test (624321)
5196 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5197 - truncate past end of file could corrupt data
5198 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5199 - a race in recovery could return ETIMEDOUT to apps (623947)
5200 - building outside the source directory was fixed
5201 * the lustre-HOWTO was brought (more) up to date (582544)
5202 * major progress was made on recovery functionality
5204 2002-10-10 Phil Schwan <phil@clusterfs.com>
5207 - recovery deadlock fix
5208 - rm -rf causes LBUG fix (617817)
5209 - file open by multiple tasks fix (618962)
5210 - directory permissions bugs (602707 and 620007)
5211 - journal_stop fixed with locking (611313)
5212 - O_APPEND failures resolved (618273, perhaps 614459)
5213 - lconf PATH fix (619770)
5214 - IA64 build fix (621450)
5215 - RPC buffer sizes scale with amount of memory
5217 2002-10-01 Phil Schwan <phil@clusterfs.com>
5220 - locks would be cancelled without throwing away data pages,
5221 resulting in inconsistent data (605627)
5222 - inode attributes were not always being refreshed (605627, 612449)
5223 - lconf now continues to cleanup after lctl reports an error
5224 - MDS now enforces user permissions (602707)
5225 - lprocfs cleanup fixed, but not yet enabled (614157)
5226 - fixed infinite server hang, should a client not respond to an AST
5227 - avoid going into recovery if user calls readlink() with a buffer
5228 that's too small (613941)
5229 - AST RPCs no longer require replies (614867) -- this may be changed
5230 - don't crash server if client sends an IOV that's too big (611336)
5231 - fixed lock conversion deadlock (611892)
5232 - fixed the following of symlinks (614622)
5233 * recovery: the server can remove locks from a client that dies, other
5234 clients can make progress
5235 * more extN patch fixes
5236 * compile-time configurable ptlrpc buffer allocations
5238 - collaborative read cache document
5239 - Lustre Lite Performance CDR document-in-progress
5241 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5244 - fix typo in patch-2.4.18
5246 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5249 - clear ptlrpc request each time in handle_incoming_request()
5250 - unlink of files now destroys the object on the OST
5252 2002-09-19 Peter Braam <braam@clusterfs.com>
5254 * add hard link support
5255 * change obdfile creation method
5256 * kernel patch changed
5258 2002-09-19 Peter Braam <braam@clusterfs.com>
5261 - stack overflow bug in extN fixed
5263 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5265 * documentation updates
5266 - add man pages for config tools
5267 - update tests/README to describe testing with new config tools
5268 - finish metadata API descriptions
5269 * bug fixes and cleanups
5270 - statfs workaround for 16TB limit
5271 - LOV stripe allocation improved, can stripe on subset of OSTs
5272 - LOV file size/IO offset was wrong for files > 4GB in size
5273 - object EA data was being dropped, caused files to be unreadable
5274 - memory overflow with non-LOV OST caused memory corruption
5275 - fixed regression tests to work with new config tools, obdfilter
5276 - fixed bug when directory size became larger than 1 block
5277 - fixed bug (for single client case) when PWD was deleted
5278 - invalidate local directory pages when doing intent-based ops
5279 - avoid LDLM oops when lock callback contained bad data
5281 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5283 * documentation updates
5284 * bug fixes and cleanups
5285 - configuration tools
5288 - 64-bit compile warnings
5289 - 64-bit internal statfs data
5291 * test_brw on persistent OST devices
5293 * lprocfs (disabled)
5295 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5297 * documentation updates
5298 * bug fixes and cleanups
5299 * configuration tools
5301 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5304 * many small fixes to 0_5_4
5305 * io/network handling
5306 * thinkos in MDS operations
5308 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5311 * crucial basic fixes to 0.5.3
5312 * IOR, Iozone work over Elan
5315 2002-08-07 Phil Schwan <phil@clusterfs.com>
5316 * version 0_5_3, our first alpha
5317 * we use the new Portals iovs
5318 * documentation updates
5319 * bug fixes and cleanups
5320 * small changes in the DLM wire protocol
5322 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5323 * version 0_5_1 with some initial stability,
5324 * locking on MD and file I/O.
5325 * documentation updates
5326 * several bug fixes since 0.5.0
5327 * small changes in wire protocol
5329 2002-07-18 Phil Schwan <phil@clusterfs.com>
5331 * delivered as Lustre Light Alpha
5332 * fixed a crash after handling invalid MDS requests
5333 * fixed directory pages for architectures with non-4k pages sizes
5335 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5336 * release version v0_4_4
5337 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5339 * Fixes a number of interruption problems with OST operations.
5340 * Update documentation for portals header changes
5341 * Move all wire protocol structs/defines to lustre_idl.h
5342 * Fixes symlink length bug.
5343 * Add tcpdump to repository.
5345 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5346 * release version v0_4_3
5347 * Fixes statfs for inodes on extN.
5348 * Fixes bug in runtests which would delete /etc/hosts.
5349 * Use 64-bit object IDs wherever possible (not into VFS though)
5350 Remove ost_get_info, which is unused by lustre, and out of date.
5352 2002-07-03 Peter Braam <braam@clusterfs.com>
5353 * release version v0_4_2 Fixes a lookup error (type not passed)
5354 * move forward to head of Portals
5355 * move forward to latest Lustre kernel
5357 2002-06-25 Peter Braam <braam@clusterfs.com>
5358 * release version v0_4_1. Hopefully stable on single node use.