1 tbd Sun Microsystems, Inc.
5 2.6.9-67.0.4.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 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
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).
23 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
24 __u64 as usingied long long type.
28 Frequency : rare, on shutdown
29 Description: race process ast vs remove callback
30 Details : removing callback before disconnect import open race with processing
35 Description: Add FMODE_EXEC to SLES10 SP1 series.
37 Severity : enhancement
39 Description: Add server support for vanilla-2.6.22.14.
43 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
44 Details : If an OST is down the MDS will hang indefinitely in
45 obd_statfs_rqset() waiting for the statfs data. While for
46 MDS QOS usage of statfs, it should not stuck in waiting.
50 Description: Disable adaptive timeouts by default
54 Frequency : on network error
55 Description: panic with double free request if network error
56 Details : mdc_finish_enqueue is finish request if any network error occuring,
57 but it's true only for synchronus enqueue, for async enqueue
58 (via ptlrpcd) this incorrect and ptlrpcd want finish request
63 Frequency : rare, on recovery
64 Description: read procfs can produce deadlock in some situation
65 Details : Holding lprocfs lock with send rpc can produce block for destroy
66 obd objects and this also block reconnect with -EALREADY. This isn't
67 fix all lprocfs bugs - but make it rare.
69 Severity : enhancement
71 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
74 Frequency : frequent on X2 node
76 Description: mdc_set_open_replay_data LBUG
77 Details : Set replay data for requests that are eligible for replay.
81 Description: lustre_mgs: operation 101 on unconnected MGS
82 Details : When MGC is disconnected from MGS long enough, MGS will evict the
83 MGC, and late on MGC cannot successfully connect to MGS and a lot
84 of the error messages complaining that MGS is not connected.
87 Frequency : rare, depends on device drivers and load
89 Description: MDS or OSS nodes crash due to stack overflow
90 Details : Code changes in 1.6.4 increased the stack usage of some functions.
91 In some cases, in conjunction with device drivers that use a lot
92 of stack, the MDS (or possibly OSS) service threads could overflow
93 the stack. One change which was identified to consume additional
94 stack has been reworked to avoid the extra stack usage.
96 Severity : enhancement
98 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
100 Severity : enhancement
102 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
104 Severity : enhancement
106 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
108 Serverity : enhancement
110 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
113 Frequency : rare on shutdown ost
115 Description: Don't allow skipping OSTs if index has been specified.
116 Details : Don't allow skipping OSTs if index has been specified, make locking
117 in internal create lots better.
121 Description: ASSERTION(!PageDirty(page)) failed
122 Details : Wrong check could lead to an assertion failure under specific
127 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
128 Details : access to bitfield in structure is always rounded to long
129 and this produce problem with not atomic change any bit.
133 Description: Lustre make rpms failed.
134 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
135 builds Lustre rpms from tarball.
137 Severity : enhancement
139 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
142 Frequency : rare on shutdown ost
144 Description: If llog cancel was not send before clean_exports phase, this can
145 produce deadlock in llog code.
146 Details : If llog thread has last reference to obd and call class_import_put
147 this produce deadlock because llog_cleanup_commit_master wait when
148 last llog_commit_thread exited, but this never success because was
149 called from llog_commit_thread.
153 Description: NULL lov_tgts causing MDS oops
154 Details : more safe checks for NULL lov_tgts for avoid oops.
156 Severity : enhancement
158 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
160 Severity : enhancement
162 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
166 Description: make mgs_setparam() handle fsname containing dash
167 Details : fsname containing a dash does not work with lctl conf_param
169 Severity : enhancement
171 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
175 Frequency : rare, in recovery and (or) destroy lovobjid file.
176 Description: rewrite lov objid code.
177 Details : Cleanup for lov objid code, remove scability problems and wrong
178 locking. Fix sending last_id into ost.
180 Severity : enhancement
182 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
184 Severity : enhancement
186 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
187 Details : Use d_move_locked instead of __d_move.
191 Frequency : rare, at shutdown
192 Description: access already free / zero obd_namespace.
193 Details : if client_disconnect_export was called without force flag set,
194 and exist connect request in flight, this can produce access to
195 NULL pointer (or already free pointer) when connect_interpret
196 store ocd flags in obd_namespace.
200 Frequency : only at startup
201 Description: not alloc memory with spinlock held.
202 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
203 if any spinlock held.
208 Description: lfs find does not continue on file error
209 Details : Continue other files processing when a file/dir is absent.
213 Description: Inconsistent usage of lustre_pack_reply()
214 Details : Standardize the usage of lustre_pack_reply() such that it
215 always generate a CERROR on failure.
218 Frequency : very rare
220 Description: Fix replay if there is an un-replied request and open
221 Details : In some cases, older replay request will revert the
222 mcd->mcd_last_xid on MDS which is used to record the client's
225 Severity : enhancement
227 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
229 Severity : enhancement
231 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
233 Severity : enhancement
235 Description: Update to SLES9 kernel-2.6.5-7.287.3.
237 Severity : enhancement
239 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
241 Severity : enhancement
243 Description: Build SLES10 patchless client fails
244 Details : The configure was broken by run ./configure with
245 --with-linux-obj=.... argument for patchless client. When the
246 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
247 can't search header adequately. Use absolute path such as
248 -I($LINUX)/include instead.
250 Severity : enhancement
252 Description: Lustre Page Accounting
253 Details : New macros for page alloc and free which enable accounting
254 of page allocation of Lustre. Use percpu counters to store memory
258 Frequency : only if debugging is disabled
260 Description: LASSERT_{REQ,REP}SWAB macros are buggy
261 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
262 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
263 these macros and replace them with their difinitions instead.
268 Description: interrupt oig_wait produce painc on resend.
269 Details : brw_redo_request can be used for resend requests from ptlrpcd and
270 private set, and this produce situation when rq_ptlrpcd_data not
271 copyed to new allocated request and triggered LBUG on assert
272 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
273 ptlrpcd set if request is changed and can be safety changed to use
276 Severity : enhancement
278 Description: organize the server-side client stats on per-nid basis
279 Details : Change the structure of stats under obdfilter and mds to
287 The "uuid"s file would list the uuids of _active_ exports.
288 And the clear entry is to clear all stats and stale nids.
293 Description: Processes looping in ll_readdir() on Lustre clients finally causing
294 a full node pseudo-hang
295 Details : Concurrent access to the same directory from multiple clients with
296 intensive file creation/removal can cause a client node to spin in
297 ll_readdir(). i_version must be increased every time the lock is
298 cancelled to ensure a revalidate is done.
303 Description: touch file failed when fs is not full
304 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
305 otherwise we can get ENOSP while fs is not full.
308 Frequency : only for Cray XT3
309 Bugzilla : 12829/13455
310 Description: Changing primary group doesn't change the group lustre assigns to
312 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
313 overridden with the primary group provided by the group upcall,
314 whereas the supplied fsgid can be trusted if it is in the list of
315 supplementary groups returned by the group upcall.
319 Description: Root Squash Functionality
320 Details : Implementation of NFS-like root squash capability. Specifically,
321 don't allow someone with root access on a client node to be able
322 to manipulate files owned by root on a server node.
326 Description: Slow trucate/writes to huge files at high offsets.
327 Details : Directly associate cached pages to lock that protect those pages,
328 this allows us to quickly find what pages to write and remove
329 once lock callback is received.
333 Description: Too many locks accumulating on client during NFS usage
334 Details : mds_open improperly used accmode to find out access mode to a
335 file. Also mdc_intent_lock logic to find out if we already have
336 lock similar to just received was flawed since introduction of
337 skiplists - locks are now added to the front of the granted
342 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
344 Details : In obd setup failure handler, obd_stopping will not necessarily to
345 be 1, and obd_set_up should also be checked to make sure whether
346 obd is completely setup.
348 Severity : enhancement
350 Description: Allow masking D_WARNING, D_ERROR messages from console
351 Details : Console messages can now be disabled via lnet.printk.
355 Description: User code with malformed file open parameter crashes client node
356 Details : Before packing join_file req, all the related reference should be
357 checked carefully in case some malformed flags cause fake join_file
363 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
364 grain of quota doing harm to quota's accuracy
365 Details : qunit size will be changed when quota limitation is too low/high;
366 record the pending quota write in order to get more accureate
367 quota; delete the patch for bug12588, which is unnecessary when
368 this patch is landed. This bug also contains fixes for bug 14526,
369 14299, 14601 and 13794.
373 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
374 Details : ldlm_completion_ast() assumes that a lock is granted when the req
375 mode is equal to the granted mode. However, it should also check
376 that LDLM_FL_CP_REQD is not set.
380 Description: Heavy nfs access might result in deadlocks
381 Details : After ELC code landed, it is now improper to enqueue any mds
382 locks under och_sem, because enqueue might want to decide to
383 cancel open locks for same inode we are holding och_sem for.
387 Description: 35% write performance drop with ldiskfs2 when quotas are on
388 Details : Enable ext3 journalled quota by default to improve performance
389 when quotas are turned on.
393 Description: Client eviction while running blogbench
394 Details : A lot of unlink operations with concurrent I/O can lead to a
395 deadlock causing evictions. To address the problem, the number of
396 oustanding OST_DESTROY requests is now throttled to
397 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
398 callbacks are processed in priority.
401 Frequency : RHEL4 only
403 Description: mkfs is very slow on IA64/RHEL4
404 Details : A performance regression has been discovered in the MPT Fusion
405 driver between versions 3.02.73rh and 3.02.99.00rh. As a
406 consequence, we have downgraded the MPT Fusion driver in the RHEL4
407 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
410 Frequency : PPC/PPC64 only
412 Description: conflicts between asm-ppc64/types.h and lustre_types.h
413 Details : fix duplicated definitions between asm-ppc64/types.h and
414 lustre_types.h on PPC.
417 Frequency : PPC/PPC64 only
419 Description: asm-ppc/segment.h does not exist
420 Details : fix compile issue on PPC.
424 Description: data checksumming impacts single node performance
425 Details : add support for several checksum algorithms. Currently, CRC32 and
426 Adler-32 are supported. The checksum type can be changed on the fly
427 through /proc/fs/lustre/osc/*/checksum_type.
431 Description: use adler32 for page checksums
432 Details : when available, use the Adler-32 algorithm instead of CRC32 for
437 Description: better handle error messages in extents code
439 Severity : enhancement
441 Description: SNMP support enhancement
442 Details : Adding total number of sampled request for an MDS node in snmp
445 Severity : enhancement
447 Description: Optimize ldlm waiting list processing for PR extent locks
448 Details : When processing waiting list for read extent lock and meeting read
449 lock that is same or wider to it that is not contended, skip
450 processing rest of the list and immediatelly return current
451 status of conflictness, since we are guaranteed there are no
452 conflicting locks in the rest of the list.
456 Description: Time out and refuse to reconnect
457 Details : When the failover node is the primary node, it is possible
458 to have two identical connections in imp_conn_list. We must
459 compare not conn's pointers but NIDs, otherwise we
460 can defeat connection throttling.
464 Description: Client not clear own cache if answer to reconnect is lost.
465 Details : Client gets evicted from server. Now client also thinks it is
466 disconnected (or gets ENOTCONN on its operation) and decides to
467 reconnect. Server receives reconnect message, but cannot find
468 export. New export is created that is fully valid (new cookie!),
469 but reply is lost and not reported to client. Client reconnects
470 again and gets back a just-created connection, but it is not new
471 so client thinks it was not evicted and does not do recovery.
475 Description: Detect stride IO mode in read-ahead
476 Details : When a client does stride read, read-ahead should detect that and
477 read-ahead pages according to the detected stride pattern.
481 Description: build for x2 fails
482 Details : fix compile issue on Cray systems.
484 Severity : enhancement
486 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
487 Details : This greatly improves speed of NFS writes on 2.6 kernels.
490 Frequency : only on PPC/SLES10
492 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
493 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
494 be defined. Add a hack in mkfs_lustre.c to work around this compile
499 Description: LASSERT on MDS when client holding flock lock dies
500 Details : ldlm pool logic depends on number of granted locks equal to
501 number of released locks which is not true for flock locks, so
502 just exclude such locks from consideration.
506 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
507 Details : Use fsfilt_sync() for both the whole filesystem sync and
508 individual file sync to eliminate dangerous inode locking
509 with I_LOCK that can lead to a deadlock.
513 Description: Update an obsolete wirecheck.c generator
514 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
516 --------------------------------------------------------------------------------
518 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
520 * Support for kernels:
521 2.6.5-7.286 (SLES 9),
522 2.6.9-55.0.9.EL (RHEL 4),
523 2.6.16.53-0.8 (SLES 10),
524 2.6.18-8.1.14.el5 (RHEL 5),
525 2.6.18.8 vanilla (kernel.org)
526 * Client support for unpatched kernels:
527 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
528 2.6.16 - 2.6.22 vanilla (kernel.org)
529 * Due to recently discovered recovery problems, we do not recommend
530 using patchless RHEL 4 clients with this or any earlier release.
531 * Recommended e2fsprogs version: 1.40.2-cfs1
532 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
535 Frequency : occasional
537 Description: MDS hang or stay in waiting lock
538 Details : If client receive lock with CBPENDING flag ldlm need send lock
539 cancel as separate rpc, to avoid situation when cancel request
540 can't processed due all i/o threads stay in wait lock.
543 Frequency : occasional
545 Description: Do not fail import if osc_interpret_create gets -EAGAIN
546 Details : If osc_interpret_create got -EAGAIN it immediately exits and
547 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
548 oscc_has_objects and see OSC has no objests and call
549 oscc_internal_create to resend create request.
551 Severity : enhancement
553 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
554 Details : Update which_patch & target file for SLES10 latest kernel.
556 Severity : enhancement
558 Description: add --type and --size parameters to lfs find
559 Details : Enhance lfs find by adding filetype and filesize parameters. Also
560 multiple OBDs can now be specified for the --obd option.
562 Severity : enhancement
564 Description: eliminate client locks in face of contention
565 Details : file contention detection and lockless i/o implementation
568 Severity : enhancement
570 Description: Remove client patches from SLES 10 kernel.
571 Details : This causes SLES 10 clients to behave as patchless clients
572 even on a Lustre-patched (server) kernel.
574 Severity : enhancement
576 Description: use i_size_read and i_size_write in 2.6 port
577 Details : replace inode->i_size access with i_size_read/write()
580 Frequency : when removing large files
582 Description: scheduling issue during removal of large Lustre files
583 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
584 It causes scheduling issues when removing large files (17TB in the
590 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
591 Details : Impossible (EPROTO is returned) to access a directory that has a
592 non-default striping and ACLs.
595 Frequency : only on ppc
597 Description: /proc/fs/lustre/devices broken on ppc
598 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
599 We should make sure the type of 'index' is loff_t and then cast
600 explicitly as needed below. Do not assign an explicitly cast
604 Frequency : only for rhel5
606 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
607 Details : Modify the target file & which_kernel.
610 Frequency : if the uninit_groups feature is enabled on ldiskfs
612 Description: e2fsck reports "invalid unused inodes count"
613 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
614 feature and only a single inode is created in a group then the
615 "bg_unused_inodes" count is incorrectly updated. Creating a
616 second inode in that group would update it correctly.
621 Description: buffer overruns could theoretically occur
622 Details : llapi_semantic_traverse() modifies the "path" argument by
623 appending values to the end of the origin string, and a buffer
624 overrun may occur. Adding buffer overrun check in liblustreapi.
626 Severity : enhancement
628 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
631 Frequency : only if filesystem is inconsistent
633 Description: handle "serious error: objid * already exists" more gracefully
634 Details : If LAST_ID value on disk is smaller than the objects existing in
635 the O/0/d* directories, it indicates disk corruption and causes an
636 LBUG(). If the object is 0-length, then we should use the existing
637 object. This will help to avoid a full fsck in most cases.
639 Severity : enhancement
641 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
642 Details : Modify vm-tunables-rhel4.patch.
644 Severity : enhancement
646 Description: Kernel config for 2.6.18-vanilla.
647 Details : Modify targets/2.6-vanilla.target.in.
648 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
649 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
650 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
651 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
654 Frequency : occasional
656 Description: improve handling recoverable errors
657 Details : If request processed with error which can be recoverable on server
658 request should be resend, otherwise page released from cache and
664 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
665 Details : Modify target file & which_patch.
666 A flaw was found in the IA32 system call emulation provided
667 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
668 value could be stored in the %RAX register, which could trigger an
669 out-of-bounds system call table access. An untrusted local user
670 could exploit this flaw to run code in the kernel
671 (ie a root privilege escalation). (CVE-2007-4573).
675 Description: change order of libsysio includes
676 Details : '#include sysio.h' should always come before '#include xtio.h'
678 Severity : enhancement
680 Description: adapt the lustre_config script to support the upgrade case
681 Details : Add "-u" option for lustre_config script to support upgrading 1.4
682 server targets to 1.6 in parallel.
687 Description: To avoid grant space > avaible space when the disk is almost
688 full. Without this patch you might see the error "grant XXXX >
689 available" or some LBUG about grant, when the disk is almost
691 Details : In filter_check_grant, for non_grant cache write, we should
692 check the left space by if (*left > ungranted + bytes), instead
693 of (*left > ungranted), because only we are sure the left space
694 is enough for another "bytes", then the ungrant space should be
695 increase. In client, we should update cl_avail_grant only there is
696 OBD_MD_FLGRANT in the reply.
701 Description: Update RHEL 4 kernel to fix local root privilege escalation.
702 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
703 described in CVE-2007-4573. This problem could allow untrusted
704 local users to gain root access.
707 Frequency : when using O_DIRECT and quotas
709 Description: Incorrect file ownership on O_DIRECT output files
710 Details : block usage reported by 'lfs quota' does not take into account
711 files that have been written with O_DIRECT.
713 --------------------------------------------------------------------------------
715 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
717 * Support for kernels:
718 2.6.5-7.286 (SLES 9),
719 2.6.9-55.0.2.EL (RHEL 4),
720 2.6.16.46-0.14 (SLES 10),
721 2.6.18-8.1.8.el5 (RHEL 5),
722 2.6.18.8 vanilla (kernel.org)
723 * Client support for unpatched kernels:
724 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
725 2.6.16 - 2.6.21 vanilla (kernel.org)
726 * Due to recently discovered recovery problems, we do not recommend
727 using patchless RHEL 4 clients with this or any earlier release.
728 * Recommended e2fsprogs version: 1.40.2-cfs1
729 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
733 Description: Fix errors in lfs documentation
734 Details : Fixes man pages
736 Severity : enhancement
738 Description: Adaptive timeouts
739 Details : RPC timeouts adapt to changing server load and network
740 conditions to reduce resend attempts and improve recovery time.
742 Severity : enhancement
744 Description: llapi_file_create() does not allow some changes
745 Details : add llapi_file_open() that allows specifying the file creation
746 mode and open flags, and also returns an open file handle.
748 Severity : enhancement
750 Description: df doesn't work properly if diskfs blocksize != 4K
751 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
754 Frequency : always on directories with default striping set
756 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
757 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
759 Severity : enhancement
761 Description: merge and cleanup kernel patches.
762 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
764 Severity : enhancement
766 Description: RedHat Update kernel for RHEL5
767 Details : Kernel config file for RHEL5.
769 Severity : enhancement
771 Description: OSS needs mutliple precreate threads
772 Details : Add ability to start more than one create thread per OSS.
775 Frequency : only with quota on the root user
777 Description: mds_obd_create error creating tmp object
778 Details : When the user sets quota on root, llog will be affected and can't
779 create files and write files.
782 Frequency : Always on ia64 patchless client, and possibly others.
784 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
785 Details : This allows the patchless client to be loaded on architectures
791 Description: disorder of journal start and llog_add cause deadlock.
792 Details : in llog_origin_connect, journal start should happen before
793 llog_add keep the same order as other functions to avoid
796 Severity : enhancement
798 Description: RedHat Update kernel for RHEL5
799 Details : Modify the kernel config file more closer RHEL5.
801 Severity : enhancement
803 Description: Build failure against Centos5 (RHEL5)
804 Details : Define PAGE_SIZE when it isn't present.
807 Frequency : occasionally when using NFS
809 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
810 Details : with NFS, the anon dentry's parent was set to itself in
811 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
812 corresponding dentry other than use rec->ur_name.
814 Severity : enhancement
816 Description: enable data checksumming by default
817 Details : enable checksum by default, allow --disable-checksum
818 configure option and "-o nochecksum" mount option. Checksums
819 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
822 Frequency : Occasionally with failover
824 Description: Client eviction due to failover config
825 Details : after a connection loss, the lustre client should attempt to
826 reconnect to the last active server first before trying the
827 other potential connections.
829 Severity : enhancement
831 Description: client-side metadata stat-ahead during readdir(directory readahead)
832 Details : perform client-side metadata stat-ahead when the client detects
833 readdir and sequential stat of dir entries therein
838 Description: when mds and osts use different quota unit(32bit and 64bit),
839 quota will be released repeatly.
840 Details : void sending multiple quota reqs to mds, which will keep the
841 status between the reqs.
844 Frequency : only with liblustre clients on XT3
846 Description: evictions taking too long
847 Details : allow llrd to evict clients directly on OSTs
851 Description: osts not allocated evenly to files
852 Details : change the condition to increase offset_idx
854 Severity : enhancement
856 Description: self-adjustable client's lru lists
857 Details : use adaptive algorithm for managing client cached locks lru
858 lists according to current server load, other client's work
859 pattern, memory activities, etc. Both, server and client
860 side namespaces provide number of proc tunables for controlling
865 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
866 Details : rewrite inherited code (uuid parsing code from ext2 utils and
867 readdir code from ext3) from scratch preserving functionality.
871 Description: Only those disconnect error should be returned by rq_status.
872 Details : In open/enqueue processs, Some errors, which will cause client
873 disconnected, should be returned by rq_status, while other
874 errors should still be returned by intent, then mdc or llite will
877 Severity : enhancement
879 Description: Tune the kernel for good SCSI performance.
880 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
881 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
885 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
886 Bugzilla : 13547, 13627
887 Description: Data corruption for OSTs that are formatted larger than 2TB
889 Details : When generating the bio request for lustre file writes the
890 sector number would overflow a temporary variable before being
891 used for the IO. The data reads correctly from Lustre (which
892 will overflow in a similar manner) but other file data or
893 filesystem metadata may be corrupted in some cases.
897 Description: "lfs find -obd UUID" prints directories
898 Details : "lfs find -obd UUID" will return all directory names instead
899 of just file names. It is incorrect because the directories
900 do not reside on the OSTs.
903 Bugzilla : 13596, 13614
904 Description: MDS hang after unclean shutdown of lots of clients
905 Details : Never resend AST requests.
909 Frequency : Always, for kernels after 2.6.16
910 Description: Fix warning idr_remove called for id=.. which is not allocated.
911 Details : Last kernels save old s_dev before kill super and not allow
912 to restore from callback - restore it before call kill_anon_super.
916 Description: Fix error on 'ls .' at the top of the Lustre mount.
917 Details : Don't revalidate dentry if it is a root dentry.
919 --------------------------------------------------------------------------------
921 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
923 * Support for kernels:
924 2.6.5-7.286 (SLES 9),
925 2.6.9-55.0.2.EL (RHEL 4),
926 2.6.16.46-0.14 (SLES 10),
927 2.6.18-8.1.8.el5 (RHEL 5),
928 2.6.18.8 vanilla (kernel.org)
929 * Client support for unpatched kernels:
930 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
931 2.6.16 - 2.6.22 vanilla (kernel.org)
932 * Due to recently discovered recovery problems, we do not recommend
933 using patchless RHEL 4 clients with this or any earlier release.
934 * Recommended e2fsprogs version: 1.39.cfs8
935 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
940 Description: block reactivating mgc import until all deactivates complete
941 Details : Fix race when failing back MDT/MGS to itself (testing)
943 Severity : enhancement
945 Description: lfs setstripe enhancement
946 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
949 Frequency : mds/oss recovery
951 Description: llog ctxt is refrenced after it has been freed.
952 Details : llog ctxt refcount was added to avoide the race between ctxt free
953 and llog recovery process. Each llog user must hold ctxt refcount
954 before it access the llog. And the llog ctxt can only be freed
955 when its refcount is zero.
957 Severity : enhancement
959 Description: randomly memory allocation failure util
960 Details : Make lustre randomly failed allocating memory for testing purpose.
963 Frequency : only for SLES10
965 Description: Update kernel patch for SLES10 SP1
966 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
968 Severity : enhancement
970 Description: omit set fsid for export NFS
971 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
972 as FS_REQUIRES_DEV for avoid problems with generate fsid.
975 Frequency : after network failures
977 Description: Add sync option to mount_lustre.c
978 Details : Client loses data written to lustre after a network interruption.
980 Severity : enhancement
982 Description: Error message improvement.
983 Details : Merging of two LCONSOLE_ERROR_MSG into one.
985 Severity : enhancement
987 Description: don't use GFP_* in generic Lustre code.
988 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
990 Severity : enhancement
992 Description: obdclass is limited by single OBD_ALLOC(idarray)
993 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
995 Severity : enhancement
997 Description: updated patchess for new RHEL4 kernel
998 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
1000 Severity : enhancement
1002 Description: warnings with build patchless client with vanila 2.6.19 and up
1003 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1004 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1006 Severity : enhancement
1008 Description: O_DIRECT bypasses client statistics.
1009 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1010 read_ahead_stats were updated. Copied stats section from
1011 osc_send_oap_rpc() into async_internal().
1014 Frequency : only for Cray XT3
1016 Description: peer credits not enough on many OST per OSS systems.
1017 Details : Use new lnet way to add credits as we need those for pings and ASTs
1020 Frequency : only with liblustre
1022 Description: Liblustre is not releasing flock locks on file close.
1023 Details : Release flock locks on file close.
1026 Frequency : only for RHEL4
1028 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
1029 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
1030 Update target file and kernel config.
1034 Description: lustre support for RHEL5
1035 Details : Add support for RHEL5.
1040 Description: ASSERTION(export != NULL) failed in target_handle_connect
1041 Details : Assetion hit is result of rare race between disconnect and connet
1042 to same nid. target_handle_connect found old connect cockie and
1043 tried to reconnect, but can't find export for this cockie.
1048 Description: umount blocks forever on error
1049 Details : In result of wrong using obd_no_recov and obd_force flags client
1050 can hand if cancel or some other requests is lost.
1053 Frequency : Only for SLES
1055 Description: sanity_quota fail test_1
1056 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1057 file, which makes TSTID[2] inunique.
1059 Severity : enhancement
1061 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1062 Details : Update target/ChangeLog/which_patch .
1064 Severity : enhancement
1066 Description: jbd statistics
1067 Details : Port older jbd statistics patch for sles10
1072 Description: lvbo_init failed for resource with missing objects.
1073 Details : Fix returning error if we do stat for file with missing/corrupted
1074 objects and i_size set to all sum of size of all avaible objects.
1075 if we to truncate/write to missing object - it is recreated.
1080 Description: Oops in read and write path when failing to allocate lock.
1081 Details : Check if lock allocation failed and return error back.
1084 Frequency : When flocks are used.
1086 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1087 Details : Flock locks might destroy just granted lock if it could be merged
1088 with another existing flock, this is done in completion handler,
1089 so teach ldlm_cli_enquque_fini that this is a valid case for
1095 Description: reply_lock_interpret crash due to race with it and lock cancel.
1096 Details : Do not replay locks that are being cancelled. Do not reference
1097 locks by their address during replay, just by their handle.
1100 Frequency : only with deactivated OSTs
1102 Description: lstripe command fails for valid OST index
1103 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
1104 instead of lov->desc.ld_active_tgt_count.
1106 Severity : enhancement
1108 Description: Build failure against Centos5 (RHEL5)
1109 Details : Use getpagesize() instead of PAGE_SIZE.
1111 --------------------------------------------------------------------------------
1113 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
1115 * Support for kernels:
1116 2.6.5-7.283 (SLES 9),
1117 2.6.9-55.EL (RHEL 4),
1118 2.6.16.46-0.14 (SLES 10),
1119 2.6.18.8 vanilla (kernel.org)
1120 * Client support for unpatched kernels:
1121 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1122 2.6.16 - 2.6.22 vanilla (kernel.org)
1123 * Due to recently discovered recovery problems, we do not recommend
1124 using patchless RHEL 4 clients with this or any earlier release.
1125 * Recommended e2fsprogs version: 1.39.cfs8
1126 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1127 * Starting with this release, the ldiskfs backing filesystem required
1128 by Lustre is now in its own package, lustre-ldiskfs. This package
1129 should be installed. It is versioned separately from Lustre and
1130 may be released separately in future.
1135 Description: server LBUG when shutting down
1136 Details : Block umount forever until the mount refcount is zero rather
1137 than giving up after an arbitrary timeout.
1139 Severity : enhancement
1141 Description: add optional extra BUILD_VERSION info
1142 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1143 to override the lustre version.
1146 Frequency : 2.6.18 servers only
1148 Description: ll_kern_mount() doesn't release the module reference
1149 Details : The ldiskfs module reference count never drops down to 0
1150 because ll_kern_mount() doesn't release the module reference.
1155 Description: server LBUG when using old ost_num_threads parameter
1156 Details : Accept the old ost_num_threads parameter but warn that it
1157 is deprecated, and fix an off-by-one error that caused an LBUG.
1162 Description: Transient SCSI error results in persistent IO issue
1163 Details : iobuf->dr_error is not reinitialized to 0 between two
1167 Frequency : sometimes when underlying device returns I/O errors
1169 Description: OSTs not going read-only during write failures
1170 Details : OSTs are not remounted read-only when the journal commit threads
1171 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1172 instead of the ext3 wrappers.
1175 Frequency : only with 10000 clients or more
1177 Description: poor connect scaling with increasing client count
1178 Details : Don't run filter_grant_sanity_check for more than 100 exports
1179 to improve scaling for large numbers of clients.
1182 Frequency : SLES10 only
1184 Description: sanity-quota.sh quotacheck failed: rc = -22
1185 Details : Quotas cannot be enabled on SLES10.
1188 Frequency : liblustre clients only
1190 Description: getdirentries does not give error when run on compute nodes
1191 Details : getdirentries does not fail when the size specified as an argument
1192 is too small to contain at least one entry
1194 Severity : enhancement
1196 Description: Add LNET router traceability for debug purposes
1197 Details : If a checksum failure occurs with a router as part of the
1198 IO path, the NID of the last router that forwarded the bulk data
1199 is printed so it can be identified.
1204 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1205 Details : Due to a race condition, liblustre clients were occasionally
1206 evicted incorrectly.
1208 Severity : enhancement
1210 Description: lfs setstripe use optional parameters instead of postional
1213 Severity : enhancement
1215 Description: Nanosecond timestamp support for ldiskfs
1216 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1217 resolution timestamps. There is not yet support for this at
1218 the Lustre filesystem level.
1221 Frequency : during server recovery
1223 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1224 Details : request with rq_no_resend flag not awake l_wait_event if they get a
1228 Frequency : nfs export on patchless client
1230 Description: connectathon hang when test nfs export over patchless client
1231 Details : Disconnected dentry cannot be found with lookup, so we do not need
1232 to unhash it or make it invalid
1235 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1236 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1237 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1242 Description: open req refcounting wrong on reconnect
1243 Details : If reconnect happened between getting open reply from server and
1244 call to mdc_set_replay_data in ll_file_open, we will schedule
1245 replay for unreferenced request that we are about to free.
1246 Subsequent close will crash in variety of ways.
1247 Check that request is still eligible for replay in
1248 mdc_set_replay_data().
1253 Description: disable writes to filesystem when reading health_check file
1254 Details : the default for reading the health_check proc file has changed
1255 to NOT do a journal transaction and write to disk, because this
1256 can cause reads of the /proc file to hang and block HA state
1257 checking on a healthy but otherwise heavily loaded system. It
1258 is possible to return to the previous behaviour during configure
1259 with --enable-health-write.
1261 Severity : enhancement
1263 Description: 64-bit inode version
1264 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1265 to the inode. This will be required for version-based recovery.
1270 Description: MDS fails to start if a duplicate client export is detected
1271 Details : in some rare cases it was possible for a client to connect to
1272 an MDS multiple times. Upon recovery the MDS would detect this
1273 and fail during startup. Handle this more gracefully.
1275 Severity : enhancement
1277 Description: Add -o localflock option to simulate old noflock behaviour.
1278 Details : This will achieve local-only flock/fcntl locks coherentness.
1283 Description: log_commit_thread vs filter_destroy race leads to crash
1284 Details : Take import reference before releasing llog record semaphore
1289 Description: Wrong request locking in request set processing
1290 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1291 imp_delayed_list, in this place should be used imp_lock.
1294 Frequency : when reconnecting
1296 Description: Grant leak when OSC reconnect to OST
1297 Details : When osc reconnect ost, OST(filter) should check whether it
1298 should grant more space to client by comparing fed_grant and
1299 cl_avail_grant, and return the granted space to client instead
1300 of "new granted" space, because client will call osc_init_grant
1301 to update the client grant space info.
1304 Frequency : when client reconnects to OST
1306 Description: Grant leak when OSC does a resend and replays bulk write
1307 Details : When osc reconnect to OST, OST(filter) should clear grant info of
1308 bulk write request, because the grant info while be sync between
1309 OSC and OST when reconnect, and we should ignore the grant info
1310 these of resend/replay write req.
1315 Description: Grant space more than avaiable space sometimes.
1316 Details : When then OST is about to be full, if two bulk writing from
1317 different clients came to OST. Accord the avaliable space of the
1318 OST, the first req should be permitted, and the second one
1319 should be denied by ENOSPC. But if the seconde arrived before
1320 the first one is commited. The OST might wrongly permit second
1321 writing, which will cause grant space > avaiable space.
1324 Frequency : when client is evicted
1326 Description: Grant might be wrongly erased when osc is evicted by OST
1327 Details : when the import is evicted by server, it will fork another
1328 thread ptlrpc_invalidate_import_thread to invalidate the
1329 import, where the grant will be set to 0. While the original
1330 thread will update the grant it got when connecting. So if
1331 the former happened latter, the grant will be wrongly errased
1332 because of this race.
1337 Description: Checking Stale with correct fid
1338 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1339 to check whether it is stale, because sometimes, we want the
1340 enqueue happened anyway, and op_data.fid2 will not be initialized.
1342 Severity : enhancement
1344 Description: update patchless client
1345 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1348 Frequency : only with 2.4 kernel
1350 Description: random memory corruption
1351 Details : size of struct ll_inode_info is to big for union inode.u and this
1352 can be cause of random memory corruption.
1357 Description: Memory leak in recovery
1358 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1359 It should also check obd_fail before fsfilt_start, otherwise if
1360 fsfilt_start return -EROFS,(failover mds during mds recovery).
1361 then the req will return with repmsg->transno = 0 and rc = EROFS.
1362 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1363 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1364 be freed no matter whether fsfilt_commit success or not.
1367 Frequency : only with huge numbers of clients
1369 Description: Prevents from taking the superblock lock in llap_from_page for
1371 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1372 for llap_from_page call in ll_removepage() prevents from taking
1373 the superblock lock for a soon died page.
1378 Description: Not check open intent error before release open handle
1379 Details : in some rare cases, the open intent error is not checked before
1380 release open handle, which may cause
1381 ASSERTION(open_req->rq_transno != 0), because it tries to release
1382 the failed open handle.
1387 Description: Set cat log bitmap only after create log success.
1388 Details : in some rare cases, the cat log bitmap is set too early. and it
1389 should be set only after create log success.
1393 Description: Accessing a block bevice can re-enable I/O when Lustre is
1394 tearing down a device.
1395 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1400 Description: service threads may hog cpus when there are a lot of requests
1401 Details : Insert cond_resched to give other threads a chance to use some CPU
1406 Description: the cat log was not initialized in recovery
1407 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1408 unlink log on mds will not be initialized until mds post
1409 recovery. And also in mds post recovery, the unlink log will
1410 initialization will be done asynchronausly, so there will be race
1411 between add unlink log and unlink log initialization.
1415 Description: brw_stats were being printed incorrectly
1416 Details : brw_stats were being printed as log2 but all of them were not
1417 recorded as log2. Also remove some code duplication arising from
1418 filter_tally_{read,write}.
1422 Frequency : rare, only in recovery.
1423 Description: ASSERTION(req->rq_type != LI_POISON) failed
1424 Details : imp_lock should be held while iterating over imp_sending_list for
1425 prevent destroy request after get timeout in ptlrpc_queue_wait.
1429 Description: replay-single.sh test 52 fails
1430 Details : A lock's skiplist need to be cleanup when it being unlinked
1431 from its resource list.
1436 Description: Short directio read returns full requested size rather than
1438 Details : Direct I/O operations should return actual amount of bytes
1439 transferred rather than requested size.
1441 Severity : enhancement
1443 Description: metadata RPC reduction (e.g. for rm performance)
1444 Details : decrease the amount of synchronous RPC between clients and servers
1445 by canceling conflicing lock before the operation on the client
1446 and packing thier handles into the main operation RPC to server.
1448 Severity : enhancement
1450 Description: add #ifdef HAVE_KERNEL_CONFIG_H
1451 Details : kernels from 2.6.19 not need include linux/config.h, but add
1452 include linux/autoconf.h in commpiler command line.
1454 Severity : enhancement
1456 Description: patchless client support for 2.6.22 kernel
1457 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
1458 removed. In this case we need drop using os depended interface to
1459 kmem_cache and use cfs_mem_cache API.
1464 Description: fix mal-formatted messages
1465 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
1469 Frequency : always in liblustre
1470 Description: wrong IS_ERR implementation in liblustre.h
1471 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
1476 Description: Correct condition for output debug message.
1477 Details : inode i_nlink equal zero is not enough for output message about
1478 disk corruption, i_ctime and i_mode should be also checked.
1482 Frequency : always in patchless client
1483 Description: add configure check for truncate_complete_page
1484 Details : improve checks for exported symbols. This allow run check without
1485 sources, but with Module.symvers shipped with kernel distribution.
1486 add check for truncate_complete_page used by patchless client.
1490 Description: sanity.sh test_77h fails with "test_77h file compare failed"
1491 Details : test_77h uses a file which was messed by other test case.
1495 Description: Not Check whether lov_tgts is NULL in some lov functions
1496 Details : Checking whether lov_tgts is NULL in some functions.
1500 Description: replace obdo_alloc() with OBDO_ALLOC macro
1501 Details : nothing special is done in obdo_alloc() function, and for
1502 debugging purpose, it needs to be replaced with macros.
1506 Description: bad return value and errno from fcntl call
1507 Details : In liblustre API, errno should be a negative value if error
1512 Description: ptlrpc_check_set() LBUG
1513 Details : In case of positive reply from server and failed client bulk
1514 callback after bulk transfer shouldn't LBUG, but process this
1515 request as erroneous.
1517 Severity : enhancement
1519 Description: VFS operations stats tool.
1520 Details : Tool which collects stats by tracking value written in pid,
1521 ppid, gid and uses llstat to generate output to plot graph using
1523 Updated lustre/utils/Makefile.am
1524 Added lustre/utils/ltrack_stats.c
1526 Severity : enhancement
1528 Description: 2.6.18 server support (lustre 1.6.1)
1529 Details : Support for 2.6.18 kernels on the server side.
1534 Description: ASSERTION(imp->imp_conn_current) failed
1535 Details : an assertion failure is hit if a client node boots and attempts to
1536 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
1539 Frequency : only for i686
1541 Description: 1.4.11 RC1 build fails for RHEL 4, i686
1542 Details : Fixed config variable for build.
1547 Description: Updated patchess for new RHEL4 kernel
1548 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
1549 Updated patch jbd-stats-2.6.9.patch
1550 Updated patch qsnet-rhel4-2.6.patch
1551 Updated patch quota-deadlock-on-pagelock-core.patch
1552 Updated patch vfs_intent-2.6-rhel4.patch
1553 Updated patch vfs_races-2.6-rhel4.patch
1554 Updated series file 2.6-rhel4-titech.series
1555 Updated series file 2.6-rhel4.series
1556 Updated kernel config files
1561 Description: lquota slave complains LBUG when reconnecting with mds
1563 Details : quota slave depends qctxt->lqc_import to send its quota request.
1564 This pointer will be invalid if mds did failover or broke its
1565 connect to osts, which leads to LBUG.
1568 Frequency : when qunit size is too small(less than 20M)
1570 Description: write is stopped by improper -EDQUOT
1571 Details : If the master is busy and qunit size is small enough(let's say 1M),
1572 the slave can not get quota from master on time, which will lead
1573 slave to trigger a -EQUOTA to client.
1578 Description: Deadlock during metadata tests
1579 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
1580 called with the per-dentry lock held.
1583 Frequency : SLES9 only
1585 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
1586 Details : set CONFIG_NR_CPUS to 128 instead of 8.
1588 Severity : enhancement
1590 Description: remove fs_prep_san_write operation and related patches
1591 Details : remove the ext3-san-jdike patches which are no longer useful.
1596 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
1597 Details : a disk corruption can cause the mballoc code to assert on a
1598 double free or other extent corruptions. Handle these with
1599 ext3_error() instead of assertions.
1602 Frequency : only with mballoc3 code and deep extent trees
1604 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
1605 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
1606 can cause the filesystem to be remounted read-only.
1611 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
1612 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
1613 few external users of this ioctl number is lfsck's call to
1614 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
1615 Add the old ioctl number to the handler so both old and new
1620 Description: parallel lock callbacks
1621 Details : Instead of sending blocking and completion callbacks as separated
1622 requests, adding them to a set and sending in parallel.
1626 Description: Disable most debugging by default
1627 Details : To improve performance, disable most logging (for debug purposes)
1628 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
1629 default, and HA includes fewer messages.
1633 Frequency : only run on patchless client.
1634 Description: use do_facet on sanity.sh for test handling recoverables errors
1635 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
1639 Description: hash tables for lists of nids, connections and uuids
1640 Details : Hash tables noticeably help when a lot of clients connect to a
1641 server, to faster identify duplicate connections or reconnects,
1642 also to faster find export to evict in manual eviction case.
1646 Description: Sometimes, when the server evict a client, and the client will
1647 not be evicted as soon as possible.
1648 Details : In enqueue req, the error was returned by intent, instead of
1649 rq_status which make ptlrpc layer not detect this error, and
1650 does not evict the client. So enqueue error should be returned
1654 Frequency : only at startup
1656 Description: Delay client connections to MDT until fist MDT->OST connect
1657 Details : If a client tried to create a new file before the MDT had
1658 connected to any OSTs, the create would return EIO. Now
1659 the client will simply block until the MDT connects to the
1660 first OST and the create can succeed.
1663 Frequency : only for SLES9
1665 Description: Routinely utilize latest Quadrics drivers in CFS releases
1666 Details : Update patch qsnet-suse-2.6.patch.
1669 Frequency : only for sles10
1671 Description: Update patches for SLES 10 SP1 kernel.
1672 Details : Update the patch vfs_intent-2.6-sles10.patch.
1677 Description: Routinely utilize latest Quadrics drivers in CFS releases
1678 Details : Update patch qsnet-rhel4-2.6.patch.
1681 Frequency : at statup only
1683 Description: mds_lov_synchronize race leads to various problems
1684 Details : simultaneous MDT->OST connections at startup can cause the
1685 sync to abort, leaving the OSC in a bad state.
1690 Description: Using wrong pointer in osc_brw_prep_request
1691 Details : Access to array[-1] can produce panic if kernel compiled with
1692 CONFIG_PAGE_ALLOC enabled
1694 Severity : enhancement
1696 Description: Async OSC create to avoid the blocking unnecessarily.
1697 Details : If a OST has no remain object, system will block on the creating
1698 when need to create a new object on this OST. Now, ways use
1699 pre-created objects when available, instead of blocking on an
1700 empty osc while others are not empty. If we must block, we block
1701 for the shortest possible period of time.
1705 Frequency : only in recovery
1706 Description: Mark OST as early accessible if his start SYNC.
1707 Details : osc_precreate return flag early accessible if oscc marked as
1708 OSCC_FLAG_SYNC_IN_PROGRESS.
1713 Description: Sometimes precreate code can triger create object on wrong ost
1714 Details : Wrong protected or not not restored variables aftre precreate loop
1715 can produce creation object on wrong ost.
1718 Frequency : oss recovery
1720 Description: llog_commit_thread cleanup should sync with llog_commit_thread
1722 Details : llog_commit_thread_count should be synced between llog_commit
1723 start and cleanup, so new llog_commit thread should not be started
1724 when llog_commit threads being stopped to avoid accessing some
1727 Severity : enhancement
1729 Description: Add printing inode info into message about error in writepage.
1731 --------------------------------------------------------------------------------
1733 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1738 Frequency : on some architectures
1740 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1741 Details : Uninitialized flags sometimes cause configuration commands to
1745 Frequency : patchless clients only
1747 Description: missing __iget() symbol export
1748 Details : The __iget() symbol export is missing. To avoid the need for
1749 this on patchless clients the deathrow inode reaper is turned
1750 off, and we depend on the VM to clean up old inodes. This
1751 dependency was during via the fix for bug 12181.
1756 Description: sanity.sh fail: test_52b
1757 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
1758 incorrect inode's flags to client.
1760 --------------------------------------------------------------------------------
1762 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1764 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1765 INTEROPERATE with older versions automatically. In many cases a
1766 special upgrade step is needed. Please read the
1767 user documentation before upgrading any part of a 1.4.x system.
1768 * WARNING: Lustre configuration and startup changes are required with
1769 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1771 * Support for kernels:
1772 2.4.21-47.0.1.EL (RHEL 3),
1773 2.6.5-7.283 (SLES 9),
1774 2.6.9-42.0.10.EL (RHEL 4),
1775 2.6.12.6 vanilla (kernel.org),
1776 2.6.16.27-0.9 (SLES10)
1777 * Client support for unpatched kernels:
1778 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
1779 2.6.16 - 2.6.19 vanilla (kernel.org),
1780 2.6.9-42.0.8EL (RHEL 4)
1781 * Recommended e2fsprogs version: 1.39.cfs6
1782 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1785 Severity : enhancement
1787 Description: Async OSC create to avoid the blocking unnecessarily.
1788 Details : If a OST has no remain object, system will block on the creating
1789 when need to create a new object on this OST. Now, ways use
1790 pre-created objects when available, instead of blocking on an
1791 empty osc while others are not empty. If we must block, we block
1792 for the shortest possible period of time.
1794 Severity : enhancement
1796 Description: MountConf
1797 Details : Lustre configuration is now managed via mkfs and mount
1798 commands instead of lmc and lconf. New obd types (MGS, MGC)
1799 are added for dynamic configuration management. See
1800 https://mail.clusterfs.com/wikis/lustre/MountConf for
1803 Severity : enhancement
1805 Description: dynamic OST addition
1806 Details : OSTs can now be added to a live filesystem
1808 Severity : enhancement
1810 Description: startup order invariance
1811 Details : MDTs and OSTs can be started in any order. Clients only
1812 require the MDT to complete startup.
1814 Severity : enhancement
1816 Description: parallel, asynchronous orphan cleanup
1817 Details : orphan cleanup is now performed in separate threads for each
1818 OST, allowing parallel non-blocking operation.
1820 Severity : enhancement
1822 Description: optimized stripe assignment
1823 Details : stripe assignments are now made based on ost space available,
1824 ost previous usage, and OSS previous usage, in order to try
1825 to optimize storage space and networking resources.
1827 Severity : enhancement
1829 Description: Permanently set tunables
1830 Details : All writable /proc/fs/lustre tunables can now be permanently
1831 set on a per-server basis, at mkfs time or on a live
1834 Severity : enhancement
1836 Description: Lustre message v2
1837 Details : Add lustre message format v2.
1839 Severity : enhancement
1841 Description: client OST exclusion list
1842 Details : Clients can be started with a list of OSTs that should be
1843 declared "inactive" for known non-responsive OSTs.
1847 Description: ENOENT returned for valid filehandle during dbench.
1848 Details : Check if a directory has children when invalidating dentries
1849 associated with an inode during lock cancellation. This fixes
1850 an incorrect ENOENT sometimes seen for valid filehandles during
1851 testing with dbench.
1854 Frequency : SFS test only (otherwise harmless)
1856 Description: SPEC SFS validation failure on NFS v2 over lustre.
1857 Details : Changes the blocksize for regular files to be 2x RPC size,
1858 and not depend on stripe size.
1860 Severity : enhancement
1862 Description: fine-grained SMP locking inside DLM
1863 Details : Improve DLM performance on SMP systems by removing the single
1864 per-namespace lock and replace it with per-resource locks.
1866 Severity : enhancement
1868 Description: don't hold multiple extent locks at one time
1869 Details : To avoid client eviction during large writes, locks are not
1870 held on multiple stripes at one time or for very large writes.
1871 Otherwise, clients can block waiting for a lock on a failed OST
1872 while holding locks on other OSTs and be evicted.
1874 Severity : enhancement
1876 Description: Multiple MD RPCs in flight.
1877 Details : Further unserialise some read-only MDT RPCs - learn about intents.
1878 To avoid overly-overloading MDT, introduce a limit on number of
1879 MDT RPCs in flight for a single client and add /proc controls
1880 to adjust this limit.
1882 Severity : enhancement
1884 Description: client read/write statistics
1885 Details : Add client read/write call usage stats for performance
1886 analysis of user processes.
1887 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1888 file access. extents_stats shows chunk size distribution.
1889 extents_stats_per_process show chunk size distribution per
1892 Severity : enhancement
1894 Description: per-client statistics on server
1895 Details : Add ldlm and operations statistics for each client in
1896 /proc/fs/lustre/mds|obdfilter/*/exports/
1898 Severity : enhancement
1900 Description: improved MDT statistics
1901 Details : Add detailed MDT operations statistics in
1902 /proc/fs/lustre/mds/*/stats
1904 Severity : enhancement
1906 Description: VFS operations stats
1907 Details : Add client VFS call stats, trackable by pid, ppid, or gid
1908 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
1913 Description: Fix client-side osc byte counters
1914 Details : The osc read/write byte counters in
1915 /proc/fs/lustre/osc/*/stats are now working
1918 Frequency : always as root on SLES
1920 Description: Failure of copying files with lustre special EAs.
1921 Details : Client side always return success for setxattr call for lustre
1922 special xattr (currently only "trusted.lov").
1927 Description: Refcount LNET uuids
1928 Details : The global LNET uuid list grew linearly with every startup;
1929 refcount repeated list entries instead of always adding to
1932 Severity : enhancement
1934 Description: Dynamic service threads
1935 Details : Within a small range, start extra service threads
1936 automatically when the request queue builds up.
1939 Frequency : mixed-endian client/server environments
1941 Description: mixed-endian crashes
1942 Details : The new msg_v2 system had some failures in mixed-endian
1945 Severity : enhancement
1947 Description: Easy OST removal
1948 Details : OSTs can be permanently deactivated with e.g. 'lctl
1949 conf_param lustre-OST0001.osc.active=0'
1951 Severity : enhancement
1953 Description: MGS proc entries
1954 Details : Added basic proc entries for the MGS showing what filesystems
1957 Severity : enhancement
1959 Description: provide MGS failover
1960 Details : Added config lock reacquisition after MGS server failover.
1962 Severity : enhancement
1964 Description: add Linux 2.4 support
1965 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1969 Description: a large application tries to do I/O to the same resource and dies
1970 in the middle of it.
1971 Details : Check the req->rq_arrival time after the call to
1972 ost_brw_lock_get(), but before we do anything about
1973 processing it & sending the BULK transfer request. This
1974 should help move old stale pending locks off the queue as
1975 quickly as obd_timeout.
1978 Frequency : when an incorrect nid is specified during startup
1980 Description: ptlrpc connect to non-existant node causes kernel crash
1981 Details : LNET can't be re-entered from an event callback, which
1982 happened when we expire a message after the export has been
1983 cleaned up. Instead, hand the zombie cleanup off to another
1986 Severity : enhancement
1988 Description: plain/inodebits lock performance improvement
1989 Details : Grouping plain/inodebits in granted list by their request modes
1990 and bits policy, thus improving the performance of search through
1994 Frequency : only if OST filesystem is corrupted
1996 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1997 Details : for a short time RPCs with bulk IO are in the replay list,
1998 but replay of bulk IOs is unimplemented. If the OST filesystem
1999 is corrupted due to disk cache incoherency and then replay is
2000 started it is possible to trip an assertion. Avoid putting
2001 committed RPCs into the replay list at all to avoid this issue.
2004 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2006 Description: System hang on startup
2007 Details : This bug allowed the liblustre (e.g. catamount) client to
2008 return to the app before handling all startup RPCs. This
2009 could leave the node unresponsive to lustre network traffic
2010 and manifested as a server ptllnd timeout.
2012 Severity : enhancement
2014 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2015 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2016 boolean to control whether to print peer debug info when a
2017 client's RPC times out.
2020 Frequency : only for kernels with patches from Lustre below 1.4.3
2022 Description: Remove old rdonly API
2023 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2026 Frequency : only for devices with external journals
2028 Description: Set external device read-only also
2029 Details : During a commanded failover stop, we set the disk device
2030 read-only while the server shuts down. We now also set any
2031 external journal device read-only at the same time.
2034 Frequency : when upgrading from 1.4 while trying to change parameters
2036 Description: The wrong (new) MDC name was used when setting parameters for
2037 upgraded MDT's. Also allows changing of OSC (and MDC)
2038 parameters if --writeconf is specified at tunefs upgrade time.
2041 Frequency : when setting specific ost indicies
2043 Description: QOS code breaks on skipped indicies
2044 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2045 created with --index need not be sequential.
2047 Severity : enhancement
2049 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2050 Details : The uninit_groups feature works in conjunction with the kernel
2051 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2052 up the pass1 processing of e2fsck. This is a read-only feature
2053 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2054 mount filesystems that have had this feature enabled.
2056 Severity : enhancement
2058 Description: Improve multi-block allocation algorithm to avoid fragmentation
2059 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2060 allocation locality and avoid filesystem fragmentation.
2065 Description: lfs quota fails with deactivated OSTS
2066 Details : With this patch, three improvements are included:
2067 1. detete the softlimit in mds and osts when use "lfs quota".
2068 2. display the inaccurate data in the output of "lfs quota".
2069 3. try to get quota info when "lfs quota" is executed.
2071 ------------------------------------------------------------------------------
2073 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2075 * Support for kernels:
2076 2.4.21-47.0.1.EL (RHEL 3)
2077 2.6.5-7.283 (SLES 9)
2078 2.6.9-42.0.10.EL (RHEL 4)
2079 2.6.12.6 vanilla (kernel.org)
2080 2.6.16.27-0.9 (SLES 10)
2081 * Recommended e2fsprogs version: 1.39.cfs5
2083 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2087 Frequency : occasional, depends on client load and configuration
2088 Bugzilla : 12181, 12203
2089 Description: data loss for recently-modified files
2091 Details : In some cases it is possible that recently written or created
2092 files may not be written to disk in a timely manner (this should
2093 normally be within 30s unless client IO load is very high).
2094 The problem appears as zero-length files or files that are a
2095 multiple of 1MB in size after a client crash or client eviction
2096 that are missing data at the end of the file.
2098 This problem is more likely to be hit on clients where files are
2099 repeatedly created and unlinked in the same directory, clients
2100 have a large amount of RAM, have many CPUs, the filesystem has
2101 many OSTs, the clients are rebooted frequently, and/or the files
2102 are not accessed by other nodes after being written.
2104 The presence of the problem can be detected by looking at
2105 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2106 smaller than the second (nr_unused) then dirty files will not
2107 be flushed automatically to disk. "sync; sleep 10" should be
2108 run several times on the node before unmounting it to update
2109 Lustre (this is also safe to run on nodes without this problem).
2111 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2112 that can cause this same problem, so customers using that kernel
2113 also need to update the kernel in addition to Lustre. In order
2114 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2116 It is normal that files written just before a client crash (less
2117 than 30s) may not yet have been flushed to disk, even for local
2121 Frequency : frequent on thin XT3 nodes
2123 Description: UUID collision on thin XT3 Linux nodes
2124 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2125 randomly, since we relied on an insufficiently-seeded PRNG.
2130 Description: OSS hangs after "All ost request buffers busy"
2131 Details : A deadlock between quota and journal operations caused OSS
2132 hangs after printing "All ost request buffers busy."
2135 Frequency : always on liblustre builds
2137 Description: Cleanup compiler warnings on liblustre
2140 Frequency : always on liblustre builds on XT3
2142 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2146 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2147 > 32000 subdirectories
2148 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2149 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2150 subdirectory count crosses 32000. This will aid e2fsck to
2151 correctly handle more than 32000 subdirectories.
2154 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2156 Description: System hang on startup
2157 Details : This bug allowed the liblustre (e.g. catamount) client to
2158 return to the app before handling all startup RPCs. This
2159 could leave the node unresponsive to lustre network traffic
2160 and manifested as a server ptllnd timeout.
2162 Severity : enhancement
2164 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2165 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2166 boolean to control whether to print peer debug info when a
2167 client's RPC times out.
2172 Description: make O_SYNC working on 2.6 kernels
2173 Details : 2.6 kernels use different method for mark pages for write,
2174 so need add a code to lustre for O_SYNC work.
2179 Description: Failure to close file and release space on NFS
2180 Details : Put inode details into lock acquired in ll_intent_file_open.
2181 Use mdc_intent_lock in ll_intent_open to properly
2182 detect all kind of errors unhandled by mdc_enqueue.
2187 Description: proc file read during shutdown sometimes raced obd removal,
2189 Details : Add lock to prevent obd access after proc file removal.
2192 Frequency : Only for files larger than 4GB on 32-bit clients.
2194 Description: improperly doing page alignment of locks
2195 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2196 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2201 Description: RPCs being resent when they shouldn't be
2202 Details : Some RPCs that should not be resent are being resent. This
2203 can cause inconsistencies in the RPC state machine. Do not
2204 resend such requests.
2207 Frequency : rare, only with NFS export
2209 Description: Crash on NFS re-export node
2210 Details : under very unusual load conditions an assertion is hit in
2211 ll_intent_file_open()
2214 Frequency : only if OST filesystem is corrupted
2216 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2217 Details : for a short time RPCs with bulk IO are in the replay list,
2218 but replay of bulk IOs is unimplemented. If the OST filesystem
2219 is corrupted due to disk cache incoherency and then replay is
2220 started it is possible to trip an assertion. Avoid putting
2221 committed RPCs into the replay list at all to avoid this issue.
2226 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2227 Details : Large single O_DIRECT read and write calls can fail to allocate
2228 a sufficiently large buffer to process the request. In case of
2229 allocation failure the allocation is retried with a smaller
2230 buffer and broken into smaller requests.
2232 Severity : enhancement
2234 Description: Add -o localflock option to simulate old noflock behaviour.
2235 Details : This will achieve local-only flock/fcntl locks coherentness.
2240 Description: versioning check is incomplete
2241 Details : Checking the version difference of client vs. server, report
2242 error if the gap is too big.
2247 Description: add support PG_writeback bit
2248 Details : add support for PG_writeback bit for Lustre, for more carefull
2249 work with page cache in 2.6 kernel. This also fix some deadlocks
2250 and remove hack for work O_SYNC with 2.6 kernel.
2252 Severity : enhancement
2254 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2255 Details : The uninit_groups feature works in conjunction with the kernel
2256 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2257 up the pass1 processing of e2fsck. This is a read-only feature
2258 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2259 mount filesystems that have had this feature enabled.
2261 Severity : enhancement
2263 Description: Improve multi-block allocation algorithm to avoid fragmentation
2264 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2265 allocation locality and avoid filesystem fragmentation.
2267 ------------------------------------------------------------------------------
2269 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2271 * Support for kernels:
2272 2.6.9-42.0.3.EL (RHEL 4)
2273 2.6.5-7.276 (SLES 9)
2274 2.4.21-47.0.1.EL (RHEL 3)
2275 2.6.12.6 vanilla (kernel.org)
2276 2.6.16.21-0.8 (SLES10)
2277 * Recommended e2fsprogs version: 1.39.cfs2-0
2279 * The backwards-compatible /proc/sys/portals symlink has been removed
2280 in this release. Before upgrading, please ensure that you change
2281 any configuration scripts or /etc/sysctl.conf files that access
2282 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2283 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2284 in advance of the upgrade on any system running Lustre 1.4.6 or
2285 newer, since /proc/sys/lnet was added in that version.
2286 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2290 Frequency : only when quota is used
2292 Description: avoid scanning export list for quota master
2293 Details : Change the algorithms to avoid scanning export list in order
2294 to improve the efficiency.
2297 Frequency : MDS failover only, very rarely
2299 Description: "went back in time" messages on mds failover
2300 Details : The greatest transno may be lost when the current operation
2301 finishes with an error (transno==0) and the client's last_rcvd
2302 record is over-written. Save the greatest transno in the
2303 mds_last_transno for this case.
2306 Frequency : always for specific kernels and striping counts
2308 Description: client may get "Matching packet too big" without ACL support
2309 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2310 when trying to access files in certain configurations. The
2311 clients should in fact be denied when mounting because they do
2312 not understand ACLs.
2315 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2317 Description: many clients connecting with IO in progress causes connect timeouts
2318 Details : Avoid synchronous journal commits to avoid delays caused by many
2319 clients connecting/disconnecting when bulk IO is in progress.
2320 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2321 OST_IO_PORTAL to avoid delays behind potentially many pending
2325 Frequency : occasionally with multiple writers to a single file
2327 Description: shared writes to file may result in wrong size reported by stat()
2328 Details : Allow growing of kms when extent lock is cancelled
2331 Frequency : always with random mmap IO to multi-striped file
2333 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2334 Details : Only if the hole is at the end of OST object so that kms is too
2335 small. Fix is to increase kms accordingly in ll_nopage.
2338 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2340 Description: writes to a missing object would leak memory on the OST
2341 Details : If there is an inconsistency between the MDS and OST filesystems,
2342 such that the MDS references an object that doesn't exist, writes
2343 to that object will leak memory due to incorrect cleanup in the
2344 error handling path, eventually running out of memory on the OST.
2349 Description: Creating too long symlink causes lustre errors
2350 Details : Check symlink and name lengths before sending requests to MDS.
2353 Frequency : only if flock is enabled (not on by default)
2355 Description: posix locks not released on fd closure on 2.6.9+
2356 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2357 kernels, this caused such locks not to be released on fd close and
2358 then assertions on fs unmount about still used locks.
2361 Frequency : MDS failover only, very rarely
2363 Description: clients may get ASSERTION(granted_lock != NULL)
2364 Details : When request was taking a long time, and a client was resending
2365 a getattr by name lock request. The were multiple lock requests
2366 with the same client lock handle and
2367 mds_getattr_name->fixup_handle_for_resent_request found one of the
2368 lock handles but later failed with ASSERTION(granted_lock != NULL).
2373 Description: handle->h_buffer_credits > 0, assertion failure
2374 Details : h_buffer_credits is zero after truncate, causing assertion
2375 failure. This patch extends the transaction or creates a new
2379 Frequency : NFS re-export or patchless client
2380 Bugzilla : 11179, 10796
2381 Description: Crash on NFS re-export node (__d_move)
2382 Details : We do not want to hash the dentry if we don't have a lock.
2383 But if this dentry is later used in d_move, we'd hit uninitialised
2384 list head d_hash, so we just do this to init d_hash field but
2385 leave dentry unhashed.
2388 Frequency : NFS re-export or patchless client
2390 Description: NFS exports has problem with symbolic link
2391 Details : lustre client didn't properly install dentry when re-exported
2392 to NFS or running patchless client.
2395 Frequency : NFS re-export or patchless client
2397 Description: Various nfs/patchless fixes.
2398 Details : fixes reuse disconected alias for lookup process - this fixes
2399 warning "find_exported_dentry: npd != pd",
2400 fix permission error with open files at nfs.
2401 fix apply umask when do revalidate.
2404 Frequency : occasional
2406 Description: Crash on NFS re-export node
2407 Details : calling clear_page() on the wrong pointer triggered oops in
2408 generic_mapping_read().
2411 Frequency : rarely, using O_DIRECT IO
2413 Description: unaligned directio crashes client with LASSERT
2414 Details : check for unaligned buffers before trying any requests.
2417 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2419 Description: stale data returned from RAID cache
2420 Details : If only a small amount of IO is done to the RAID device before
2421 reading it again it is possible to get stale data from the RAID
2422 cache instead of reading it from disk.
2425 Frequency : always for sles10 kernel
2427 Description: sles10 support
2428 Details : ll_follow_link: compile fixes and using of nd_set_link
2429 under newer kernels.
2432 Frequency : depends on arch, kernel and compiler version, always on sles10
2435 Description: recursive or deep enough symlinks cause stack overflow
2436 Details : getting rid of large stack-allocated variable in
2440 Frequency : depends on hardware
2442 Description: lustre write performance loss in the SLES10 kernel
2443 Details : the performance loss is caused by using of write barriers in the
2444 ext3 code. The SLES10 kernel turns barrier support on by
2445 default. The fix is to undo that change for ldiskfs.
2447 ------------------------------------------------------------------------------
2449 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2451 * Support for kernels:
2452 2.6.9-42.0.3EL (RHEL 4)
2453 2.6.5-7.276 (SLES 9)
2454 2.4.21-47.0.1.EL (RHEL 3)
2455 2.6.12.6 vanilla (kernel.org)
2459 Frequency : quota enabled and large files being deleted
2461 Description: releasing more than 4GB of quota at once hangs OST
2462 Details : If a user deletes more than 4GB of files on a single OST it
2463 will cause the OST to spin in an infinite loop. Release
2464 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2469 Description: statfs data retrieved from /proc may be stale or zero
2470 Details : When reading per-device statfs data from /proc, in the
2471 {kbytes,files}_{total,free,avail} files, it may appear
2472 as zero or be out of date.
2475 Frequency : systems with MD RAID1 external journal devices
2477 Description: lconf's call to blkid is confused by RAID1 journal devices
2478 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2479 of returning all block devices that match the journal UUID.
2482 Frequency : always, for aggregate stripe size over 4GB
2484 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2485 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2486 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2489 Frequency : always if "lfs find" used on a local file/directory
2491 Description: "lfs find" segfaults if used on a local file/directory
2492 Details : The case where a directory component was not specified wasn't
2493 handled correctly. Handle this properly.
2496 Frequency : always on ppc64
2498 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2499 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2500 perform endianness conversion.
2503 Frequency : rarely (truncate to non-zero file size after write under load)
2504 Bugzilla : 10730, 10687
2505 Description: Files padded with zeros to next 4K multiple
2506 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2507 it is possible that files that are truncated to a non-zero size
2508 immediately after being written are filled with zero bytes beyond
2509 the truncated size. No file data is lost.
2511 Severity : enhancement
2513 Description: Allow recovery/failover for liblustre clients.
2514 Details : liblustre clients were unaware of failover configurations until
2517 Severity : enhancement
2519 Description: user file locks should fail when not mounting with flock option
2520 Details : Set up an error-returning stub in ll_file_operations.lock field
2521 to prevent incorrect behaviour when client is mounted without
2522 flock option. Also, set up properly f_op->flock field for
2526 Frequency : always on ia64
2528 Description: "lfs df" loops on printing out MDS statfs information
2529 Details : The obd_ioctl_data was not initialized and in some systems
2530 this caused a failure during the ioctl that did not return
2531 an error. Initialize the struct and return an error on failure.
2534 Frequency : SLES 9 only
2536 Description: Error of copying files with lustre special EAs as root
2537 Details : Client side always return success for setxattr call for lustre
2538 special xattr (currently only "trusted.lov").
2541 Frequency : rarely on clusters with both ia64+i386 clients
2543 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2544 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2545 clients are doing concurrent writes to the same file it is
2546 possible that the ia64 clients may LASSERT because the OST
2547 extent locks are not PAGE_SIZE aligned. Ensure that grown
2548 locks are always aligned on the request boundary.
2551 Frequency : specific use, occasional
2553 Description: Overwriting in use executable truncates on-disk binary image
2554 Details : If one node attempts to overwrite an executable in use by
2555 another node, we now correctly return ETXTBSY instead of
2556 truncating the file.
2558 Severity : enhancement
2560 Description: Async OSC create to avoid the blocking unnecessarily.
2561 Details : If a OST has no remain object, system will block on the creating
2562 when need to create a new object on this OST. Now, ways use
2563 pre-created objects when available, instead of blocking on an
2564 empty osc while others are not empty. If we must block, we block
2565 for the shortest possible period of time.
2570 Description: chmod on Lustre root is propagated to other clients
2571 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2572 invalid by the follow_mount time.
2575 Frequency : liblustre clients only
2577 Description: Race in 'instant cancel' lock handling could lead to such locks
2578 never to be granted in case of SMP MDS
2579 Details : Do not destroy not yet granted but cbpending locks in
2583 Frequency : replay/resend of open
2585 Description: non null lock assetion failure in mds_intent_policy
2586 Details : Trying to replay/resend lockless open requests resulted in
2587 mds_open() returning 0 with no lock. Now it sets a flag if
2588 a lock is going to be returned.
2590 Severity : enhancement
2592 Description: Checksum enhancements
2593 Details : New checksum enhancements allow for resending RPCs that failed
2596 Severity : enhancement
2598 Description: Tunables on number of dirty pages in cacche
2599 Details : Allow to set limit on number of dirty pages cached.
2604 Description: client crash on unmount - lock still has references
2605 Details : In some error handling cases it was possible to leak a lock
2606 reference on a client while accessing a file. This was not
2607 harmful to the client during operation, but would cause the
2608 client to crash when the filesystem is unmounted.
2611 Frequency : specific case, rare
2613 Description: ETXTBSY on mds though file not in use by client
2614 Details : ETXTBSY is no longer incorrectly returned when attempting to
2615 chmod or chown a directory that the user previously tried to
2616 execute or a currently-executing binary.
2619 Frequency : extremely rare except on liblustre-based clients
2621 Description: Lustre space not freed when files are deleted
2622 Details : Clean up open-unlinked files after client eviction. Previously
2623 the unlink was skipped and the files remained as orphans.
2628 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2629 Details : In some cases it was possible to send a blocking callback to
2630 a client doing a glimpse, even though that client didn't get
2631 a lock granted. When the glimpse lock is cancelled on the OST
2632 the freed lock is left on the waiting list and corrupted the list.
2635 Frequency : all core dumps
2637 Description: Broke core dumps to lustre
2638 Details : Negative dentry may be unhashed if parent does not have UPDATE
2639 lock, but some callers, e.g. do_coredump, expect dentry to be
2640 hashed after successful create, hash it in ll_create_it.
2642 ------------------------------------------------------------------------------
2644 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2646 * Support for kernels:
2647 2.6.9-42.0.2.EL (RHEL 4)
2648 2.6.5-7.276 (SLES 9)
2649 2.4.21-40.EL (RHEL 3)
2650 2.6.12.6 vanilla (kernel.org)
2654 Frequency : always on RHEL 3
2656 Description: Number of open files grows over time
2657 Details : The number of open files grows over time, whether or not
2658 Lustre is started. This was due to a filp leak introduced
2659 by one of our kernel patches.
2661 ------------------------------------------------------------------------------
2663 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
2665 * Support for kernels:
2666 2.6.9-42.EL (RHEL 4)
2667 2.6.5-7.267 (SLES 9)
2668 2.4.21-40.EL (RHEL 3)
2669 2.6.12.6 vanilla (kernel.org)
2674 Bugzilla : 5719, 9635, 9792, 9684
2675 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2676 Details : If a server is under heavy load and cannot reply to new
2677 connection requests before the client resends the (re)connect,
2678 the connection handling code can behave badly if two service
2679 threads are concurrently handing separate (re)connections from
2680 the same client. Add better locking to the connection handling
2681 code, and ensure that only a single connection will be processed
2682 for a given client UUID, even if the lock is dropped.
2684 Severity : enhancement
2686 Description: add TCP zero-copy support to kernel
2687 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2688 sends if the hardware supports scatter-gather and checksumming.
2689 This allows socklnd to do client-write and server-read more
2690 efficiently and reduce CPU utilization from skbuf copying.
2693 Frequency : only if NFS exporting from client
2695 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2696 Details : If setting attributes on a file created under NFS that had
2697 never been opened it would be possible to oops the client
2698 if the file had no objects.
2702 Bugzilla : 9326, 10402, 10897
2703 Description: client crash in ptlrpcd_wake() thread when sending async RPC
2704 Details : It is possible that ptlrpcd_wake() dereferences a freed async
2705 RPC. In rare cases the ptlrpcd thread alread processed the RPC
2706 before ptlrpcd_wake() was called and the request was freed.
2709 Frequency : always for liblustre
2711 Description: liblustre client does MDS+OSTs setattr RPC for each write
2712 Details : When doing a write from a liblustre client, the client
2713 incorrectly issued an RPC to the MDS and each OST the file was
2714 striped over in order to update the timestamps. When writing
2715 with small chunks and many clients this could overwhelm the MDS
2716 with RPCs. In all cases it would slow down the write because
2717 these RPCs are unnecessary.
2719 Severity : enhancement
2721 Description: allow number of MDS service threads to be changed at module load
2722 Details : It is now possible to change the number of MDS service threads
2723 running. Adding "options mds mds_num_threads={N}" to the MDS's
2724 /etc/modprobe.conf will set the number of threads for the next
2725 time Lustre is restarted (assuming the "mds" module is also
2726 reloaded at that time). The default number of threads will
2727 stay the same, 32 for most systems.
2732 Description: OST crash if filesystem is unformatted or corrupt
2733 Details : If an OST is started on a device that has never been formatted
2734 or if the filesystem is corrupt and cannot even mount then the
2735 error handling cleanup routines would dereference a NULL pointer.
2740 Description: NULL pointer deref in llap_from_page.
2741 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2742 incorrect) mapping if the page was truncated/reclaimed while it was
2743 searched for. Check for this condition and skip such pages when
2744 doing readahead. Introduce extra check to llap_from_page() to
2745 verify page->mapping->host is non-NULL (so page is not anonymous).
2748 Frequency : Sometimes when using sys_sendfile
2750 Description: "page not covered by a lock" warnings from ll_readpage
2751 Details : sendfile called ll_readpage without right page locks present.
2752 Now we introduced ll_file_sendfile that does necessary locking
2753 around call to generic_file_sendfile() much like we do in
2757 Frequency : with certain MDS communication failures at client mount time
2759 Description: NULL pointer deref after failed client mount
2760 Details : a client connection request may delayed by the network layer
2761 and not be sent until after the PTLRPC layer has timed out the
2762 request. If the client fails the mount immediately it will try
2763 to clean up before the network times out the request. Add a
2764 reference from the request import to the obd device and delay
2765 the cleanup until the network drops the request.
2768 Frequency : occasionally during client (re)connect
2770 Description: assertion failure during client (re)connect
2771 Details : processing a client connection request may be delayed by the
2772 client or server longer than the client connect timeout. This
2773 causes the client to resend the connection request. If the
2774 original connection request is replied in this interval, the
2775 client may trip an assertion failure in ptlrpc_connect_interpret()
2776 which thought it would be the only running connect process.
2779 Frequency : only with obd_echo servers and clients that are rebooted
2781 Description: kernel BUG accessing uninitialized data structure
2782 Details : When running an obd_echo server it did not start the ping_evictor
2783 thread, and when a client was evicted an uninitialized data
2784 structure was accessed. Start the ping_evictor in the RPC
2785 service startup instead of the OBD startup.
2787 Severity : enhancement
2788 Bugzilla : 10193 (patchless)
2789 Description: Remove dependency on various unexported kernel interfaces.
2790 Details : No longer need reparent_to_init, exit_mm, exit_files,
2791 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2794 Frequency : rare (only users of deprecated and unsupported LDAP config)
2796 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2797 Details : LDAP apparently contains 'lustreName' attributes instead of
2798 'name'. A simple remapping of the name is sufficient.
2801 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2803 Description: waiting_locks_callback trips kernel BUG if client is evicted
2804 Details : Running with the dump_on_timeout debug flag turned on makes
2805 it possible that the waiting_locks_callback() can try to dump
2806 the Lustre kernel debug logs from an interrupt handler. Defer
2807 this log dumping to the expired_lock_main() thread.
2809 Severity : enhancement
2811 Description: Support NFS exporting on 2.6 kernels.
2812 Details : Implement non-rawops metadata methods for NFS server to use without
2813 changing NFS server code.
2816 Frequency : very rare (synthetic metadata workload only)
2818 Description: two racing renames might cause an MDS thread to deadlock
2819 Details : Running the "racer" program may cause one MDS thread to rename
2820 a file from being the source of a rename to being the target of
2821 a rename at exactly the same time that another thread is doing
2822 so, and the second thread has already enqueued these locks after
2823 doing a lookup of the target and is trying to relock them in
2824 order. Ensure that we don't try to re-lock the same resource.
2827 Frequency : only very large systems with liblustre clients
2829 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2830 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2831 all OSTs in parallel. This allows the eviction of stale liblustre
2832 clients to proceed much faster than if they were done in series,
2833 and also offers similar improvements for other set_info RPCs.
2838 Description: excessive CPU usage during initial read phase on client
2839 Details : During the initial read phase on a client, it would agressively
2840 retry readahead on the file, consuming too much CPU and impacting
2841 performance (since 1.4.5.8). Improve the readahead algorithm
2842 to avoid this, and also improve some other common cases (read
2843 of small files in particular, where "small" is files smaller than
2844 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2849 Description: MDS crash when receiving packet with unknown intent.
2850 Details : Do not LBUG in unknown intent case, just return -EFAULT
2852 Severity : enhancement
2853 Bugzilla : 9293, 9385
2854 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2855 Details : Do not serialize getattr (non-intent version) and statfs.
2858 Frequency : occasional, when OST network is overloaded/intermittent
2860 Description: client evicted by OST after bulk IO timeout
2861 Details : If a client sends a bulk IO request (read or write) the OST
2862 may evict the client if it is unresposive to its data GET/PUT
2863 request. This is incorrect if the network is overloaded (takes
2864 too long to transfer the RPC data) or dropped the OST GET/PUT
2865 request. There is no need to evict the client at all, since
2866 the pinger and/or lock callbacks will handle this, and the
2867 client can restart the bulk request.
2870 Frequency : Always when mmapping file with no objects
2872 Description: client crashes when mmapping file with no objects
2873 Details : Check that we actually have objects in a file before doing any
2874 operations on objects in ll_vm_open, ll_vm_close and
2880 Description: Request leak when working with deleted CWD
2881 Details : Introduce advanced request refcount tracking for requests
2882 referenced from lustre intent.
2884 Severity : Enhancement
2886 Description: Cache open file handles on client.
2887 Details : MDS now will return special lock along with openhandle, if
2888 requested and client is allowed to hold openhandle, even if unused,
2889 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2890 closing files for every read/write openration.
2892 Severity : Enhancement
2894 Description: Cache open negative dentries on client when possible.
2895 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2896 negative dentries on lock revocation.
2901 Description: Remounting a client read-only wasn't possible with a zconf mount
2902 Details : It wasn't possible to remount a client read-only with llmount.
2904 Severity : enhancement
2905 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2906 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2907 MPICH is included to add Lustre-specific ADIO interfaces.
2908 This is based closely on the UFS ADIO layer and only differs
2909 in file creation, in order to allow the OST striping to be set.
2910 This is user-contributed code and not supported by CFS.
2915 Description: extended inode attributes (immutable, append-only) work improperly
2916 when 2.4 and 2.6 kernels are used on client/server or vice versa
2917 Details : Introduce kernel-independent values for these flags.
2919 Severity : enhancement
2922 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2923 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2924 granularity. Now, Fractional MB granularity is be supported,
2925 this is very useful for low memory system.
2927 Severity : enhancement
2929 Description: Getattr by fid
2930 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2931 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2932 have locks and avoids getattr rpc requests for it on every stat.
2935 Frequency : Always, for filesystems larger than 2TB
2937 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2938 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2939 filesystems larger than 2TB to be mounted with mballoc enabled.
2942 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2944 Description: filesystem corruption for non-standard kernels and very large OSTs
2945 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2946 filesystems larger than 2TB is mounted then the kernel will
2947 silently corrupt the start of the filesystem. CONFIG_LBD is
2948 enabled for all CFS-supported kernels, but the possibility of
2949 this happening with a modified kernel config exists.
2951 Severity : enhancement
2953 Description: add client O_DIRECT support for 2.6 kernels
2954 Details : It is now possible to do O_DIRECT reads and writes to files
2955 in the Lustre client mountpoint on 2.6 kernel clients.
2957 Severity : enhancement
2959 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2960 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2961 parallel, not waiting for response from every OST before sending
2962 a rpc to the next OST.
2967 Description: setattr vs write race when updating file timestamps
2968 Details : Client processes that update a file timestamp into the past
2969 right after writing to the file (e.g. tar) it is possible that
2970 the updated file modification time can be reset to the current
2971 time due to a race between processing the setattr and write RPC.
2973 Severity : enhancement
2975 Description: Bring 'lfs find' closer in line with regular Linux find.
2976 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2977 -print0 options and obtains all the needed info through the lustre
2980 Severity : enhancement
2982 Description: support up to 1024 configured devices on one node
2983 Details : change obd_dev array from statically allocated to dynamically
2984 allocated structs as they are first used to reduce memory usage
2989 Description: Flush dirty partially truncated pages during truncate
2990 Details : Immediatelly flush partially truncated pages in filter_setattr,
2991 this way we completely avoid having any pages in page cache on OST
2992 and can retire ugly workarounds during writes to flush such pages.
2997 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2998 Details : protect inode from truncation within vfs_unlink() context
2999 just take a reference before calling vfs_unlink() and release it
3000 when parent's i_sem is free.
3003 Frequency : always, if extents are used on OSTs
3005 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3006 Details : OSTs using the extents format would not zero the high 16 bits of
3007 the index physical block number. This is not a problem for any
3008 OST filesystems smaller than 16TB, and no kernels support ext3
3009 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3010 new/modified files) and can be fixed for existing filesystems
3011 with e2fsprogs-1.39-cfs1.
3016 Description: import connection selection may be incorrect if timer wraps
3017 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3018 import connections to be ignored if the 32-bit jiffies counter
3019 wraps. Use a 64-bit jiffies counter.
3022 Frequency : during server recovery
3024 Description: crash after server is denying duplicate export
3025 Details : If clients are resending connect requests to the server, the
3026 server refuses to allow a client to connect multiple times.
3027 Fixed a bug in the handling of this case.
3030 Frequency : very large clusters immediately after boot
3032 Description: LNET request buffers exhausted under heavy short-term load
3033 Details : If a large number of client requests are generated on a service
3034 that has previously never seen so many requests it is possible
3035 that the request buffer growth cannot keep up with the spike in
3036 demand. Instead of dropping incoming requests, they are held in
3037 the LND until the RPC service can accept more requests.
3040 Frequency : Sometimes during replay
3042 Description: Assertion failure in ll_local_open after replay.
3043 Details : If replay happened on an open request reply before we were able
3044 to set replay handler, reply will become not swabbed tripping the
3045 assertion in ll_local_open. Now we set the handler right after
3046 recognising of open request
3049 Frequency : very rare
3051 Description: kernel reports "badness in vsnprintf"
3052 Details : Reading from the "recovery_status" /proc file in small chunks
3053 may cause a negative length in lprocfs_obd_rd_recovery_status()
3054 call to vsnprintf() (which is otherwise harmless). Exit early
3055 if there is no more space in the output buffer.
3057 Severity : enhancement
3059 Description: clear OBD RPC statistics by writing to them
3060 Details : It is now possible to clear the OBD RPC statistics by writing
3061 to the "stats" file.
3066 Description: Client mtime is not the same on different clients after utimes
3067 Details : In some cases, the client was using the utimes() syscall on
3068 a file cached on another node. The clients now validate the
3069 ctime from the MDS + OSTs to determine which one is right.
3074 Description: Inability to activate failout mode
3075 Details : lconf script incorrectly assumed that in python string's numeric
3076 value is used in comparisons.
3079 Frequency : always with multiple stripes per file
3081 Description: Inefficient object allocation for mutli-stripe files
3082 Details : When selecting which OSTs to stripe files over, for files with
3083 a stripe count that divides evenly into the number of OSTs,
3084 the MDS is always picking the same starting OST for each file.
3085 Return the OST selection heuristic to the original design.
3090 Description: mount failures may take full timeout to return an error
3091 Details : Under some heavy load conditions it is possible that a
3092 failed mount can wait for the full obd_timeout interval,
3093 possibly several minutes, before reporting an error.
3094 Instead return an error as soon as the status is known.
3096 ------------------------------------------------------------------------------
3098 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
3100 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3101 INTEROPERATE with older versions automatically. Please read the
3102 user documentation before upgrading any part of a live system.
3103 * WARNING: Lustre networking configuration changes are required with
3104 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3107 * Support for kernels:
3108 2.6.9-22.0.2.EL (RHEL 4)
3109 2.6.5-7.244 (SLES 9)
3110 2.6.12.6 vanilla (kernel.org)
3113 Severity : enhancement
3114 Bugzilla : 7981/8208
3115 Description: Introduced Lustre Networking (LNET)
3116 Details : LNET is new networking infrastructure for Lustre, it includes
3117 a reorganized network configuration mode (see the user
3118 documentation for full details) as well as support for routing
3119 between different network fabrics. Lustre Networking Devices
3120 (LNDs) for the supported network fabrics have also been
3121 created for this new infrastructure.
3123 Severity : enhancement
3124 Description: Introduced Access control lists
3125 Details : clients can set ACLs on files and directories in order to have
3126 more fine-grained permissions than the standard Unix UGO+RWX.
3127 The MDS must be started with the "-o acl" mount option.
3129 Severity : enhancement
3130 Description: Introduced filesystem quotas
3131 Details : Administrators may now establish per-user quotas on the
3134 Severity : enhancement
3136 Description: Configuration change for the XT3
3137 The PTLLND is now used to run Lustre over Portals on the XT3
3138 The configure option(s) --with-cray-portals are no longer used.
3139 Rather --with-portals=<path-to-portals-includes> is used to
3140 enable building on the XT3. In addition to enable XT3 specific
3141 features the option --enable-cray-xt3 must be used.
3146 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3147 Details : A race between allocating a new llcd and re-getting the llcd_lock
3148 allowed another thread to grab newly-allocated llcd.
3150 Severity : enhancement
3152 Description: 2.6 OST async journal commit and locking fix to improve performance
3153 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3154 2.6 kernels are now async as they already were in 2.4 kernels so
3155 that they can commit concurrently with the network bulk transfer.
3156 For block-allocated files the filter allocation semaphore is held
3157 to avoid filesystem fragmentation during allocation. BKL lock
3158 removed for 2.6 xattr operations where it is no longer needed.
3163 Description: lconf incorrectly determined whether two IP networks could talk
3164 Details : In some more complicated routing and multiple-network
3165 configurations, lconf will avoid trying to make a network
3166 connection to a disjoint part of the IP space. It was doing the
3167 math incorrectly for one set of cases.
3172 Description: Fix for potential infinite loop processing records in an llog.
3173 Details : If an llog record is corrupted/zeroed, it is possible to loop
3174 forever in llog_process(). Validate the llog record length
3175 and skip the remainder of the block on error.
3178 Frequency : occasional (liblustre only)
3180 Description: liblustre could not open files whose last component is a symlink
3181 Details : sysio_path_walk() would incorrectly pass the open intent to
3182 intermediate path components.
3185 Frequency : rare (liblustre only with non-standard tuning)
3186 Bugzilla : 7201 (7350)
3187 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3188 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3189 after it has been released, only do this for referenced locks
3191 Severity : enhancement
3193 Description: specifying an (invalid) directory default stripe_size of -1
3194 would reset the directory default striping
3195 Details : stripe_size -1 was used internally to signal directory stripe
3196 removal, now use "all default" to signal dir stripe removal
3197 as a directory striping of "all default" is not useful
3200 Frequency : common for large clusters running liblustre clients
3202 Description: doing an ls when liblustre clients are running is slow
3203 Details : sending a glimpse AST to a liblustre client waits for every AST
3204 to time out, as liblustre clients will not respond. Since they
3205 cannot cache data we refresh the OST lock LVB from disk instead.
3207 Severity : enhancement
3209 Description: doing an ls at the same time as file IO can be slow
3210 Details : enqueue and other "small" requests can be blocked behind many
3211 large IO requests. Create a new OST IO portal for non-IO
3212 requests so they can be processed faster.
3215 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3217 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3218 created file with mode 000
3219 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3220 allow an ftruncate() as a non-root user to a file with mode 000.
3221 Lustre now respects this flag to disable mode checks when
3222 truncating a file owned by the user
3225 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3228 Description: Revoking locks from clients that went dead or catatonic might take
3230 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3231 cancellation of such locks instant on servers without waiting for
3232 any reply from clients. Clients drops these locks when cancel
3233 notification from server is received without replying.
3236 Frequency : liblustre-only, when liblustre client dies or becomes busy
3238 Description: Doing ls on Linux clients can take a long time with active
3240 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3241 granting such locks to it in the first place if possible. Locks
3242 are taken by proxy on the OST during the read or write and
3243 dropped immediately afterward. Add connect flags handling, do
3244 not grant locks to liblustre clients for glimpse ASTs.
3246 Severity : enhancement
3248 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3249 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3250 sequential reads and behaves badly with applications doing random
3251 reads. Improve it by reading ahead at least read region, and
3252 avoiding excessive large RPC for small reads.
3254 Severity : enhancement
3256 Description: Creating more than 1000 files for a single job may cause a load
3257 imbalance on the OSTs if there are also a large number of OSTs.
3258 Details : qos_prep_create() uses an OST index reseed value that is an
3259 even multiple of the number of available OSTs so that if the
3260 reseed happens in the middle of the object allocation it will
3261 still utilize the OSTs as uniformly as possible.
3266 Description: OST or MDS may oops in ping_evictor_main()
3267 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3268 but doesn't restart at beginning of obd_exports_timed list
3271 Severity : enhancement
3273 Description: improve by-nid export eviction on the MDS and OST
3274 Details : allow multiple exports with the same NID to be evicted at one
3275 time without re-searching the exports list.
3278 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3280 Description: MDS may oops in groups_free()
3281 Details : in rare race conditions a newly allocated group_info struct is
3282 freed again, and this can be NULL. The 2.4 compatibility code
3283 for groups_free() checked for a NULL pointer, but 2.6 did not.
3286 Frequency : common for liblustre clients doing little filesystem IO
3287 Bugzilla : 9352, 7313
3288 Description: server may evict liblustre clients accessing contended locks
3289 Details : if a client is granted a lock or receives a completion AST
3290 with a blocking AST already set it would not reply to the AST
3291 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3294 Frequency : lfs setstripe, only systems with more than 160 OSTs
3296 Description: unable to set striping with a starting offset beyond OST 160
3297 Details : llapi_create_file() incorrectly limited the starting stripe
3298 index to the maximum single-file stripe count.
3301 Frequency : LDAP users only
3303 Description: lconf did not handle in-kernel recovery with LDAP properly
3304 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3306 Severity : enhancement
3308 Description: bind OST threads to NUMA nodes to improve performance
3309 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3310 node and do their allocations there to localize memory access
3312 Severity : enhancement
3314 Description: llmount can determine client NID directly from Myrinet (GM)
3315 Details : the client NID code from gmnalnid was moved directly into
3316 llmount, removing the need to use this or specifying the
3317 client NID explicitly when mounting GM clients with zeroconf
3320 Frequency : if client is started with down MDS
3322 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3323 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3324 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3325 ptlrpc_import_delayed_req() because we didn't check if the
3326 request was interrupted, and we also didn't break out of the
3327 event loop if there was no timeout
3332 Description: data loss during non-page-aligned writes to a single file from
3333 both multiple nodes and multiple threads on one node at same time
3334 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3335 inconsistency led to false short-reads, that were cached and later
3336 used by ->prepare_write() to fill in partially written page,
3337 leading to data loss.
3340 Frequency : always, if lconf --abort_recovery used
3342 Description: lconf --abort_recovery fails with 'Operation not supported'
3343 Details : lconf was attempting to abort recovery on the MDT device and not
3346 Severity : enhancement
3348 Description: remove cleanup logs
3349 Details : replace lconf-generated cleanup logs with lustre internal
3350 cleanup routines. Eliminates the need for client-cleanup and
3353 Severity : enhancement
3355 Description: add support for EAs (user and system) on lustre filesystems
3356 Details : it is now possible to store extended attributes in the Lustre
3357 client filesystem, and with the user_xattr mount option it
3358 is possible to allow users to store EAs on their files also
3360 Severity : enhancement
3362 Description: Add possibility (config option) to show minimal available OST free
3364 Details : When compiled with --enable-mindf configure option, statfs(2)
3365 (and so, df) will return least minimal free space available from
3366 all OSTs as amount of free space on FS, instead of summary of
3367 free spaces of all OSTs.
3369 Severity : enhancement
3371 Description: do not expand extent locks acquired on OST-side
3372 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3373 by server: they are not cached anyway.
3376 Frequency : when mmap is used/binaries executed from Lustre
3378 Description: Unmmap pages before throwing them away from read cache.
3379 Details : llap_shrink cache now attempts to unmap pages before discarding
3380 them (if unmapping failed - do not discard). SLES9 kernel has
3381 extra checks that trigger if this unmapping is not done first.
3386 Description: lconf didn't resolve symlinks before checking to see whether a
3387 given mountpoint was already in use
3390 Frequency : when migrating failover services
3391 Bugzilla : 6395, 9514
3392 Description: When migrating a subset of services from a node (e.g. failback
3393 from a failover service node) the remaining services would
3394 time out and evict clients.
3395 Details : lconf --force (implied by --failover) sets the global obd_timeout
3396 to 5 seconds in order to quickly disconnect, but this caused
3397 other RPCs to time out too quickly. Do not change the global
3398 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3400 Severity : enhancement
3401 Frequency : if MDS is started with down OST
3402 Bugzilla : 9439,5706
3403 Description: Allow startup/shutdown of an MDS without depending on the
3404 availability of the OSTs.
3405 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3406 Add appropriate locking and lov-osc refcounts for safe
3407 cleaning. Add osc abort_inflight calls in case the
3408 synchronize never started.
3411 Frequency : occasional (Cray XT3 only)
3413 Description: root not authorized to access files in CRAY_PORTALS environment
3414 Details : The client process capabilities were not honoured on the MDS in
3415 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3416 been accessed by an authorized user then root was able to access
3417 the file on the local client also. The root user capabilities
3418 are now allowed on the MDS, as this environment has secure UID.
3421 Frequency : occasional
3423 Description: ldiskfs "too long searching" message happens too often
3424 Details : A debugging message (otherwise harmless) prints too often on
3425 the OST console. This has been reduced to only happen when
3426 there are fragmentation problems on the filesystem.
3431 Description: Division by zero in statfs when all OSCs are inactive
3432 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3433 lov_statfs divides by value returned by lov_get_stripecnt().
3437 Bugzilla : 9489, 3273
3438 Description: First write from each client to each OST was only 4kB in size,
3439 to initialize client writeback cache, which caused sub-optimal
3440 RPCs and poor layout on disk for the first writen file.
3441 Details : Clients now request an initial cache grant at (re)connect time
3442 and so that they can start streaming writes to the cache right
3443 away and always do full-sized RPCs if there is enough data.
3444 If the OST is rebooted the client also re-establishes its grant
3445 so that client cached writes will be honoured under the grant.
3450 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3451 Details : Now I/O RPCs go to different portal number and (presumably) fast
3452 lock requests (and glimses) and other RPCs get their own service
3453 threads pool that should be able to service those RPCs
3456 Severity : enhancement
3458 Description: Ability to exchange lustre version between client and servers and
3459 issue warnings at client side if client is too old. Also for
3460 liblustre clients there is ability to refuse connection of too old
3462 Details : New 'version' field is added to connect data structure that is
3463 filled with version info. That info is later checked by server and
3467 Frequency : rare, liblustre only.
3468 Bugzilla : 9296, 9581
3469 Description: Two simultaneous writes from liblustre at offset within same page
3470 might proceed at the same time overwriting eachother with stale
3472 Details : I/O lock withing llu_file_prwv was released too early, before data
3473 actually was hitting the wire. Extended lock-holding time until
3474 server acknowledges receiving data.
3477 Frequency : extremely rare. Never observed in practice.
3479 Description: avoid generating lustre_handle cookie of 0.
3480 Details : class_handle_hash() generates handle cookies by incrementing
3481 global counter, and can hit 0 occasionaly (this is unlikely, but
3482 not impossible, because initial value of cookie counter is
3483 selected randonly). Value of 0 is used as a sentinel meaning
3484 "unassigned handle" --- avoid it. Also coalesce two critical
3485 sections in this function into one.
3487 Severity : enhancement
3489 Description: allow liblustre clients to delegate truncate locking to OST
3490 Details : To avoid overhead of locking, liblustre client instructs OST to
3491 take extent lock in ost_punch() on client's behalf. New connection
3492 flag is added to handle backward compatibility.
3494 Severity : enhancement
3495 Bugzilla : 4928, 7341, 9758
3496 Description: allow number of OST service threads to be specified
3497 Details : a module parameter allows the number of OST service threads
3498 to be specified via "options ost ost_num_threads={N}" in the
3499 OSS's /etc/modules.conf or /etc/modprobe.conf.
3503 Bugzilla : 6146, 9635, 9895
3504 Description: servers crash with bad pointer in target_handle_connect()
3505 Details : In rare cases when a client is reconnecting it was possible that
3506 the connection request was the last reference for that export.
3507 We would temporarily drop the export reference and get a new
3508 one, but this may have been the last reference and the export
3509 was just destroyed. Get new reference before dropping old one.
3511 Severity : enhancement
3512 Frequency : if client is started with failover MDS
3514 Description: Allow multiple MDS hostnames in the mount command
3515 Details : Try to read the configuration from all specified MDS
3516 hostnames during a client mount in case the "primary"
3519 Severity : enhancement
3521 Description: Stop sending data to evicted clients as soon as possible.
3522 Details : Check if the client we are about to send or are sending data to
3523 was evicted already. (Check is done every second of waiting,
3524 for which l_wait_event interface was extended to allow checking
3525 of exit condition at specified intervals).
3528 Frequency : rare, normally only when NFS exporting is done from client
3530 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3532 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3535 Frequency : always, if srand() is called before liblustre initialization
3537 Description: Liblustre uses system PRNG disturbing its usage by user application
3538 Details : Introduce internal to lustre fast and high-quality PRNG for
3539 lustre usage and make liblustre and some other places in generic
3540 lustre code to use it.
3542 Severity : enhancement
3543 Bugzilla : 9477, 9557, 9870
3544 Description: Verify that the MDS configuration logs are updated when xml is
3545 Details : Check if the .xml configuration logs are newer than the config
3546 logs stored on the MDS and report an error if this is the case.
3547 Request --write-conf, or allow starting with --old_conf.
3549 Severity : enhancement
3551 Description: Handle symlinks in the path when checking if Lustre is mounted.
3552 Details : Resolve intermediate symlinks when checking if a client has
3553 mounted a filesystem to avoid duplicate client mounts.
3558 Description: lconf can hit an error exception but still return success.
3559 Details : The lconf command catches the Command error exception at the top
3560 level script context and will exit with the associated exit
3561 status, but doesn't ensure that this exit status is non-zero.
3566 Description: failure of ptlrpc thread startup can cause oops
3567 Details : Starting a ptlrpc service thread can fail if there are a large
3568 number of threads or the server memory is very fragmented.
3569 Handle this without oopsing.
3572 Frequency : always, only if liblustre and non-default acceptor port was used
3574 Description: liblustre cannot connect to servers with non-default acceptor port
3575 Details : tcpnal_set_default_params() was not called and was therefore
3576 ignoring the environment varaible TCPNAL_PORT, as well as other
3577 TCPNAL_ environment variables
3582 Description: two objects could be created on the same OST for a single file
3583 Details : If an OST is down, in some cases it was possible to create two
3584 objects on a single OST for a single file. No problems other
3585 than potential performance impact and spurious error messages.
3589 Bugzilla : 5681, 9562
3590 Description: Client may oops in ll_unhash_aliases
3591 Details : Client dcache may become inconsistent in race condition.
3592 In some cases "getcwd" can fail if the current directory is
3598 Description: Inode refcounting problems in NFS export code
3599 Details : link_raw functions used to call d_instantiate without obtaining
3600 extra inode reference first.
3604 Bugzilla : 9942, 9903
3605 Description: Referencing freed requests leading to crash, memleaks with NFS.
3606 Details : We used to require that call to ll_revalidate_it was always
3607 followed by ll_lookup_it. Also with revalidate_special() it is
3608 possible to call ll_revalidate_it() twice for the same dentry
3609 even if first occurence returned success. This fix changes semantic
3610 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3611 reference on a request referred from the intent.
3612 ll_intent_release() then releases such a request.
3615 Frequency : rare, normally benchmark loads only
3617 Description: unlinked inodes were kept in memory on the client
3618 Details : If a client is repeatedly creating and unlinking files it
3619 can accumulate a lot of stale inodes in the inode slab cache.
3620 If there is no other client load running this can cause the
3621 client node to run out of memory. Instead flush old inodes
3622 from client cache that have the same inode number as a new inode.
3625 Frequency : SLES9 2.6.5 kernel and long filenames only
3626 Bugzilla : 9969, 10379
3627 Description: utime reports stale NFS file handle
3628 Details : SLES9 uses out-of-dentry names in some cases, which confused
3629 the lustre dentry revalidation. Change it to always use the
3633 Frequency : rare, unless heavy write-truncate concurrency is continuous
3634 Bugzilla : 4180, 6984, 7171, 9963, 9331
3635 Description: OST becomes very slow and/or deadlocked during object unlink
3636 Details : filter_destroy() was holding onto the parent directory lock
3637 while truncating+unlinking objects. For very large objects this
3638 may block other threads for a long time and slow overall OST
3639 responsiveness. It may also be possible to get a lock ordering
3640 deadlock in this case, or run out of journal credits because of
3641 the combined truncate+unlink. Solution is to do object truncate
3642 first in one transaction without parent lock, and then do the
3643 final unlink in a new transaction with the parent lock. This
3644 reduces the lock hold time dramatically.
3647 Frequency : rare, 2.4 kernels only
3649 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3650 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3651 before unmounting their filesystem in order to avoid deadlock.
3652 The kernel_locked() function in 2.4 kernels only checks whether
3653 the kernel lock is held, not whether it is this process that is
3654 holding it as 2.6 kernels do.
3659 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3660 Details : The client ptlrpc code may be trying to reconnect to a down
3661 server before a previous connection attempt has timed out.
3662 Increase the reconnect interval to be longer than the connection
3663 timeout interval to avoid sending duplicate connections to
3667 Frequency : echo_client brw_test command
3669 Description: fix echo_client to work with OST preallocated code
3670 Details : OST preallocation code (5137) didn't take echo_client IO path
3671 into account: echo_client calls filter methods outside of any
3672 OST thread and, hence, there is no per-thread preallocated
3673 pages and buffers to use. Solution: hijack pga pages for IO. As
3674 a byproduct, this avoids unnecessary data copying.
3678 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3679 Description: Client can oops in mdc_commit_close() after open replay
3680 Details : It was possible for the MDS to return an open request with no
3681 transaction number in mds_finish_transno() if the client was
3682 evicted, but without actually returning an error. Clients
3683 would later try to replay that open and may trip an assertion
3684 Simplify the client close codepath, and always return an error
3685 from the MDS in case the open is not successful.
3688 Frequency : rare, 2.6 OSTs only
3690 Description: OST may deadlock under high load on fragmented files
3691 Details : If there was a heavy load and highly-fragmented OST filesystems
3692 it was possible to have all the OST threads deadlock waiting on
3693 allocation of biovecs, because the biovecs were not released
3694 until the entire RPC IO was completed. Instead, release biovecs
3695 as soon as they are complete to ensure forward IO progress.
3697 Severity : enhancement
3699 Description: Support for specifying external journal device at mount
3700 Details : If an OST or MDS device is formatted with an external journal
3701 device, this device major/minor is stored in the ext3 superblock
3702 and may not be valid for failover. Allow detecting and
3703 specifying the external journal at mount time.
3708 Description: Mounting an MDS with pending unlinked files may cause oops
3709 Details : target_finish_recovery() calls mds_postrecov() which returned
3710 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3711 considers this an error and immediately begins cleaning up the
3712 lov, just after starting the mds_lov process
3714 Severity : enhancement
3716 Description: Implement 'lfs df' to report actual free space on per-OST basis
3717 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3718 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3719 usage of objects; '-h' to report in human readable format.
3721 ------------------------------------------------------------------------------
3723 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
3730 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3731 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3732 at startup during orphan unlinking. Instead, skip buddy update
3733 before setup, as it will be regenerated after recovery is complete.
3738 Description: If an OST is inactive, its locks might reference stale inodes.
3739 Details : lov_change_cbdata() must iterate over all namespaces, even if
3740 they are inactive to clear inode references from the lock.
3742 Severity : enhancement
3743 Frequency : occasional, if non-standard max_dirty_mb used
3745 Description: Client will block write RPCs if not enough grant
3746 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3747 then the client will block writes while waiting for another RPC
3748 to complete instead of consuming its dirty limit. With change
3749 we get improved performance when max_dirty_mb is small.
3751 Severity : enhancement
3752 Bugzilla : 3389, 6253
3753 Description: Add support for supplementary groups on the MDS.
3754 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3755 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3756 lookups for user supplementary groups into a cache.
3760 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3761 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3762 for liblustre O_EXCL is enforced by MDS.
3766 Description: readdir never returns NULL in liblustre.
3767 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3768 dentry in struct dirent.
3772 Frequency : liblustre only, depends on application IO pattern
3773 Description: liblustre clients evicted if not contacting servers
3774 Details : Don't put liblustre clients into the ping_evictor list, so
3775 they will not be evicted by the pinger ever.
3777 Severity : enhancement
3779 Description: Add ability to evict clients by NID from MDS.
3780 Details : By echoing "nid:$NID" string into
3781 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3782 $NID would be instantly evicted from this MDS and from all active
3783 OSTs connected to it.
3787 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3788 Details : lookup_it_finish() used to query file size from OSTs that was not
3793 Description: service threads change working directory to that of init
3794 Details : Starting lustre service threads may pin the working directory
3795 of the parent thread, making that filesystem busy. Threads
3796 now change to the working directory of init to avoid this.
3800 Frequency : during shutdown only
3801 Description: shutdown with a failed MDS or OST can cause unmount to hang
3802 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3807 Frequency : 2.6 only
3808 Description: chmod/chown may include an extra supplementary group
3809 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3810 supplementary group and if none is specified this is used.
3813 Bugzilla : 5479 (6816)
3815 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3816 Details : If lookup is in progress on a file that is unlinked we might try
3817 to revalidate the inode and fail in revalidate after lookup is
3818 complete and ll_file_open() enqueues the open again but
3819 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3822 Frequency : always, if lconf --abort_recovery used
3824 Description: lconf --abort_recovery fails with 'Operation not supported'
3825 Details : lconf was attempting to abort recovery on the MDT device and not
3828 ------------------------------------------------------------------------------
3830 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3835 Frequency : rare (only unsupported configurations with a node running as an
3837 Bugzilla : 6514, 5137
3838 Description: Mounting a Lustre file system on a node running as an OST could
3840 Details : OSTs now preallocates memory needed to write out data at
3841 startup, instead of when needed, to avoid having to
3842 allocate memory in possibly low memory situations.
3843 Specifically, if the file system is mounted on on OST,
3844 memory pressure could force it to try to write out data,
3845 which it needed to allocate memory to do. Due to the low
3846 memory, it would be unable to do so and the node would
3847 become unresponsive.
3849 Severity : enhancement
3851 Description: Addition of lconf --service command line option
3852 Details : lconf now accepts a '--service <arg>' option, which is
3853 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3855 Severity : enhancement
3857 Description: Failover mode is now the default for OSTs.
3858 Details : By default, OSTs will now run in failover mode. To return to
3859 the old behaviour, add '--failout' to the lmc line for OSTs.
3861 Severity : enhancement
3863 Description: Health checks are now provided for MDS and OSTs
3864 Details : Additional detailed health check information on MSD and OSTs
3865 is now provided through the procfs health_check value.
3868 Frequency : occasional, depends on IO load
3870 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3871 after numerous create/delete cycles
3872 Details : The ext3 inode allocation policy would not allocate new inodes
3873 very well on the OSTs because there are no new directories
3874 being created. Instead we look for groups with free space if
3875 the parent directories are nearly full.
3880 Description: Network or server problems during mount may cause partially
3881 mounted clients instead of returning an error.
3882 Details : The config llog parsing code may overwrite the error return
3883 code during mount error handling, returning success instead
3889 Description: MDS can fail to allocate large reply buffers
3890 Details : After long uptimes the MDS can fail to allocate large reply
3891 buffers (e.g. zconf client mount config records) due to memory
3892 fragmentation or consumption by the buffer cache. Preallocate
3893 some large reply buffers so that these replies can be sent even
3894 under memory pressure.
3898 Frequency : rare (liblustre)
3899 Description: fsx running with liblustre complained that using truncate() to
3900 extend the file doesn't work. This patch corrects that issue.
3901 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3902 ATTR_SIZE and lsm use in llu_setattr_raw.
3906 Frequency : rare, only 2.6 kernels
3907 Description: Unusual file access patterns on the MDS may result in inode
3908 data being lost in very rare circumstances.
3909 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3910 code in the 2.6 kernel caused the inode and/or EA data not to
3911 be read from disk, causing single-file corruption.
3915 Frequency : rare, only 2.6 filesystems using extents
3916 Description: Heavy concurrent write and delete load may cause data corruption.
3917 Details : It was possible under high-load situations to have an extent
3918 metadata block in the block device cache from a just-unlinked
3919 file overwrite a newly-allocated data block. We now unmap any
3920 metadata buffers that alias just-allocated data blocks.
3924 Frequency : filesystems with default stripe_count larger than 77
3925 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3926 Details : lconf specifies an inode size of 4096 bytes when the default
3927 stripe_count is larger than 77. This conflicts with the default
3928 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3929 this case to avoid pinning too much memory for large EAs.
3931 ------------------------------------------------------------------------------
3933 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3938 Frequency : rare (extremely heavy IO load with hundreds of clients)
3940 Description: Client is evicted, gets IO error writing to file
3941 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3942 caused clients to be evicted for AST timeouts. The fixes in
3943 bug 5192 mean we no longer need to have such short AST timeouts
3944 so ldlm_timeout has been increased.
3947 Frequency : occasional during --force or --failover shutdown under load
3948 Bugzilla : 5949, 4834
3949 Description: Server oops/LBUG if stopped with --force or --failover under load
3950 Details : a collection of import/export refcount and cleanup ordering
3951 issues fixed for safer force cleanup
3954 Frequency : only filesystems larger than 120 OSTs
3955 Bugzilla : 5990, 6223
3956 Description: lfs getstripe would oops on a very large filesystem
3957 Details : lov_getconfig used kfree on vmalloc'd memory
3960 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3961 Bugzilla : 6242, 6243
3962 Description: reading from files that had been truncated to a non-zero size
3963 but never opened returned no data
3964 Details : ll_file_read() reads zeros from no-object files to EOF
3969 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3970 Details : There's little harm in aborting MDS/OSS recovery and letting it
3971 try again, so I removed the LASSERT and return an error instead.
3973 Severity : enhancement
3975 Description: New debugging infrastructure for tracking down data corruption
3976 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3977 (b) cover more of the client-side code path, and (c) try to narrow
3978 down where problems occurred
3982 Bugzilla : 3819, 4364, 4397, 6313
3983 Description: Racing close and eviction MDS could cause assertion in mds_close
3984 Details : It was possible to get multiple mfd references during close and
3985 client eviction, leading to one thread referencing a freed mfd.
3987 Severity: : enhancement
3988 Bugzilla : 3262, 6359
3989 Description: Attempts to reconnect to servers are now more aggressive.
3990 Details : This builds on the enhanced upcall-less recovery that was added
3991 in 1.4.2. When trying to reconnect to servers, clients will
3992 now try each server in the failover group every 10 seconds. By
3993 default, clients would previously try one server every 25 seconds.
3998 Description: After recovery, certain operations trigger a failed
3999 assertion on a client.
4000 Details : Failing over an mds, using lconf -d --failover, while a
4001 client was doing a readdir() call would cause the client to
4002 LBUG after recovery completed and the readdir() was resent.
4004 Severity : enhancement
4006 Description: Default groups are now added by lconf
4007 Details : You can now run lconf --group <servicename> without having to
4008 manually add groups with lmc.
4011 Frequency : occasional
4013 Description: Nodes with an elan id of 0 trigger a failed assertion
4016 Frequency : always when accessing e.g. tty/console device nodes
4018 Description: tty and some other devices nodes cannot be used on lustre
4019 Details : file's private_data field is used by device data and lustre
4020 values in there got lost. New field was added to struct file to
4021 store fs-specific private data.
4024 Frequency : when exporting Lustre via NFS
4026 Description: NFSD failed occasionally when looking up a path component
4027 Details : NFSD is looking up ".." which was broken in ext3 directories
4028 that had grown large enough to become hashed.
4031 Frequency : Clusters with multiple interfaces not on the same subnet
4033 Description: Nodes will repeatedly try to reconnect to an interface which it
4034 cannot reach and report an error to the log.
4035 Details : Extra peer list entries will be created by lconf with some peers
4036 unreachable. lconf now validates the peer before adding it.
4039 Frequency : Only if a default stripe is set on the filesystem root.
4041 Description: Setting a default stripe on the filesystem root prevented the
4042 filesystem from being remounted.
4043 Details : The client was sending extra request flags in the root getattr
4044 request and did not allocate a reply buffer for the dir EA.
4047 Frequency : occasional, higher if lots of files are accessed by one client
4048 Bugzilla : 6159, 6097
4049 Description: Client trips assertion regarding lsm mismatch/magic
4050 Details : While revalidating inodes the VFS looks up inodes with ifind()
4051 and in rare cases can find an inode that is being freed.
4052 The ll_test_inode() code will free the lsm during ifind()
4053 when it finds an existing inode and then the VFS later attaches
4054 this free lsm to a new inode.
4058 Bugzilla : 6422, 7030
4059 Description: MDS deadlock between mkdir and client eviction
4060 Details : Creating a new file via mkdir or mknod (starting a transaction
4061 and getting the ns lock) can deadlock with client eviction
4062 (gets ns lock and trying to finish a synchronous transaction).
4065 Frequency : occasional
4066 Description: While starting a server, the fsfilt_ext3 module could not be
4068 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4069 kernels. Previously, lconf would still use the ext3 name
4070 when trying to load modules. Now, it will correctly use
4071 ext3 on 2.4 and ldiskfs on 2.6.
4073 Severity : enhancement
4074 Description: The default stripe count has been changed to 1
4075 Details : The interpretation of the default stripe count (0, to lfs
4076 or lmc) has been changed to mean striping across a single
4077 OST, rather than all available. For general usage we have
4078 found a stripe count of 1 or 2 works best.
4080 Severity : enhancement
4081 Description: Add support for compiling against Cray portals.
4082 Details : Conditional compiling for some areas that are different
4086 Frequency : occasional
4087 Bugzilla : 6409, 6834
4088 Description: Creating files with an explicit stripe count may lead to
4089 a failed assertion on the MDS
4090 Details : If some OSTs are full or unavailable, creating files may
4091 trigger a failed assertion on the MDS. Now, Lustre will
4092 try to use other servers or return an error to the
4096 Frequency : occasional
4098 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4099 caused lock timeout during readahead (since 1.4.2).
4100 Details : Processes doing readahead might match a lock that hasn't been
4101 granted yet if there are overlapping and conflicting lock
4102 requests. The readahead process waits on ungranted lock
4103 (original lock is CBPENDING), while OST waits for that process
4104 to cancel CBPENDING read lock and eventually evicts client.
4106 Severity : enhancement
4108 Description: Initial enabling of flock support for clients
4109 Details : Implements fcntl advisory locking and file status functions.
4110 This feature is provided as an optional mount flag (default
4111 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4112 locking are implemented yet, and those that are are not guaranteed
4113 to be completely correct in production environments.
4114 mount -t lustre -o [flock|noflock] ...
4117 Frequency : occasional
4119 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4120 assertion in the ext3 JBD (journaling) layer.
4121 Details : The b_committed_data struct is protected by the big kernel lock
4122 in 2.4 kernels, serializing journal_commit_transaction() and
4123 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4124 access to this struct. In 2.6 kernels there is finer grained
4125 locking to improve SMP performance of the JBD layer.
4129 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4131 -----------------------------------------------------------------------------
4133 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4135 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4136 versions of Lustre. Please update all servers and clients to
4137 version 1.4.2 or later at the same time. You must also run
4138 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4139 to update the configuration logs.
4141 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4142 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4143 - swab llog records of type '0' so we get proper header size/idx (5861)
4144 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4145 - fix rename of one directory over another leaking an inode (5953)
4146 - avoid SetPageDirty on 2.6 (5981)
4147 - don't re-add just-being-destroyed locks to the waiting list (5653)
4148 - when creating new directories, inherit the parent's custom
4149 striping settings if present parent (3048)
4150 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4151 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4152 - don't hold client locks on temporary worklist from l_lru (5666)
4153 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4154 - automatically evict dead clients (5921)
4155 - Update file size properly in create+truncate+fstat case (6196)
4156 - Do not unhash mountpoint dentries, do not allow removal of
4158 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4159 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4160 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4161 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4162 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4163 - Correctly handle reads of files with no objects (6243)
4164 - lctl recover will also mark a device active if deactivate used (5933)
4166 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4167 - allow --write-conf on an MDS with different nettype than client (5619)
4168 - don't write config llogs to MDS for mounts not from that MDS (5617)
4169 - lconf should create multiple TCP connections from a client (5201)
4170 - init scripts are now turned off by default; run chkconfig --on
4171 lustre and chkconfig --on lustrefs to use them
4172 - upcalls are no longer needed for clients to recover to failover
4174 - add --abort-recovery option to lconf to abort recovery on device
4176 - add support for an arbitrary number of OSTs (3026)
4177 - Quota support protocol changes.
4178 - forward compatibility changes to wire structs (6007)
4179 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4180 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4181 - improved "lustre status" script
4182 - initialize blocksize for non-regular files (6062)
4183 - added --disable-server and --disable-client configure options (5782)
4184 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4185 - Vanilla 2.4.29 support
4186 - increase maximum number of obd devices to 520 (6242)
4187 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4188 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4189 SLES 9 2.6.5 kernels
4190 - limit stripes per file to 160 (the maximum EA size) (6093)
4192 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4195 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4196 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4197 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4198 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4199 - small ext3 extents cleanups and fixes (5733)
4200 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4201 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4202 - avoid needless client->OST connect, fix handle mismatch (5317)
4203 - fix DLM error path that led to out-of-sync client, long delays (5779)
4204 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4205 - fix several locking issues related to i_size (5492,5624,5654,5672)
4206 - don't move pending lock onto export if it is already evicted (5683)
4207 - fix kernel oops when creating .foo in unlinked directory (5548)
4208 - fix deadlock in obdfilter statistics vs. object create (5811)
4209 - use time_{before,after} to avoid timer jiffies wrap (5882)
4210 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4211 - Do not leak request if server was not able to process it (5154)
4212 - If mds_open unable to find parent dir, make that negative lookup(5154)
4213 - don't create new directories with extent-mapping (5909, 5936)
4215 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4216 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4217 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4218 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4219 - support for RHEL4 kernel on i686 (5773)
4220 - provide error messages when incompatible logs are encountered (5898)
4222 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4223 * version 1.4.0.10 (1.4.1 release candidate 1)
4225 - don't keep a lock reference when lock is not granted (4238)
4226 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4227 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4228 - drop import inflight refcount on signal_completed_replay error (5255)
4229 - unlock page after async write error during send (3677)
4230 - handle missing objects in filter_preprw_read properly (5265)
4231 - no transno return for symlink open, don't save no-trasno open (3440)
4232 - don't try to complete elan receive that already failed (4012)
4233 - free RPC server reply state on error (5406)
4234 - clean up thread from ptlrpc_start_thread() on error (5160)
4235 - readahead could read extra page into cache that wasn't ejected (5388)
4236 - prevent races in class_attach/setup/cleanup/detach (5260)
4237 - don't dereference de->d_inode after l_dput of de (5458)
4238 - use "int" for stripe value returned from lock_to_stripe (5544)
4239 - mballoc allocation and error-checking fixes in 2.6 (5504)
4240 - block device patches to fix I/O request sizes in 2.6 (5482)
4241 - look up hostnames for IB nals (5602)
4242 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4243 - don't start multiple acceptors for the same port (5277)
4244 - fix incorrect LASSERT in mds_getattr_name (5635)
4245 - export a proc file for general "ping" checking (5628)
4246 - fix "lfs check" to not block when the MDS is down (5628)
4248 - service request history (4965)
4249 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4250 - create an "evict_client" /proc entry on OSTs, like the MDS has
4251 - fix mount usage message, return errors per mount(8) (5168)
4252 - change grep [] to grep "[]" in tests so they work in more UMLs
4253 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4254 - remove ancient LOV_MAGIC_V0 EA support (5047)
4255 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4256 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4257 - allow readahead allocations to fail when low on memory (5383)
4258 - mmap locking landed again, after considerable improvement (2828)
4259 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4261 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4264 - send OST transaction number in read/write reply to free req (4966)
4265 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4266 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4267 - the watchdog thread now runs as interruptible (5246)
4268 - flock/lockf fixes (but it's still disabled, pending 5135)
4269 - don't use EXT3 constants in llite code (5094)
4270 - memory shortage at startup could cause assertion (5176)
4272 - reorganization of lov code
4273 - single portals codebase
4275 - add extents/mballoc support (5025)
4276 - direct I/O reads in the obdfilter (4048)
4277 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4279 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4282 - send OST transaction number in read/write reply to free req (4966)
4283 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4284 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4285 - the watchdog thread now runs as interruptible (5246)
4286 - handle missing objects in filter_preprw_read properly (5265)
4287 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4288 - drop import inflight refcount on signal_completed_replay error (5255)
4289 - unlock page after async write error during send (3677)
4290 - return original error code on reconstructed replies (3761)
4291 - no transno return for symlink open, don't save no-trasno open (3440)
4293 - add pid to ldlm debugging output (4922)
4294 - bump the watchdog timeouts -- we can't handle 30sec yet
4295 - extra debugging for orphan dentry/inode bug (5259)
4297 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4300 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4301 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4302 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4303 - fix NULL dereference in /proc/sys/portals/routes (4827)
4304 - allow failed mdc_close() operations to be interrupted (4561)
4305 - stop precreate on OST before MDS would time out on it (4778)
4306 - don't send partial-page writes before EOF from client (4410)
4307 - discard client grant for sub-page writes on large-page clients (4520)
4308 - don't free dentries not owned by NFS code, check generation (4806)
4309 - fix lsm leak if mds_create_objects() fails (4801)
4310 - limit debug_daemon file size, always print CERROR messages (4789)
4311 - use transno after validating reply (3892)
4312 - process timed out requests if import state changes (3754)
4313 - update mtime on OST during writes, return in glimpse (4829)
4314 - add mkfsoptions to LDAP (4679)
4315 - use ->max_readahead method instead of zapping global ra (5039)
4316 - don't interrupt __l_wait_event() during strace
4318 - add software watchdogs to catch hung threads quickly (4941)
4319 - make lustrefs init script start after nfs is mounted
4320 - fix CWARN/ERROR duplication (4930)
4321 - return async write errors to application if possible (2248)
4322 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4323 - print NAL number in %x format (4645)
4324 - update barely-supported suse-2.4.21-171 series (4842)
4325 - support for sles 9 %post scripts
4326 - support for building 2.6 kernel-source packages
4327 - support for sles km_* packages
4329 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4332 - ignore -ENOENT errors in osc_destroy (3639)
4333 - notify osc create thread that OSC is being cleaned up (4600)
4334 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4335 - reconstruct ost_handle() like mds_handle() (4657)
4336 - create a new thread to do import eviction to avoid deadlock (3969)
4337 - let lconf resolve symlinked-to devices (4629)
4338 - don't unlink "objects" from directory with default EA (4554)
4339 - hold socknal file ref over connect in case target is down (4394)
4340 - allow more than 32000 subdirectories in a single directory (3244)
4341 - fix blocks count for O_DIRECT writes (3751)
4342 - OST returns ENOSPC from object create when no space left (4539)
4343 - don't send truncate RPC if file size isn't changing (4410)
4344 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4345 - bind to privileged port in socknal and tcpnal (3689)
4347 - rate limit CERROR/CWARN console message to avoid overload (4519)
4348 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4349 - basic mmap support (3918)
4350 - kernel patch series update from b1_4 (4711)
4352 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4355 - avoid crash during MDS cleanup with OST shut down (2775)
4356 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4357 - don't use bad inodes on the MDS (3744)
4358 - dynamic object preallocation to improve recovery speed (4236)
4359 - don't hold spinlock over lock dumping or change debug flags (4401)
4360 - don't zero obd_dev when it is force cleaned (3651)
4361 - print grants to console if they go negative (4431)
4362 - "lctl deactivate" will stop automatic recovery attempts (3406)
4363 - look for existing locks in ldlm_handle_enqueue() (3764)
4364 - don't resolve lock handle twice in recovery avoiding race (4401)
4365 - revalidate should check working dir is a directory (4134)
4367 - don't always mark "slow" obdfilter messages as errors (4418)
4369 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4372 - don't close LustreDB during write_conf until it is done (3860)
4373 - fix typo in lconf for_each_profile (3821)
4374 - allow dumping logs from multiple threads at one time (3820)
4375 - don't allow multiple threads in OSC recovery (3812)
4376 - fix debug_size parameters (3864)
4377 - fix mds_postrecov to initialize import for llog ctxt (3121)
4378 - replace config semaphore with spinlock (3306)
4379 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4380 - don't allow enqueue to complete on a destroyed export (3822)
4381 - down write_lock before checking llog header bitmap (3825)
4382 - recover from lock replay timeout (3764)
4383 - up llog sem before sending rpc (3652)
4384 - reduce ns lock hold times when setting kms (3267)
4385 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4386 - fix NULL deref and obd_dev leak on setup error (3312)
4387 - replace some LBUG about llog ops with error handling (3841)
4388 - don't match INVALID dentries from d_lookup and spin (3784)
4389 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4390 - fix invalid assertion in ptlrpc_set_wait (3880)
4392 - add libwrap support for the TCP acceptor (3996)
4393 - add /proc/sys/portals/routes for non-root route listing (3994)
4394 - allow setting MDS UUID in .xml (2580)
4395 - print the stack of a process that LBUGs (4228)
4397 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4400 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4401 - make sure to unset replay flag from failed open requests (3440)
4402 - if default stripe count is 0, use OST count for inode size (3636)
4403 - update parent mtime/ctime on client for create/unlink (2611)
4404 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4405 - free recovery state on server during a forced cleanup (3571)
4406 - unregister_reply for resent reqs (3063)
4407 - loop back devices mounting and status check on 2.6 (3563)
4408 - fix resource-creation race that can provoke i_size == 0 (3513)
4409 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4410 - more debugging for page-accounting assertion (3746)
4411 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4412 - avoid deadlock after precreation failure (3758)
4413 - fix race and lock order deadlock in orphan handling (3450, 3750)
4414 - add validity checks when grabbing inodes from l_ast_data (3599)
4416 - add /proc/.../recovery_status to obdfilter (3428)
4417 - lightweight CDEBUG infrastructure, debug daemon (3668)
4418 - change default OSC RPC parameters to be better on small clusters
4419 - turn off OST read cache for files smaller than 32MB
4420 - install man pages and include them in rpms (3100)
4421 - add new init script for (un)mounting lustre filesystems (2593)
4422 - run chkconfig in %post for init scripts (3701)
4423 - drop scimac NAL (unmaintained)
4425 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4428 - clean kiobufs before and after use (3485)
4429 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4430 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4431 - change init script to fail more clearly if not run as root (1528)
4432 - allow clients to reconnect during replay (1742)
4433 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4434 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4435 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4436 - fix cond_resched (3554)
4437 - only evict unfinished clients after recovery (3515)
4438 - allow bulk resend, prevent data loss (3570)
4439 - dynamic ptlrpc request buffer allocation (2102)
4440 - don't allow unlinking open directory if it isn't empty (2904)
4441 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4442 - remove extraneous obd dereference causing LASSERT failure (3334)
4443 - don't use get_cycles() when creating temp. files on the mds (3156)
4444 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4445 - handle EEXIST for set-stripe, set proper directory name (3336)
4447 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4448 - fix ksocknal_fmb_callback() error messages (2918)
4450 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4453 - don't copy lvb into (possibly NULL) reply on error (2983)
4454 - don't deref dentry after dput, don't free lvb on error (2922)
4455 - use the kms to determine writeback rpc length (2947)
4456 - increment oti_logcookies when osc is inactive (2948)
4457 - update client's i_blocks count via lvb messages (2543)
4458 - handle intent open/close of special files properly (1557)
4459 - mount MDS with errors=remount-ro, like obdfilter (2009)
4460 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4461 - don't use cancelling-locks' kms values (2947)
4462 - use highest lock extent for kms, not last one (2925)
4463 - don't dereference ERR_PTR() dentry in error handling path (3107)
4464 - fix thread race in portals_debug_dumplog() (3122)
4465 - create lprocfs device entries at setup instead of at attach (1519)
4466 - common AST error handler, don't evict client on completion race (3145)
4467 - zero nameidata in detach_mnt in 2.6 (3118)
4468 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4469 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4470 - initialize RPC timeout timer earlier for 2.6 (3219)
4471 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4472 - print nal/nid for unknown nid (3258)
4473 - additional checks for oscc recovery before doing precreate (3284)
4474 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4475 - don't crash in mdc_close for bad permissions on open (3285)
4476 - zero i_rdev for non-device files (3147)
4477 - clear page->private before handing to FS, better assertion (3119)
4478 - tune the read pipeline (3236)
4479 - fix incorrect decref of invalidated dentry (2350)
4480 - provide read-ahead stats and refine rpc in flight stats (3328)
4481 - don't hold journal transaction open across create RPC (3313)
4482 - update atime on MDS at close time (3265)
4483 - close LDAP connection when recovering to avoid server load (3315)
4484 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4485 - don't leak open file on MDS after open resend (3325)
4486 - serialize filter_precreate and filter_destroy_precreated (3329)
4487 - loop device shouldn't call sync_dev() for nul device (3092)
4488 - clear page cache after eviction (2766)
4489 - resynchronize MDS->OST in background (2824)
4490 - refuse to mount the same filesystem twice on same mountpoint (3394)
4491 - allow llmount to create routes for mounting behind routers (3320)
4492 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4493 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4494 - fix rare problem with rename on htree directories (3417)
4496 - allow default OST striping configuration per directory (1414)
4497 - fix compilation for qswnal for 2.6 kernels (3125)
4498 - increase maximum number of MDS request buffers for large systems
4499 - change liblustreapi to be useful for external progs like lfsck (3098)
4500 - increase local configuration timeout for slow disks (3353)
4501 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4503 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4506 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4507 - don't overwrite extent policy data in reply if lock was blocked (2901)
4508 - drop filter export grants atomically with removal from device (2663)
4509 - del obd_self_export from work_list in class_disconnect_exports (2908)
4510 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4511 - swab reply message in mdc_close, other PPC fixes (2464)
4512 - fix destroying of named logs (2325)
4513 - overwrite old logs when running lconf --write_conf (2264)
4514 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4515 - fix race in target_handle_connect (2898)
4516 - mds_reint_create() should take same inode create lock (2926)
4517 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4518 - don't close files for self_export to avoid uninitialized obd (2936)
4519 - allow MDS with the same name as client node (2939)
4520 - hold dentry reference for closed log files for unlink (2325)
4521 - reserve space for all logs during transactions (2059)
4522 - don't evict page beyond end of stripe extent (2925)
4523 - don't oops on a deleted current working directory (2399)
4524 - handle hard links to targets without a parent properly (2517)
4525 - don't dereference NULL lock when racing during eviction (2867)
4526 - don't grow lock extents when lots of conflicting locks (2919)
4528 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4531 - account for cache space usage on clients to avoid data loss (974)
4532 - lfsck support in lustre kernel code (2349)
4533 - reduce journal credits needed for BRW writes (2370)
4534 - orphan handling to avoid losing space on client/server crashes
4535 - ptlrpcd can be blocked, stopping ALL progress (2477)
4536 - use lock value blocks to assist in proper KMS, faster stat (1021)
4537 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4538 - recovery for initial connections (2355)
4539 - fixes for mds_cleanup_orphans (1934)
4540 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4541 - block all file creations until orphan recovery completes (1901)
4542 - client remove rq_connection from request struct (2423)
4543 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4544 - recovery timer race (2670)
4545 - mdc_close recovey bug (2532)
4546 - ptlrpc cleanup bug (2710)
4547 - mds timeout on local locks (2588)
4548 - namespace lock held during RPCs (2431)
4549 - handle interrupted sync write properly (2503)
4550 - don't try to handle a message that hasn't been replied to (2699)
4551 - client assert failure during cleanup after abort recovery (2701)
4552 - leak mdc device after failed mount (2712)
4553 - ptlrpc_check_set allows timedout requests to complete (2714)
4554 - wait for inflight reqs when ptlrpcd finishes (2710)
4555 - make sure unregistered services are removed from the srv_list
4556 - reset bulk XID's when resending them (caught by 1138 test)
4557 - unregister_bulk after timeout
4558 - fix lconf error (2694)
4559 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4560 - readahead locks pages, leaves pending causing memory pressure (2673)
4561 - increase OST request buffers to 4096 on large machines (2729)
4562 - fix up permission of existing directories in simple_mkdir (2661)
4563 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4564 - don't assign transno to errored transactions (2742)
4565 - don't delete objects on OST if given a bogus objid from MDS (2751)
4566 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4567 - if rq_no_resend, then timeout request after recovery (2432)
4568 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4569 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4570 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4571 - MDS recovery shouldn't delete objects that it has given out (2730)
4572 - if enqueue arrives after completion, don't clobber LVB (2819)
4573 - don't unlock pages twice when trigger_group_io returns error (2814)
4574 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4575 - don't write pages to disk if there was an error (1450)
4576 - don't ping imports that have recovery disabled (2676)
4577 - take buffered bytes into account when balancing socknal conn (2817)
4578 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4579 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4580 - remove little-endian swabbing of llog records (1987)
4581 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4582 - retry reposting request buffers if they fail (1191)
4583 - grow extent at grant time to avoid granting a revoked lock (2809)
4584 - lock revoke doesn't evict page if covered by a second lock (2765)
4585 - disable VM readahead to avoid reading outside lock extents (2805)
4587 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4588 - updated kernel patches for hp-2.4.20 kernel (2681)
4590 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4593 - fix truncated write corruption (2366)
4594 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4596 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4597 - ensure len(uuid) < 37 in lmc (1171)
4598 - fix ia64 OOPS in llog_test (2255)
4599 - zero end of page at obdfilter for partial page writes (2648)
4600 - don't leave stale dentries around after renames (bug 2428)
4601 - fix timeouts when evicting a client with a single lock held (2642)
4602 - set deadline for the initial HELLO message to drain (2634)
4603 - print out dotted-quad IP addresses in the socknal (2302)
4605 - additional debugging for MDS client eviction problem (2443)
4606 - fix mkfsoptions support for osts (2603, 2604)
4608 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4611 - add series for the vanilla 2.6.0 kernel
4612 - add series for the vanilla 2.4.24 kernel
4613 - add series for a cray x86/64 UL kernel drop
4614 - fix xattr patches for the vanilla 2.4.19 series
4616 - generate true UUIDs in lmc (1171)
4617 - have portals stack dumping break in UML (2466)
4618 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4619 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4620 - wake callback waiting threads on client eviction (2460)
4621 - Add --ptldebug and --subsystem to lmc (1719)
4622 - update assertion to allow safe interrupt allocation
4623 - set rq_no_resend for cancel requests (2432)
4624 - recalculate ptlrpcd timeout after resend (2494)
4625 - call vfs_rmdir when removing pending directories (2368)
4626 - fix renaming a file to itself (2429)
4627 - lmc creates a default one-stripe lov (2454)
4628 - expand procfs space to handle large clusters (2326)
4629 - increase UML stack to avoid overflow
4630 - update lconf's list of debug and subsystem masks
4631 - fix lfs find --obd (2510)
4632 - /proc tunable for disabling filter read caching (2591)
4633 - stop rpm packages from altering slapd.conf (2301)
4634 - disable nagle in the socknal under 0conf (2578)
4635 - choose mds inode size based on stripe count (2572)
4636 - fix kernel-source rpm problems (2516)
4638 - add --disable-doc to avoid pdf generation (2421)
4639 - update documentation, tests, type-os, comments
4640 - avoid format warnings on ia64
4641 - remove the TOE NAL
4642 - tiny code cleanups by removing unused fields
4644 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4647 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4648 - remove the most blatant lies from BUILDING file (2371)
4649 - change default debug level to reasonable production setting
4650 - reduce client side cache size to reduce cache flush time
4651 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4652 - make TCP zerocopy and pinger support enabled by default (2476)
4653 - sync writes completed after process exits caused crashes (2319)
4654 - maintain correct mount count on the MDS (2356)
4655 - backout 1557, because 2316 wasn't really fixed
4656 - better file I/O statistics gathering in /proc
4657 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4658 - another kernel patch to fix zero-copy TCP function export
4659 - don't take duplicate lock when processing re-sent getattr (2420)
4660 - lctl uses obd_self_export instead of creating new conn (2353)
4661 - MDS/OST recovery case which requires object creation asserted (2425)
4662 - move lfs from /usr/sbin to /usr/bin in packages
4663 - fix race between mds_client_add and mds_client_free (2417)
4664 - use kmalloc instead of slabs in portals (2430)
4665 - don't create duplicate records when a failover MDS is present (2442)
4666 - remove unnecessary mount age check (2332)
4667 - don't remove directory inodes from locks prematurely (2451)
4668 - don't break if MDS service name is the same as hostname (2103)
4669 - fix races in client write RPC generation when cache full (2482)
4671 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4674 - remove now-unused request->rq_obd (278)
4675 - if an allocation fails, print out how much memory we've used (1933)
4676 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4677 - add the "configurable stack size" patch to most series files (1256)
4678 - ability to write large log records, for 100+ OST configs (2306)
4679 - fix NULL deref when filter_prep fails (2314)
4680 - fix operator precedence error in filter_sync
4681 - dynamic allocation of socknal TX descriptors (2315)
4682 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4683 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4684 - ensure that CWARN messages in Portals always get to the syslog
4685 - __init/__exit are not for prototype decls (ldlm_init/exit)
4686 - x86-64 compile warning fixes
4687 - fix gateway LMC keyword conflict (2318)
4688 - fix MDS lock inversions in getattr/reint paths (1844)
4689 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4690 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4691 - DLM race condition prevented some lock evictions (2328)
4692 - ENOMEM detection and retry on socknal sends (2230)
4693 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4694 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4695 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4696 - better error messages when a client is rejected during recovery (1505)
4697 - avoid cancelling locks which were never granted, after failure (2330)
4698 - fix i_sem/journal inversion in mds_client_add (2333)
4699 - fix truncate/getattr lock cycle deadlock (2334)
4700 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4701 - fix two rare exit paths which could leak an l_lock() ref (2321)
4702 - fencepost error in MDS/OST orphan recovery (2226)
4703 - make log record alignment 8 bytes (1988)
4704 - lstripe now fails when requested offset > ost_count (2237)
4705 - ensure that all kernel series have a complete list.h (1607)
4706 - fix crashes in special-file operations (2316)
4707 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4709 - allow configurable automake binary, for testing new versions
4710 - small update to the lfs documentation
4712 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4714 * fix negative export reference count in fsfilt_sync (2312)
4716 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4717 * release candidate 0.9.1
4719 - orphans are moved into the PENDING directory for possible recovery
4720 - replayed opens now open by fid for orphan/rename safety (1042)
4721 - last close of an orphan inode generates a transno (683)
4722 - chdir() and mount() now pin the directory entry (1020)
4723 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4724 - discard very old requests without processing them (1502)
4725 - remove some common, well-understood CERRORs (1505)
4726 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4727 - clear "grant" flags in OST replies until OST grant code lands (1644)
4728 - fix read performance by not clobbering i_blksize on client (1598)
4729 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4730 - make LRU size a /proc tunable, clears locks when reduced (707)
4731 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4732 - server locks take references on exports now (1558)
4733 - build fixes for 2.4.20-rh trees (1663)
4734 - return an error from lov_create if all OSCs are inactive (1751)
4735 - fix import levels when a reconnect happens without a timeout (1597)
4736 - exit early from mds_open if we get a lookup error (1749)
4737 - partial page read at EOF wouldn't wait for disk before sending (1642)
4738 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4739 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4740 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4741 - write/truncate lock inversion (1639)
4742 - Don't auto-load obdclass, portals modules during cleanup (1495)
4743 - fix timestamps from jumping to "now" (1763)
4744 - extra journal assertions (1648)
4745 - add an extra multiunlink test (1771)
4746 - fix read_record/write_record API (1776)
4747 - fix leak of offset_extent, possible incorrect i_size later (1772)
4748 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4749 - Debugging for the kqswnal_get_idle_tx problems (1820)
4750 - Allow recovery to be attempted multiple times (1536)
4751 - Write out MDS last_rcvd file after it is first created (1600)
4752 - Fix tx_descriptor leak in failed transmit situations (1827)
4753 - ext3 journaling fixes for assertion failure after IO error (1871)
4754 - class_export_put() on freed export after completion AST error (1896)
4755 - Fix revalidate looping in VFS (1322)
4756 - Don't access a freed export during MDS_REINT timeout (1521)
4757 - Add open-unlink recovery support on the MDS (1673,1764)
4758 - Return an error if no MDS data was read from last_rcvd (1946)
4759 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4760 - Don't setup a disk device that doesn't match exported UUID (317)
4761 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4762 - avoid committing NULL handle in force close
4763 - local.sh is now a one-stripe LOV configuration
4764 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4765 - fix ext3 htree duplicate directory entry corruption (1516)
4766 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4767 - update to new LOV EA format (2097)
4768 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4770 2003-06-15 Phil Schwan <phil@clusterfs.com>
4773 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4774 - per-import/export recovery handling (958, 931, 959)
4775 - multiple last-rcvd slots, for serving multiple FSes (949)
4776 - connections are again shared between multiple imp/exports (963, 964)
4777 - "umount -f" would hang if any requests needed to be sent (393, 978)
4778 - avoid pinning large req buffer by copying for queued messages (989)
4779 - add "uuid" to "lctl device" command to help upcalls (991)
4780 - "open" RPCs with transnos would confuse recovery counters (1037)
4781 - do proper endian conversion of all wire messages (288, 340, 891)
4782 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4783 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4784 - fix ptlrpc_connection leak in target_handle_connect (1174)
4785 - fix import refcounting bug in OST and MDS cleanup (1134)
4786 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4787 - fix very unlikely obd_types race condition (501)
4788 - remove osc_open hack for echo_client (1187)
4789 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4790 - a failure in read_inode2 leads to deadlock (1139)
4791 - cancel ack-locks as soon as transaction is committed (1072)
4792 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4793 - make sure to commitrw after any preprw to avoid deadlock (1162)
4794 - failing to execute a file in a lustre FS would lock inode (1203)
4795 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4796 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4797 - obd_uuid2tgt would walk off the end of the list (1255)
4798 - on IA64 the osc would give portals incorrect bulk size (1258)
4799 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4800 - fix lock inversion caused by new llite matching code (1282)
4801 - limit the number of dirty pages on a client to 10MB (1286)
4802 - timed out locks were not being corrected cancelled (1289)
4803 - fix O_DIRECT above 4GB on IA-32 (1292)
4804 * major user-visible changes
4805 - fail out/fail over policy now controlled by the upcall (993)
4807 - add OBD_PING to check server availability and failure (954)
4808 - lustre messages are now sent in sending host order (288, 340, 891)
4809 - add eadatalen to MDS getattr reply (340)
4810 - OST read replies may contain second buffer, with per-page status (593)
4812 2003-03-11 Phil Schwan <phil@clusterfs.com>
4815 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4816 - failed open()s could cause deadlock; fixed (867, 869)
4817 - stop cancelling OST locks when files are closed (481)
4818 - overlapping XID spaces caused network corruption (851, 853)
4819 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4820 - setattr_raw added, to do single-RPC, server-side setattrs
4821 - lmc/lconf syntax change for OST UUIDs
4822 - fix crashy race condition between ptlrpc_free_req and osc_close
4823 - don't use request in mdc_enqueue if we hit a timeout (889)
4824 - don't set the inode i_size for regular files from the MDS (896)
4825 - handle out of order completion AST (842)
4826 - don't LBUG if a lock request times out after receiving AST (913)
4827 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4828 - if a bad lock AST arrives, send an error instead of dropping entirely
4829 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4830 - fix leak in bulk IO when only partially completed (899, 900, 926)
4831 - fix O_DIRECT for ia64 (55)
4832 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4833 - C-z after timeout could hang a process forever; fixed (977)
4835 - client-side I/O cache (678, 924, 929, 941, 970)
4837 - READPAGE and SETATTRs which don't take server-side locks get
4840 2003-02-11 Phil Schwan <phil@clusterfs.com>
4843 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4844 - Fix fsfilt_extN_readpage() to read a full page of directory
4845 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4846 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4847 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4848 - Most intent operations were removed, in favour of a new RPC mode
4849 that does a single RPC to the server and bypasses most of the VFS
4850 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4851 - Aggressively cancel local locks on DLM servers
4852 - mds_reint_unlink sends EA to the client if it's the last nlink.
4853 client uses that EA to unlink OST objects.
4854 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4855 - recursive symlinks were fixed (439)
4856 - fixed NULL deref in DEBUG_REQ
4857 - filter_update_lastobjid no longer calls sync, which annoyed extN
4858 - fixed multi-client small-writes to a single file problem (445)
4859 - fixed mtime updates during file writes (607)
4860 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4861 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4862 - fixed Linux OST size reporting problem (444, 656)
4863 - OST now updates object mtime with writes or setattr (607, 619)
4864 - client verifies file size before zeroing page past EOF (445)
4865 - OST now writes last allocated objid to disk with allocation (108)
4866 - LOV on echo now works (409)
4868 - mds_reint_unlink sends a new buffer, with the EA included. this
4869 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4870 set if a regular file was being unlinked, and it was the last link
4871 - use PtlGet from the target for bulk writes (315)
4872 - OST now updates object mtime with writes or setattr (607, 619)
4873 - LDLM now has a grant-time callback to revalidate locked items, if
4875 - Many MDS operations were reorganized to combat race conditions
4877 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4878 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4880 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4883 - Fully reactivate OST imports after reconnection (512, others)
4884 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4885 - More graceful error handling for truncating on dead OST (515)
4886 - Don't error out unless we're actually accessing dead stripes (474)
4887 - Fix garbage sizes when stripes are missing (410)
4888 - LRU counters were broken, causing constant lock purge (433, 432)
4889 - garbage on read from stripes with failed OSTs (441)
4890 - mark OSCs as active before reconnecting during recovery (438)
4891 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4892 - lfind did not preserve OST order in output (443)
4893 - symlinks cause hung clients, incorrect data (439)
4894 - stop dereferencing request after dropping refcount (457)
4895 - don't LASSERT(spin_is_locked) on non-SMP (455)
4896 - fixes for many rename() bugs
4897 - fstat didn't correctly synchronize attributes (399)
4898 - server must handle lock cancellation during blocking AST prep (487)
4899 - bulk descriptors were free()d too soon (511)
4900 - fix paths in lconf, which would load incorrect modules (451, 507)
4901 - fix confusing lconf 'host not found' error message (386)
4902 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4903 - fix race condition in mdc_blocking_ast() for inode access (526)
4904 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4905 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4906 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4907 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4908 - fix lstripe in several ways (kernel side) (527)
4909 - fix request leak in ldlm_cli_enqueue (262)
4910 - incorrect OSC was marked inactive after OST failure
4911 - call mds_fs_cleanup before unmounting filesystem (524)
4912 - fix races between taking ns_lock and ldlm_lock_change_resource
4913 - fix races updating LOV export open file list
4914 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4915 - fix recovery NULL deref in ldlm_cli_cancel_unused
4916 - fix some DLM races by using new hash table for lock handles (419)
4917 - permit the client to specify desired inodes, at replay
4918 - duplicate requests when we queue them for replay reintegration
4919 - fix last_rcvd offset calculation
4920 - sync after each recovered transaction, so we always make progress
4921 - never, not always, ERESTART requests without transnos
4922 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4924 - skip replay if the MDS says that the client is already connected
4925 - don't check for a recovery-enabled export to match lctl's UUID
4926 - don't INC_USE_COUNT for phantom exports
4927 - don't crash when cleaning up phantom exports (567)
4928 - don't double-finish or set replay data for errored mdc_open requests
4929 - abort requests when they time out, so we don't get old replies
4930 - send/receive replies for AST messages again
4931 - if the client says that it doesn't have the lock, cancel it on the
4933 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4934 mark it as destroyed, it will all work out when decref is called
4935 - fix module use counts (22, 581)
4937 - ASTs now expect a reply (server cancels lock on error reply)
4939 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4942 - fix many simultaneous client startup (392)
4943 - fix dentry->d_it clobbering
4944 - credentials weren't being shipped for readdir/getattr operations
4945 - remove invalid assertions triggered during some concurrent MD
4947 - proper Lustre versions added (336, 389)
4948 - fix memory leak for create error case (398)
4949 - fix LOV locking bug that would get cli/srv out of sync
4950 - fix echo client over LOV (409)
4951 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4952 - fix double-O_EXCL intent crash (424)
4953 - avoid sending multiple lock CANCELs (352)
4955 - MDS can do multi-client recovery (modulo bugs in new code)
4957 - many updates, edits, cleanups
4959 2002-11-18 Phil Schwan <phil@clusterfs.com>
4962 - fix null d_it dereference (346)
4963 - fix full OST/dbench hang (333)
4964 - fix permission problem with file removal (286)
4965 - fix removal of OSCs from LOV when they fail
4966 - fix NULL deref during bulk timeout (214)
4967 - fix problems related to multiple filesystems on one MDS (241)
4968 - fixed serious subtle metadata locking bugs
4969 - free locks on clients when inodes are removed due to memory
4971 - fix inode pointer in lock data (285)
4972 - partial support for multiple MDS on a single host (241)
4973 - data locks weren't cancelled at clear_inode time (290, 311)
4974 - intent locks could lead to unbounded lock growth (205)
4975 - added a maximum lock count, an LRU list, and a flusher
4976 - fix multiple rename (365)
4977 - properly abstracted the echo client
4978 - OSC locked 1 byte too many; fixed
4979 - rewrote brw callback code:
4980 - fixed recovery bugs related to LOVs (306)
4981 - fixed too-many-pages-in-one-write crash (191)
4982 - fixed (again) crash in sync_io_timeout (214)
4983 - probably fixed callback-related race (385)
4985 - Add capability to MDS protocol
4986 - LDLM cancellations and callbacks on different portals
4988 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4991 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4992 - defer open object destruction to close time (601981, 138)
4993 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4994 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4995 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4996 - don't call obd_setattr in ll_file_release for destroyed objects
4997 * protocol change to lustre_msg: move |version| and add |flags|
4998 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4999 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5000 * added replay of create, unlink, link and rename operations during
5001 MDS failover; recovery should be much more robust now
5002 * remove failed OSCs from LOVs (only lov_create uses this so far)
5003 * the lustre-HOWTO was brought (more) up to date (582544)
5005 2002-10-23 Phil Schwan <phil@clusterfs.com>
5008 - in-use dentries weren't being reused properly (617851)
5009 - prevent multiple LDLM setup (599178)
5010 - fix LOV size calculations for truncate (617853)
5011 - fix client handling of MDS intent errors (POSIX)
5012 - fix permission bug in lovstripe.c test (624321)
5013 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5014 - truncate past end of file could corrupt data
5015 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5016 - a race in recovery could return ETIMEDOUT to apps (623947)
5017 - building outside the source directory was fixed
5018 * the lustre-HOWTO was brought (more) up to date (582544)
5019 * major progress was made on recovery functionality
5021 2002-10-10 Phil Schwan <phil@clusterfs.com>
5024 - recovery deadlock fix
5025 - rm -rf causes LBUG fix (617817)
5026 - file open by multiple tasks fix (618962)
5027 - directory permissions bugs (602707 and 620007)
5028 - journal_stop fixed with locking (611313)
5029 - O_APPEND failures resolved (618273, perhaps 614459)
5030 - lconf PATH fix (619770)
5031 - IA64 build fix (621450)
5032 - RPC buffer sizes scale with amount of memory
5034 2002-10-01 Phil Schwan <phil@clusterfs.com>
5037 - locks would be cancelled without throwing away data pages,
5038 resulting in inconsistent data (605627)
5039 - inode attributes were not always being refreshed (605627, 612449)
5040 - lconf now continues to cleanup after lctl reports an error
5041 - MDS now enforces user permissions (602707)
5042 - lprocfs cleanup fixed, but not yet enabled (614157)
5043 - fixed infinite server hang, should a client not respond to an AST
5044 - avoid going into recovery if user calls readlink() with a buffer
5045 that's too small (613941)
5046 - AST RPCs no longer require replies (614867) -- this may be changed
5047 - don't crash server if client sends an IOV that's too big (611336)
5048 - fixed lock conversion deadlock (611892)
5049 - fixed the following of symlinks (614622)
5050 * recovery: the server can remove locks from a client that dies, other
5051 clients can make progress
5052 * more extN patch fixes
5053 * compile-time configurable ptlrpc buffer allocations
5055 - collaborative read cache document
5056 - Lustre Lite Performance CDR document-in-progress
5058 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5061 - fix typo in patch-2.4.18
5063 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5066 - clear ptlrpc request each time in handle_incoming_request()
5067 - unlink of files now destroys the object on the OST
5069 2002-09-19 Peter Braam <braam@clusterfs.com>
5071 * add hard link support
5072 * change obdfile creation method
5073 * kernel patch changed
5075 2002-09-19 Peter Braam <braam@clusterfs.com>
5078 - stack overflow bug in extN fixed
5080 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5082 * documentation updates
5083 - add man pages for config tools
5084 - update tests/README to describe testing with new config tools
5085 - finish metadata API descriptions
5086 * bug fixes and cleanups
5087 - statfs workaround for 16TB limit
5088 - LOV stripe allocation improved, can stripe on subset of OSTs
5089 - LOV file size/IO offset was wrong for files > 4GB in size
5090 - object EA data was being dropped, caused files to be unreadable
5091 - memory overflow with non-LOV OST caused memory corruption
5092 - fixed regression tests to work with new config tools, obdfilter
5093 - fixed bug when directory size became larger than 1 block
5094 - fixed bug (for single client case) when PWD was deleted
5095 - invalidate local directory pages when doing intent-based ops
5096 - avoid LDLM oops when lock callback contained bad data
5098 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5100 * documentation updates
5101 * bug fixes and cleanups
5102 - configuration tools
5105 - 64-bit compile warnings
5106 - 64-bit internal statfs data
5108 * test_brw on persistent OST devices
5110 * lprocfs (disabled)
5112 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5114 * documentation updates
5115 * bug fixes and cleanups
5116 * configuration tools
5118 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5121 * many small fixes to 0_5_4
5122 * io/network handling
5123 * thinkos in MDS operations
5125 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5128 * crucial basic fixes to 0.5.3
5129 * IOR, Iozone work over Elan
5132 2002-08-07 Phil Schwan <phil@clusterfs.com>
5133 * version 0_5_3, our first alpha
5134 * we use the new Portals iovs
5135 * documentation updates
5136 * bug fixes and cleanups
5137 * small changes in the DLM wire protocol
5139 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5140 * version 0_5_1 with some initial stability,
5141 * locking on MD and file I/O.
5142 * documentation updates
5143 * several bug fixes since 0.5.0
5144 * small changes in wire protocol
5146 2002-07-18 Phil Schwan <phil@clusterfs.com>
5148 * delivered as Lustre Light Alpha
5149 * fixed a crash after handling invalid MDS requests
5150 * fixed directory pages for architectures with non-4k pages sizes
5152 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5153 * release version v0_4_4
5154 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5156 * Fixes a number of interruption problems with OST operations.
5157 * Update documentation for portals header changes
5158 * Move all wire protocol structs/defines to lustre_idl.h
5159 * Fixes symlink length bug.
5160 * Add tcpdump to repository.
5162 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5163 * release version v0_4_3
5164 * Fixes statfs for inodes on extN.
5165 * Fixes bug in runtests which would delete /etc/hosts.
5166 * Use 64-bit object IDs wherever possible (not into VFS though)
5167 Remove ost_get_info, which is unused by lustre, and out of date.
5169 2002-07-03 Peter Braam <braam@clusterfs.com>
5170 * release version v0_4_2 Fixes a lookup error (type not passed)
5171 * move forward to head of Portals
5172 * move forward to latest Lustre kernel
5174 2002-06-25 Peter Braam <braam@clusterfs.com>
5175 * release version v0_4_1. Hopefully stable on single node use.