1 tbd Sun Microsystems, Inc.
5 2.6.9-67.0.1.EL (RHEL 4),
6 2.6.16.54-0.2.3 (SLES 10),
7 2.6.18-53.1.4.el5 (RHEL 5),
8 2.6.18.8 vanilla (kernel.org)
9 * Client support for unpatched kernels:
10 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
11 2.6.16 - 2.6.22 vanilla (kernel.org)
12 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
13 we do not recommend using patchless RHEL4 clients with kernels
14 prior to 2.6.9-55EL (RHEL4U5).
15 * Recommended e2fsprogs version: 1.40.4-cfs1
16 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
17 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
18 removed cwd "./" (refer to Bugzilla 14399).
21 Frequency : rare on shutdown ost
23 Description: Don't allow skipping OSTs if index has been specified.
24 Details : Don't allow skipping OSTs if index has been specified, make locking
25 in internal create lots better.
29 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
30 Details : access to bitfield in structure is always rounded to long
31 and this produce problem with not atomic change any bit.
35 Description: Lustre make rpms failed.
36 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
37 builds Lustre rpms from tarball.
39 Severity : enhancement
41 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
44 Frequency : rare on shutdown ost
46 Description: If llog cancel was not send before clean_exports phase, this can
47 produce deadlock in llog code.
48 Details : If llog thread has last reference to obd and call class_import_put
49 this produce deadlock because llog_cleanup_commit_master wait when
50 last llog_commit_thread exited, but this never success because was
51 called from llog_commit_thread.
55 Description: NULL lov_tgts causing MDS oops
56 Details : more safe checks for NULL lov_tgts for avoid oops.
58 Severity : enhancement
60 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
62 Severity : enhancement
64 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
68 Description: make mgs_setparam() handle fsname containing dash
69 Details : fsname containing a dash does not work with lctl conf_param
71 Severity : enhancement
73 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
77 Frequency : rare, in recovery and (or) destroy lovobjid file.
78 Description: rewrite lov objid code.
79 Details : Cleanup for lov objid code, remove scability problems and wrong
80 locking. Fix sending last_id into ost.
82 Severity : enhancement
84 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
86 Severity : enhancement
88 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
89 Details : Use d_move_locked instead of __d_move.
93 Frequency : rare, at shutdown
94 Description: access already free / zero obd_namespace.
95 Details : if client_disconnect_export was called without force flag set,
96 and exist connect request in flight, this can produce access to
97 NULL pointer (or already free pointer) when connect_interpret
98 store ocd flags in obd_namespace.
102 Frequency : only at startup
103 Description: not alloc memory with spinlock held.
104 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
105 if any spinlock held.
110 Description: lfs find does not continue on file error
111 Details : Continue other files processing when a file/dir is absent.
115 Description: Inconsistent usage of lustre_pack_reply()
116 Details : Standardize the usage of lustre_pack_reply() such that it
117 always generate a CERROR on failure.
120 Frequency : very rare
122 Description: Fix replay if there is an un-replied request and open
123 Details : In some cases, older replay request will revert the
124 mcd->mcd_last_xid on MDS which is used to record the client's
127 Severity : enhancement
129 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
131 Severity : enhancement
133 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
135 Severity : enhancement
137 Description: Update to SLES9 kernel-2.6.5-7.287.3.
139 Severity : enhancement
141 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
143 Severity : enhancement
145 Description: Build SLES10 patchless client fails
146 Details : The configure was broken by run ./configure with
147 --with-linux-obj=.... argument for patchless client. When the
148 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
149 can't search header adequately. Use absolute path such as
150 -I($LINUX)/include instead.
152 Severity : enhancement
154 Description: Lustre Page Accounting
155 Details : New macros for page alloc and free which enable accounting
156 of page allocation of Lustre. Use percpu counters to store memory
160 Frequency : only if debugging is disabled
162 Description: LASSERT_{REQ,REP}SWAB macros are buggy
163 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
164 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
165 these macros and replace them with their difinitions instead.
170 Description: interrupt oig_wait produce painc on resend.
171 Details : brw_redo_request can be used for resend requests from ptlrpcd and
172 private set, and this produce situation when rq_ptlrpcd_data not
173 copyed to new allocated request and triggered LBUG on assert
174 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
175 ptlrpcd set if request is changed and can be safety changed to use
178 Severity : enhancement
180 Description: organize the server-side client stats on per-nid basis
181 Details : Change the structure of stats under obdfilter and mds to
189 The "uuid"s file would list the uuids of _active_ exports.
190 And the clear entry is to clear all stats and stale nids.
195 Description: Processes looping in ll_readdir() on Lustre clients finally causing
196 a full node pseudo-hang
197 Details : Concurrent access to the same directory from multiple clients with
198 intensive file creation/removal can cause a client node to spin in
199 ll_readdir(). i_version must be increased every time the lock is
200 cancelled to ensure a revalidate is done.
205 Description: touch file failed when fs is not full
206 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
207 otherwise we can get ENOSP while fs is not full.
210 Frequency : only for Cray XT3
211 Bugzilla : 12829/13455
212 Description: Changing primary group doesn't change the group lustre assigns to
214 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
215 overridden with the primary group provided by the group upcall,
216 whereas the supplied fsgid can be trusted if it is in the list of
217 supplementary groups returned by the group upcall.
221 Description: Root Squash Functionality
222 Details : Implementation of NFS-like root squash capability. Specifically,
223 don't allow someone with root access on a client node to be able
224 to manipulate files owned by root on a server node.
228 Description: Slow trucate/writes to huge files at high offsets.
229 Details : Directly associate cached pages to lock that protect those pages,
230 this allows us to quickly find what pages to write and remove
231 once lock callback is received.
235 Description: Too many locks accumulating on client during NFS usage
236 Details : mds_open improperly used accmode to find out access mode to a
237 file. Also mdc_intent_lock logic to find out if we already have
238 lock similar to just received was flawed since introduction of
239 skiplists - locks are now added to the front of the granted
244 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
246 Details : In obd setup failure handler, obd_stopping will not necessarily to
247 be 1, and obd_set_up should also be checked to make sure whether
248 obd is completely setup.
250 Severity : enhancement
252 Description: Allow masking D_WARNING, D_ERROR messages from console
253 Details : Console messages can now be disabled via lnet.printk.
257 Description: User code with malformed file open parameter crashes client node
258 Details : Before packing join_file req, all the related reference should be
259 checked carefully in case some malformed flags cause fake
260 join_file req on client.
265 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
266 grain of quota doing harm to quota's accuracy
267 Details : qunit size will be changed when quota limitation is too low/high;
268 record the pending quota write in order to get more accureate
269 quota; delete the patch for bug12588, which is unnecessary when
270 this patch is landed. This bug also contains fixes for bug 14526,
271 14299, 14601 and 13794.
275 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
276 Details : ldlm_completion_ast() assumes that a lock is granted when the req
277 mode is equal to the granted mode. However, it should also check
278 that LDLM_FL_CP_REQD is not set.
282 Description: Heavy nfs access might result in deadlocks
283 Details : After ELC code landed, it is now improper to enqueue any mds
284 locks under och_sem, because enqueue might want to decide to
285 cancel open locks for same inode we are holding och_sem for.
289 Description: 35% write performance drop with ldiskfs2 when quotas are on
290 Details : Enable ext3 journalled quota by default to improve performance
291 when quotas are turned on.
295 Description: Client eviction while running blogbench
296 Details : A lot of unlink operations with concurrent I/O can lead to a
297 deadlock causing evictions. To address the problem, the number of
298 oustanding OST_DESTROY requests is now throttled to
299 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
300 callbacks are processed in priority.
302 --------------------------------------------------------------------------------
304 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
306 * Support for kernels:
307 2.6.5-7.286 (SLES 9),
308 2.6.9-55.0.9.EL (RHEL 4),
309 2.6.16.53-0.8 (SLES 10),
310 2.6.18-8.1.14.el5 (RHEL 5),
311 2.6.18.8 vanilla (kernel.org)
312 * Client support for unpatched kernels:
313 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
314 2.6.16 - 2.6.22 vanilla (kernel.org)
315 * Due to recently discovered recovery problems, we do not recommend
316 using patchless RHEL 4 clients with this or any earlier release.
317 * Recommended e2fsprogs version: 1.40.2-cfs1
318 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
321 Frequency : occasional
323 Description: MDS hang or stay in waiting lock
324 Details : If client receive lock with CBPENDING flag ldlm need send lock
325 cancel as separate rpc, to avoid situation when cancel request
326 can't processed due all i/o threads stay in wait lock.
329 Frequency : occasional
331 Description: Do not fail import if osc_interpret_create gets -EAGAIN
332 Details : If osc_interpret_create got -EAGAIN it immediately exits and
333 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
334 oscc_has_objects and see OSC has no objests and call
335 oscc_internal_create to resend create request.
337 Severity : enhancement
339 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
340 Details : Update which_patch & target file for SLES10 latest kernel.
342 Severity : enhancement
344 Description: add --type and --size parameters to lfs find
345 Details : Enhance lfs find by adding filetype and filesize parameters. Also
346 multiple OBDs can now be specified for the --obd option.
348 Severity : enhancement
350 Description: eliminate client locks in face of contention
351 Details : file contention detection and lockless i/o implementation
354 Severity : enhancement
356 Description: Remove client patches from SLES 10 kernel.
357 Details : This causes SLES 10 clients to behave as patchless clients
358 even on a Lustre-patched (server) kernel.
360 Severity : enhancement
362 Description: use i_size_read and i_size_write in 2.6 port
363 Details : replace inode->i_size access with i_size_read/write()
366 Frequency : when removing large files
368 Description: scheduling issue during removal of large Lustre files
369 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
370 It causes scheduling issues when removing large files (17TB in the
376 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
377 Details : Impossible (EPROTO is returned) to access a directory that has a
378 non-default striping and ACLs.
381 Frequency : only on ppc
383 Description: /proc/fs/lustre/devices broken on ppc
384 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
385 We should make sure the type of 'index' is loff_t and then cast
386 explicitly as needed below. Do not assign an explicitly cast
390 Frequency : only for rhel5
392 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
393 Details : Modify the target file & which_kernel.
396 Frequency : if the uninit_groups feature is enabled on ldiskfs
398 Description: e2fsck reports "invalid unused inodes count"
399 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
400 feature and only a single inode is created in a group then the
401 "bg_unused_inodes" count is incorrectly updated. Creating a
402 second inode in that group would update it correctly.
407 Description: buffer overruns could theoretically occur
408 Details : llapi_semantic_traverse() modifies the "path" argument by
409 appending values to the end of the origin string, and a buffer
410 overrun may occur. Adding buffer overrun check in liblustreapi.
412 Severity : enhancement
414 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
417 Frequency : only if filesystem is inconsistent
419 Description: handle "serious error: objid * already exists" more gracefully
420 Details : If LAST_ID value on disk is smaller than the objects existing in
421 the O/0/d* directories, it indicates disk corruption and causes an
422 LBUG(). If the object is 0-length, then we should use the existing
423 object. This will help to avoid a full fsck in most cases.
425 Severity : enhancement
427 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
428 Details : Modify vm-tunables-rhel4.patch.
430 Severity : enhancement
432 Description: Kernel config for 2.6.18-vanilla.
433 Details : Modify targets/2.6-vanilla.target.in.
434 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
435 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
436 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
437 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
440 Frequency : occasional
442 Description: improve handling recoverable errors
443 Details : If request processed with error which can be recoverable on server
444 request should be resend, otherwise page released from cache and
450 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
451 Details : Modify target file & which_patch.
452 A flaw was found in the IA32 system call emulation provided
453 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
454 value could be stored in the %RAX register, which could trigger an
455 out-of-bounds system call table access. An untrusted local user
456 could exploit this flaw to run code in the kernel
457 (ie a root privilege escalation). (CVE-2007-4573).
461 Description: change order of libsysio includes
462 Details : '#include sysio.h' should always come before '#include xtio.h'
464 Severity : enhancement
466 Description: adapt the lustre_config script to support the upgrade case
467 Details : Add "-u" option for lustre_config script to support upgrading 1.4
468 server targets to 1.6 in parallel.
473 Description: To avoid grant space > avaible space when the disk is almost
474 full. Without this patch you might see the error "grant XXXX >
475 available" or some LBUG about grant, when the disk is almost
477 Details : In filter_check_grant, for non_grant cache write, we should
478 check the left space by if (*left > ungranted + bytes), instead
479 of (*left > ungranted), because only we are sure the left space
480 is enough for another "bytes", then the ungrant space should be
481 increase. In client, we should update cl_avail_grant only there is
482 OBD_MD_FLGRANT in the reply.
487 Description: Update RHEL 4 kernel to fix local root privilege escalation.
488 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
489 described in CVE-2007-4573. This problem could allow untrusted
490 local users to gain root access.
493 Frequency : when using O_DIRECT and quotas
495 Description: Incorrect file ownership on O_DIRECT output files
496 Details : block usage reported by 'lfs quota' does not take into account
497 files that have been written with O_DIRECT.
499 --------------------------------------------------------------------------------
501 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
503 * Support for kernels:
504 2.6.5-7.286 (SLES 9),
505 2.6.9-55.0.2.EL (RHEL 4),
506 2.6.16.46-0.14 (SLES 10),
507 2.6.18-8.1.8.el5 (RHEL 5),
508 2.6.18.8 vanilla (kernel.org)
509 * Client support for unpatched kernels:
510 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
511 2.6.16 - 2.6.21 vanilla (kernel.org)
512 * Due to recently discovered recovery problems, we do not recommend
513 using patchless RHEL 4 clients with this or any earlier release.
514 * Recommended e2fsprogs version: 1.40.2-cfs1
515 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
519 Description: Fix errors in lfs documentation
520 Details : Fixes man pages
522 Severity : enhancement
524 Description: Adaptive timeouts
525 Details : RPC timeouts adapt to changing server load and network
526 conditions to reduce resend attempts and improve recovery time.
528 Severity : enhancement
530 Description: llapi_file_create() does not allow some changes
531 Details : add llapi_file_open() that allows specifying the file creation
532 mode and open flags, and also returns an open file handle.
534 Severity : enhancement
536 Description: df doesn't work properly if diskfs blocksize != 4K
537 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
540 Frequency : always on directories with default striping set
542 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
543 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
545 Severity : enhancement
547 Description: merge and cleanup kernel patches.
548 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
550 Severity : enhancement
552 Description: RedHat Update kernel for RHEL5
553 Details : Kernel config file for RHEL5.
555 Severity : enhancement
557 Description: OSS needs mutliple precreate threads
558 Details : Add ability to start more than one create thread per OSS.
561 Frequency : only with quota on the root user
563 Description: mds_obd_create error creating tmp object
564 Details : When the user sets quota on root, llog will be affected and can't
565 create files and write files.
568 Frequency : Always on ia64 patchless client, and possibly others.
570 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
571 Details : This allows the patchless client to be loaded on architectures
577 Description: disorder of journal start and llog_add cause deadlock.
578 Details : in llog_origin_connect, journal start should happen before
579 llog_add keep the same order as other functions to avoid
582 Severity : enhancement
584 Description: RedHat Update kernel for RHEL5
585 Details : Modify the kernel config file more closer RHEL5.
587 Severity : enhancement
589 Description: Build failure against Centos5 (RHEL5)
590 Details : Define PAGE_SIZE when it isn't present.
593 Frequency : occasionally when using NFS
595 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
596 Details : with NFS, the anon dentry's parent was set to itself in
597 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
598 corresponding dentry other than use rec->ur_name.
600 Severity : enhancement
602 Description: enable data checksumming by default
603 Details : enable checksum by default, allow --disable-checksum
604 configure option and "-o nochecksum" mount option. Checksums
605 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
608 Frequency : Occasionally with failover
610 Description: Client eviction due to failover config
611 Details : after a connection loss, the lustre client should attempt to
612 reconnect to the last active server first before trying the
613 other potential connections.
615 Severity : enhancement
617 Description: client-side metadata stat-ahead during readdir(directory readahead)
618 Details : perform client-side metadata stat-ahead when the client detects
619 readdir and sequential stat of dir entries therein
624 Description: when mds and osts use different quota unit(32bit and 64bit),
625 quota will be released repeatly.
626 Details : void sending multiple quota reqs to mds, which will keep the
627 status between the reqs.
630 Frequency : only with liblustre clients on XT3
632 Description: evictions taking too long
633 Details : allow llrd to evict clients directly on OSTs
637 Description: osts not allocated evenly to files
638 Details : change the condition to increase offset_idx
640 Severity : enhancement
642 Description: self-adjustable client's lru lists
643 Details : use adaptive algorithm for managing client cached locks lru
644 lists according to current server load, other client's work
645 pattern, memory activities, etc. Both, server and client
646 side namespaces provide number of proc tunables for controlling
651 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
652 Details : rewrite inherited code (uuid parsing code from ext2 utils and
653 readdir code from ext3) from scratch preserving functionality.
657 Description: Only those disconnect error should be returned by rq_status.
658 Details : In open/enqueue processs, Some errors, which will cause client
659 disconnected, should be returned by rq_status, while other
660 errors should still be returned by intent, then mdc or llite will
663 Severity : enhancement
665 Description: Tune the kernel for good SCSI performance.
666 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
667 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
671 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
672 Bugzilla : 13547, 13627
673 Description: Data corruption for OSTs that are formatted larger than 2TB
675 Details : When generating the bio request for lustre file writes the
676 sector number would overflow a temporary variable before being
677 used for the IO. The data reads correctly from Lustre (which
678 will overflow in a similar manner) but other file data or
679 filesystem metadata may be corrupted in some cases.
683 Description: "lfs find -obd UUID" prints directories
684 Details : "lfs find -obd UUID" will return all directory names instead
685 of just file names. It is incorrect because the directories
686 do not reside on the OSTs.
689 Bugzilla : 13596, 13614
690 Description: MDS hang after unclean shutdown of lots of clients
691 Details : Never resend AST requests.
695 Frequency : Always, for kernels after 2.6.16
696 Description: Fix warning idr_remove called for id=.. which is not allocated.
697 Details : Last kernels save old s_dev before kill super and not allow
698 to restore from callback - restore it before call kill_anon_super.
702 Description: Fix error on 'ls .' at the top of the Lustre mount.
703 Details : Don't revalidate dentry if it is a root dentry.
705 --------------------------------------------------------------------------------
707 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
709 * Support for kernels:
710 2.6.5-7.286 (SLES 9),
711 2.6.9-55.0.2.EL (RHEL 4),
712 2.6.16.46-0.14 (SLES 10),
713 2.6.18-8.1.8.el5 (RHEL 5),
714 2.6.18.8 vanilla (kernel.org)
715 * Client support for unpatched kernels:
716 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
717 2.6.16 - 2.6.22 vanilla (kernel.org)
718 * Due to recently discovered recovery problems, we do not recommend
719 using patchless RHEL 4 clients with this or any earlier release.
720 * Recommended e2fsprogs version: 1.39.cfs8
721 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
726 Description: block reactivating mgc import until all deactivates complete
727 Details : Fix race when failing back MDT/MGS to itself (testing)
729 Severity : enhancement
731 Description: lfs setstripe enhancement
732 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
735 Frequency : mds/oss recovery
737 Description: llog ctxt is refrenced after it has been freed.
738 Details : llog ctxt refcount was added to avoide the race between ctxt free
739 and llog recovery process. Each llog user must hold ctxt refcount
740 before it access the llog. And the llog ctxt can only be freed
741 when its refcount is zero.
743 Severity : enhancement
745 Description: randomly memory allocation failure util
746 Details : Make lustre randomly failed allocating memory for testing purpose.
749 Frequency : only for SLES10
751 Description: Update kernel patch for SLES10 SP1
752 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
754 Severity : enhancement
756 Description: omit set fsid for export NFS
757 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
758 as FS_REQUIRES_DEV for avoid problems with generate fsid.
761 Frequency : after network failures
763 Description: Add sync option to mount_lustre.c
764 Details : Client loses data written to lustre after a network interruption.
766 Severity : enhancement
768 Description: Error message improvement.
769 Details : Merging of two LCONSOLE_ERROR_MSG into one.
771 Severity : enhancement
773 Description: don't use GFP_* in generic Lustre code.
774 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
776 Severity : enhancement
778 Description: obdclass is limited by single OBD_ALLOC(idarray)
779 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
781 Severity : enhancement
783 Description: updated patchess for new RHEL4 kernel
784 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
786 Severity : enhancement
788 Description: warnings with build patchless client with vanila 2.6.19 and up
789 Details : change old ctl_table style and replace ctl_table/ctl_table_header
790 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
792 Severity : enhancement
794 Description: O_DIRECT bypasses client statistics.
795 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
796 read_ahead_stats were updated. Copied stats section from
797 osc_send_oap_rpc() into async_internal().
800 Frequency : only for Cray XT3
802 Description: peer credits not enough on many OST per OSS systems.
803 Details : Use new lnet way to add credits as we need those for pings and ASTs
806 Frequency : only with liblustre
808 Description: Liblustre is not releasing flock locks on file close.
809 Details : Release flock locks on file close.
812 Frequency : only for RHEL4
814 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
815 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
816 Update target file and kernel config.
820 Description: lustre support for RHEL5
821 Details : Add support for RHEL5.
826 Description: ASSERTION(export != NULL) failed in target_handle_connect
827 Details : Assetion hit is result of rare race between disconnect and connet
828 to same nid. target_handle_connect found old connect cockie and
829 tried to reconnect, but can't find export for this cockie.
834 Description: umount blocks forever on error
835 Details : In result of wrong using obd_no_recov and obd_force flags client
836 can hand if cancel or some other requests is lost.
839 Frequency : Only for SLES
841 Description: sanity_quota fail test_1
842 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
843 file, which makes TSTID[2] inunique.
845 Severity : enhancement
847 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
848 Details : Update target/ChangeLog/which_patch .
850 Severity : enhancement
852 Description: jbd statistics
853 Details : Port older jbd statistics patch for sles10
858 Description: lvbo_init failed for resource with missing objects.
859 Details : Fix returning error if we do stat for file with missing/corrupted
860 objects and i_size set to all sum of size of all avaible objects.
861 if we to truncate/write to missing object - it is recreated.
866 Description: Oops in read and write path when failing to allocate lock.
867 Details : Check if lock allocation failed and return error back.
870 Frequency : When flocks are used.
872 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
873 Details : Flock locks might destroy just granted lock if it could be merged
874 with another existing flock, this is done in completion handler,
875 so teach ldlm_cli_enquque_fini that this is a valid case for
881 Description: reply_lock_interpret crash due to race with it and lock cancel.
882 Details : Do not replay locks that are being cancelled. Do not reference
883 locks by their address during replay, just by their handle.
886 Frequency : only with deactivated OSTs
888 Description: lstripe command fails for valid OST index
889 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
890 instead of lov->desc.ld_active_tgt_count.
892 Severity : enhancement
894 Description: Build failure against Centos5 (RHEL5)
895 Details : Use getpagesize() instead of PAGE_SIZE.
897 --------------------------------------------------------------------------------
899 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
901 * Support for kernels:
902 2.6.5-7.283 (SLES 9),
903 2.6.9-55.EL (RHEL 4),
904 2.6.16.46-0.14 (SLES 10),
905 2.6.18.8 vanilla (kernel.org)
906 * Client support for unpatched kernels:
907 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
908 2.6.16 - 2.6.22 vanilla (kernel.org)
909 * Due to recently discovered recovery problems, we do not recommend
910 using patchless RHEL 4 clients with this or any earlier release.
911 * Recommended e2fsprogs version: 1.39.cfs8
912 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
913 * Starting with this release, the ldiskfs backing filesystem required
914 by Lustre is now in its own package, lustre-ldiskfs. This package
915 should be installed. It is versioned separately from Lustre and
916 may be released separately in future.
921 Description: server LBUG when shutting down
922 Details : Block umount forever until the mount refcount is zero rather
923 than giving up after an arbitrary timeout.
925 Severity : enhancement
927 Description: add optional extra BUILD_VERSION info
928 Details : add a new environment variable (namely LUSTRE_VERS) which allows
929 to override the lustre version.
932 Frequency : 2.6.18 servers only
934 Description: ll_kern_mount() doesn't release the module reference
935 Details : The ldiskfs module reference count never drops down to 0
936 because ll_kern_mount() doesn't release the module reference.
941 Description: server LBUG when using old ost_num_threads parameter
942 Details : Accept the old ost_num_threads parameter but warn that it
943 is deprecated, and fix an off-by-one error that caused an LBUG.
948 Description: Transient SCSI error results in persistent IO issue
949 Details : iobuf->dr_error is not reinitialized to 0 between two
953 Frequency : sometimes when underlying device returns I/O errors
955 Description: OSTs not going read-only during write failures
956 Details : OSTs are not remounted read-only when the journal commit threads
957 get I/O errors because fsfilt_ext3 calls journal_start/stop()
958 instead of the ext3 wrappers.
961 Frequency : only with 10000 clients or more
963 Description: poor connect scaling with increasing client count
964 Details : Don't run filter_grant_sanity_check for more than 100 exports
965 to improve scaling for large numbers of clients.
968 Frequency : SLES10 only
970 Description: sanity-quota.sh quotacheck failed: rc = -22
971 Details : Quotas cannot be enabled on SLES10.
974 Frequency : liblustre clients only
976 Description: getdirentries does not give error when run on compute nodes
977 Details : getdirentries does not fail when the size specified as an argument
978 is too small to contain at least one entry
980 Severity : enhancement
982 Description: Add LNET router traceability for debug purposes
983 Details : If a checksum failure occurs with a router as part of the
984 IO path, the NID of the last router that forwarded the bulk data
985 is printed so it can be identified.
990 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
991 Details : Due to a race condition, liblustre clients were occasionally
994 Severity : enhancement
996 Description: lfs setstripe use optional parameters instead of postional
999 Severity : enhancement
1001 Description: Nanosecond timestamp support for ldiskfs
1002 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1003 resolution timestamps. There is not yet support for this at
1004 the Lustre filesystem level.
1007 Frequency : during server recovery
1009 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1010 Details : request with rq_no_resend flag not awake l_wait_event if they get a
1014 Frequency : nfs export on patchless client
1016 Description: connectathon hang when test nfs export over patchless client
1017 Details : Disconnected dentry cannot be found with lookup, so we do not need
1018 to unhash it or make it invalid
1021 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1022 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1023 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1028 Description: open req refcounting wrong on reconnect
1029 Details : If reconnect happened between getting open reply from server and
1030 call to mdc_set_replay_data in ll_file_open, we will schedule
1031 replay for unreferenced request that we are about to free.
1032 Subsequent close will crash in variety of ways.
1033 Check that request is still eligible for replay in
1034 mdc_set_replay_data().
1039 Description: disable writes to filesystem when reading health_check file
1040 Details : the default for reading the health_check proc file has changed
1041 to NOT do a journal transaction and write to disk, because this
1042 can cause reads of the /proc file to hang and block HA state
1043 checking on a healthy but otherwise heavily loaded system. It
1044 is possible to return to the previous behaviour during configure
1045 with --enable-health-write.
1047 Severity : enhancement
1049 Description: 64-bit inode version
1050 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1051 to the inode. This will be required for version-based recovery.
1056 Description: MDS fails to start if a duplicate client export is detected
1057 Details : in some rare cases it was possible for a client to connect to
1058 an MDS multiple times. Upon recovery the MDS would detect this
1059 and fail during startup. Handle this more gracefully.
1061 Severity : enhancement
1063 Description: Add -o localflock option to simulate old noflock behaviour.
1064 Details : This will achieve local-only flock/fcntl locks coherentness.
1069 Description: log_commit_thread vs filter_destroy race leads to crash
1070 Details : Take import reference before releasing llog record semaphore
1075 Description: Wrong request locking in request set processing
1076 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1077 imp_delayed_list, in this place should be used imp_lock.
1080 Frequency : when reconnecting
1082 Description: Grant leak when OSC reconnect to OST
1083 Details : When osc reconnect ost, OST(filter) should check whether it
1084 should grant more space to client by comparing fed_grant and
1085 cl_avail_grant, and return the granted space to client instead
1086 of "new granted" space, because client will call osc_init_grant
1087 to update the client grant space info.
1090 Frequency : when client reconnects to OST
1092 Description: Grant leak when OSC does a resend and replays bulk write
1093 Details : When osc reconnect to OST, OST(filter) should clear grant info of
1094 bulk write request, because the grant info while be sync between
1095 OSC and OST when reconnect, and we should ignore the grant info
1096 these of resend/replay write req.
1101 Description: Grant space more than avaiable space sometimes.
1102 Details : When then OST is about to be full, if two bulk writing from
1103 different clients came to OST. Accord the avaliable space of the
1104 OST, the first req should be permitted, and the second one
1105 should be denied by ENOSPC. But if the seconde arrived before
1106 the first one is commited. The OST might wrongly permit second
1107 writing, which will cause grant space > avaiable space.
1110 Frequency : when client is evicted
1112 Description: Grant might be wrongly erased when osc is evicted by OST
1113 Details : when the import is evicted by server, it will fork another
1114 thread ptlrpc_invalidate_import_thread to invalidate the
1115 import, where the grant will be set to 0. While the original
1116 thread will update the grant it got when connecting. So if
1117 the former happened latter, the grant will be wrongly errased
1118 because of this race.
1123 Description: Checking Stale with correct fid
1124 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1125 to check whether it is stale, because sometimes, we want the
1126 enqueue happened anyway, and op_data.fid2 will not be initialized.
1128 Severity : enhancement
1130 Description: update patchless client
1131 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1134 Frequency : only with 2.4 kernel
1136 Description: random memory corruption
1137 Details : size of struct ll_inode_info is to big for union inode.u and this
1138 can be cause of random memory corruption.
1143 Description: Memory leak in recovery
1144 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1145 It should also check obd_fail before fsfilt_start, otherwise if
1146 fsfilt_start return -EROFS,(failover mds during mds recovery).
1147 then the req will return with repmsg->transno = 0 and rc = EROFS.
1148 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1149 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1150 be freed no matter whether fsfilt_commit success or not.
1153 Frequency : only with huge numbers of clients
1155 Description: Prevents from taking the superblock lock in llap_from_page for
1157 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1158 for llap_from_page call in ll_removepage() prevents from taking
1159 the superblock lock for a soon died page.
1164 Description: Not check open intent error before release open handle
1165 Details : in some rare cases, the open intent error is not checked before
1166 release open handle, which may cause
1167 ASSERTION(open_req->rq_transno != 0), because it tries to release
1168 the failed open handle.
1173 Description: Set cat log bitmap only after create log success.
1174 Details : in some rare cases, the cat log bitmap is set too early. and it
1175 should be set only after create log success.
1179 Description: Accessing a block bevice can re-enable I/O when Lustre is
1180 tearing down a device.
1181 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1186 Description: service threads may hog cpus when there are a lot of requests
1187 Details : Insert cond_resched to give other threads a chance to use some CPU
1192 Description: the cat log was not initialized in recovery
1193 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1194 unlink log on mds will not be initialized until mds post
1195 recovery. And also in mds post recovery, the unlink log will
1196 initialization will be done asynchronausly, so there will be race
1197 between add unlink log and unlink log initialization.
1201 Description: brw_stats were being printed incorrectly
1202 Details : brw_stats were being printed as log2 but all of them were not
1203 recorded as log2. Also remove some code duplication arising from
1204 filter_tally_{read,write}.
1208 Frequency : rare, only in recovery.
1209 Description: ASSERTION(req->rq_type != LI_POISON) failed
1210 Details : imp_lock should be held while iterating over imp_sending_list for
1211 prevent destroy request after get timeout in ptlrpc_queue_wait.
1215 Description: replay-single.sh test 52 fails
1216 Details : A lock's skiplist need to be cleanup when it being unlinked
1217 from its resource list.
1222 Description: Short directio read returns full requested size rather than
1224 Details : Direct I/O operations should return actual amount of bytes
1225 transferred rather than requested size.
1227 Severity : enhancement
1229 Description: metadata RPC reduction (e.g. for rm performance)
1230 Details : decrease the amount of synchronous RPC between clients and servers
1231 by canceling conflicing lock before the operation on the client
1232 and packing thier handles into the main operation RPC to server.
1234 Severity : enhancement
1236 Description: add #ifdef HAVE_KERNEL_CONFIG_H
1237 Details : kernels from 2.6.19 not need include linux/config.h, but add
1238 include linux/autoconf.h in commpiler command line.
1240 Severity : enhancement
1242 Description: patchless client support for 2.6.22 kernel
1243 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
1244 removed. In this case we need drop using os depended interface to
1245 kmem_cache and use cfs_mem_cache API.
1250 Description: fix mal-formatted messages
1251 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
1255 Frequency : always in liblustre
1256 Description: wrong IS_ERR implementation in liblustre.h
1257 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
1262 Description: Correct condition for output debug message.
1263 Details : inode i_nlink equal zero is not enough for output message about
1264 disk corruption, i_ctime and i_mode should be also checked.
1268 Frequency : always in patchless client
1269 Description: add configure check for truncate_complete_page
1270 Details : improve checks for exported symbols. This allow run check without
1271 sources, but with Module.symvers shipped with kernel distribution.
1272 add check for truncate_complete_page used by patchless client.
1276 Description: sanity.sh test_77h fails with "test_77h file compare failed"
1277 Details : test_77h uses a file which was messed by other test case.
1281 Description: Not Check whether lov_tgts is NULL in some lov functions
1282 Details : Checking whether lov_tgts is NULL in some functions.
1286 Description: replace obdo_alloc() with OBDO_ALLOC macro
1287 Details : nothing special is done in obdo_alloc() function, and for
1288 debugging purpose, it needs to be replaced with macros.
1292 Description: bad return value and errno from fcntl call
1293 Details : In liblustre API, errno should be a negative value if error
1298 Description: ptlrpc_check_set() LBUG
1299 Details : In case of positive reply from server and failed client bulk
1300 callback after bulk transfer shouldn't LBUG, but process this
1301 request as erroneous.
1303 Severity : enhancement
1305 Description: VFS operations stats tool.
1306 Details : Tool which collects stats by tracking value written in pid,
1307 ppid, gid and uses llstat to generate output to plot graph using
1309 Updated lustre/utils/Makefile.am
1310 Added lustre/utils/ltrack_stats.c
1312 Severity : enhancement
1314 Description: 2.6.18 server support (lustre 1.6.1)
1315 Details : Support for 2.6.18 kernels on the server side.
1320 Description: ASSERTION(imp->imp_conn_current) failed
1321 Details : an assertion failure is hit if a client node boots and attempts to
1322 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
1325 Frequency : only for i686
1327 Description: 1.4.11 RC1 build fails for RHEL 4, i686
1328 Details : Fixed config variable for build.
1333 Description: Updated patchess for new RHEL4 kernel
1334 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
1335 Updated patch jbd-stats-2.6.9.patch
1336 Updated patch qsnet-rhel4-2.6.patch
1337 Updated patch quota-deadlock-on-pagelock-core.patch
1338 Updated patch vfs_intent-2.6-rhel4.patch
1339 Updated patch vfs_races-2.6-rhel4.patch
1340 Updated series file 2.6-rhel4-titech.series
1341 Updated series file 2.6-rhel4.series
1342 Updated kernel config files
1347 Description: lquota slave complains LBUG when reconnecting with mds
1349 Details : quota slave depends qctxt->lqc_import to send its quota request.
1350 This pointer will be invalid if mds did failover or broke its
1351 connect to osts, which leads to LBUG.
1354 Frequency : when qunit size is too small(less than 20M)
1356 Description: write is stopped by improper -EDQUOT
1357 Details : If the master is busy and qunit size is small enough(let's say 1M),
1358 the slave can not get quota from master on time, which will lead
1359 slave to trigger a -EQUOTA to client.
1364 Description: Deadlock during metadata tests
1365 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
1366 called with the per-dentry lock held.
1369 Frequency : SLES9 only
1371 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
1372 Details : set CONFIG_NR_CPUS to 128 instead of 8.
1374 Severity : enhancement
1376 Description: remove fs_prep_san_write operation and related patches
1377 Details : remove the ext3-san-jdike patches which are no longer useful.
1382 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
1383 Details : a disk corruption can cause the mballoc code to assert on a
1384 double free or other extent corruptions. Handle these with
1385 ext3_error() instead of assertions.
1388 Frequency : only with mballoc3 code and deep extent trees
1390 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
1391 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
1392 can cause the filesystem to be remounted read-only.
1397 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
1398 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
1399 few external users of this ioctl number is lfsck's call to
1400 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
1401 Add the old ioctl number to the handler so both old and new
1406 Description: parallel lock callbacks
1407 Details : Instead of sending blocking and completion callbacks as separated
1408 requests, adding them to a set and sending in parallel.
1412 Description: Disable most debugging by default
1413 Details : To improve performance, disable most logging (for debug purposes)
1414 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
1415 default, and HA includes fewer messages.
1419 Frequency : only run on patchless client.
1420 Description: use do_facet on sanity.sh for test handling recoverables errors
1421 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
1425 Description: hash tables for lists of nids, connections and uuids
1426 Details : Hash tables noticeably help when a lot of clients connect to a
1427 server, to faster identify duplicate connections or reconnects,
1428 also to faster find export to evict in manual eviction case.
1432 Description: Sometimes, when the server evict a client, and the client will
1433 not be evicted as soon as possible.
1434 Details : In enqueue req, the error was returned by intent, instead of
1435 rq_status which make ptlrpc layer not detect this error, and
1436 does not evict the client. So enqueue error should be returned
1440 Frequency : only at startup
1442 Description: Delay client connections to MDT until fist MDT->OST connect
1443 Details : If a client tried to create a new file before the MDT had
1444 connected to any OSTs, the create would return EIO. Now
1445 the client will simply block until the MDT connects to the
1446 first OST and the create can succeed.
1449 Frequency : only for SLES9
1451 Description: Routinely utilize latest Quadrics drivers in CFS releases
1452 Details : Update patch qsnet-suse-2.6.patch.
1455 Frequency : only for sles10
1457 Description: Update patches for SLES 10 SP1 kernel.
1458 Details : Update the patch vfs_intent-2.6-sles10.patch.
1463 Description: Routinely utilize latest Quadrics drivers in CFS releases
1464 Details : Update patch qsnet-rhel4-2.6.patch.
1467 Frequency : at statup only
1469 Description: mds_lov_synchronize race leads to various problems
1470 Details : simultaneous MDT->OST connections at startup can cause the
1471 sync to abort, leaving the OSC in a bad state.
1476 Description: Using wrong pointer in osc_brw_prep_request
1477 Details : Access to array[-1] can produce panic if kernel compiled with
1478 CONFIG_PAGE_ALLOC enabled
1480 Severity : enhancement
1482 Description: Async OSC create to avoid the blocking unnecessarily.
1483 Details : If a OST has no remain object, system will block on the creating
1484 when need to create a new object on this OST. Now, ways use
1485 pre-created objects when available, instead of blocking on an
1486 empty osc while others are not empty. If we must block, we block
1487 for the shortest possible period of time.
1491 Frequency : only in recovery
1492 Description: Mark OST as early accessible if his start SYNC.
1493 Details : osc_precreate return flag early accessible if oscc marked as
1494 OSCC_FLAG_SYNC_IN_PROGRESS.
1499 Description: Sometimes precreate code can triger create object on wrong ost
1500 Details : Wrong protected or not not restored variables aftre precreate loop
1501 can produce creation object on wrong ost.
1504 Frequency : oss recovery
1506 Description: llog_commit_thread cleanup should sync with llog_commit_thread
1508 Details : llog_commit_thread_count should be synced between llog_commit
1509 start and cleanup, so new llog_commit thread should not be started
1510 when llog_commit threads being stopped to avoid accessing some
1513 Severity : enhancement
1515 Description: Add printing inode info into message about error in writepage.
1517 --------------------------------------------------------------------------------
1519 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1524 Frequency : on some architectures
1526 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1527 Details : Uninitialized flags sometimes cause configuration commands to
1531 Frequency : patchless clients only
1533 Description: missing __iget() symbol export
1534 Details : The __iget() symbol export is missing. To avoid the need for
1535 this on patchless clients the deathrow inode reaper is turned
1536 off, and we depend on the VM to clean up old inodes. This
1537 dependency was during via the fix for bug 12181.
1542 Description: sanity.sh fail: test_52b
1543 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
1544 incorrect inode's flags to client.
1546 --------------------------------------------------------------------------------
1548 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1550 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1551 INTEROPERATE with older versions automatically. In many cases a
1552 special upgrade step is needed. Please read the
1553 user documentation before upgrading any part of a 1.4.x system.
1554 * WARNING: Lustre configuration and startup changes are required with
1555 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1557 * Support for kernels:
1558 2.4.21-47.0.1.EL (RHEL 3),
1559 2.6.5-7.283 (SLES 9),
1560 2.6.9-42.0.10.EL (RHEL 4),
1561 2.6.12.6 vanilla (kernel.org),
1562 2.6.16.27-0.9 (SLES10)
1563 * Client support for unpatched kernels:
1564 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
1565 2.6.16 - 2.6.19 vanilla (kernel.org),
1566 2.6.9-42.0.8EL (RHEL 4)
1567 * Recommended e2fsprogs version: 1.39.cfs6
1568 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1571 Severity : enhancement
1573 Description: Async OSC create to avoid the blocking unnecessarily.
1574 Details : If a OST has no remain object, system will block on the creating
1575 when need to create a new object on this OST. Now, ways use
1576 pre-created objects when available, instead of blocking on an
1577 empty osc while others are not empty. If we must block, we block
1578 for the shortest possible period of time.
1580 Severity : enhancement
1582 Description: MountConf
1583 Details : Lustre configuration is now managed via mkfs and mount
1584 commands instead of lmc and lconf. New obd types (MGS, MGC)
1585 are added for dynamic configuration management. See
1586 https://mail.clusterfs.com/wikis/lustre/MountConf for
1589 Severity : enhancement
1591 Description: dynamic OST addition
1592 Details : OSTs can now be added to a live filesystem
1594 Severity : enhancement
1596 Description: startup order invariance
1597 Details : MDTs and OSTs can be started in any order. Clients only
1598 require the MDT to complete startup.
1600 Severity : enhancement
1602 Description: parallel, asynchronous orphan cleanup
1603 Details : orphan cleanup is now performed in separate threads for each
1604 OST, allowing parallel non-blocking operation.
1606 Severity : enhancement
1608 Description: optimized stripe assignment
1609 Details : stripe assignments are now made based on ost space available,
1610 ost previous usage, and OSS previous usage, in order to try
1611 to optimize storage space and networking resources.
1613 Severity : enhancement
1615 Description: Permanently set tunables
1616 Details : All writable /proc/fs/lustre tunables can now be permanently
1617 set on a per-server basis, at mkfs time or on a live
1620 Severity : enhancement
1622 Description: Lustre message v2
1623 Details : Add lustre message format v2.
1625 Severity : enhancement
1627 Description: client OST exclusion list
1628 Details : Clients can be started with a list of OSTs that should be
1629 declared "inactive" for known non-responsive OSTs.
1633 Description: ENOENT returned for valid filehandle during dbench.
1634 Details : Check if a directory has children when invalidating dentries
1635 associated with an inode during lock cancellation. This fixes
1636 an incorrect ENOENT sometimes seen for valid filehandles during
1637 testing with dbench.
1640 Frequency : SFS test only (otherwise harmless)
1642 Description: SPEC SFS validation failure on NFS v2 over lustre.
1643 Details : Changes the blocksize for regular files to be 2x RPC size,
1644 and not depend on stripe size.
1646 Severity : enhancement
1648 Description: fine-grained SMP locking inside DLM
1649 Details : Improve DLM performance on SMP systems by removing the single
1650 per-namespace lock and replace it with per-resource locks.
1652 Severity : enhancement
1654 Description: don't hold multiple extent locks at one time
1655 Details : To avoid client eviction during large writes, locks are not
1656 held on multiple stripes at one time or for very large writes.
1657 Otherwise, clients can block waiting for a lock on a failed OST
1658 while holding locks on other OSTs and be evicted.
1660 Severity : enhancement
1662 Description: Multiple MD RPCs in flight.
1663 Details : Further unserialise some read-only MDT RPCs - learn about intents.
1664 To avoid overly-overloading MDT, introduce a limit on number of
1665 MDT RPCs in flight for a single client and add /proc controls
1666 to adjust this limit.
1668 Severity : enhancement
1670 Description: client read/write statistics
1671 Details : Add client read/write call usage stats for performance
1672 analysis of user processes.
1673 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1674 file access. extents_stats shows chunk size distribution.
1675 extents_stats_per_process show chunk size distribution per
1678 Severity : enhancement
1680 Description: per-client statistics on server
1681 Details : Add ldlm and operations statistics for each client in
1682 /proc/fs/lustre/mds|obdfilter/*/exports/
1684 Severity : enhancement
1686 Description: improved MDT statistics
1687 Details : Add detailed MDT operations statistics in
1688 /proc/fs/lustre/mds/*/stats
1690 Severity : enhancement
1692 Description: VFS operations stats
1693 Details : Add client VFS call stats, trackable by pid, ppid, or gid
1694 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
1699 Description: Fix client-side osc byte counters
1700 Details : The osc read/write byte counters in
1701 /proc/fs/lustre/osc/*/stats are now working
1704 Frequency : always as root on SLES
1706 Description: Failure of copying files with lustre special EAs.
1707 Details : Client side always return success for setxattr call for lustre
1708 special xattr (currently only "trusted.lov").
1713 Description: Refcount LNET uuids
1714 Details : The global LNET uuid list grew linearly with every startup;
1715 refcount repeated list entries instead of always adding to
1718 Severity : enhancement
1720 Description: Dynamic service threads
1721 Details : Within a small range, start extra service threads
1722 automatically when the request queue builds up.
1725 Frequency : mixed-endian client/server environments
1727 Description: mixed-endian crashes
1728 Details : The new msg_v2 system had some failures in mixed-endian
1731 Severity : enhancement
1733 Description: Easy OST removal
1734 Details : OSTs can be permanently deactivated with e.g. 'lctl
1735 conf_param lustre-OST0001.osc.active=0'
1737 Severity : enhancement
1739 Description: MGS proc entries
1740 Details : Added basic proc entries for the MGS showing what filesystems
1743 Severity : enhancement
1745 Description: provide MGS failover
1746 Details : Added config lock reacquisition after MGS server failover.
1748 Severity : enhancement
1750 Description: add Linux 2.4 support
1751 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1755 Description: a large application tries to do I/O to the same resource and dies
1756 in the middle of it.
1757 Details : Check the req->rq_arrival time after the call to
1758 ost_brw_lock_get(), but before we do anything about
1759 processing it & sending the BULK transfer request. This
1760 should help move old stale pending locks off the queue as
1761 quickly as obd_timeout.
1764 Frequency : when an incorrect nid is specified during startup
1766 Description: ptlrpc connect to non-existant node causes kernel crash
1767 Details : LNET can't be re-entered from an event callback, which
1768 happened when we expire a message after the export has been
1769 cleaned up. Instead, hand the zombie cleanup off to another
1772 Severity : enhancement
1774 Description: plain/inodebits lock performance improvement
1775 Details : Grouping plain/inodebits in granted list by their request modes
1776 and bits policy, thus improving the performance of search through
1780 Frequency : only if OST filesystem is corrupted
1782 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1783 Details : for a short time RPCs with bulk IO are in the replay list,
1784 but replay of bulk IOs is unimplemented. If the OST filesystem
1785 is corrupted due to disk cache incoherency and then replay is
1786 started it is possible to trip an assertion. Avoid putting
1787 committed RPCs into the replay list at all to avoid this issue.
1790 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1792 Description: System hang on startup
1793 Details : This bug allowed the liblustre (e.g. catamount) client to
1794 return to the app before handling all startup RPCs. This
1795 could leave the node unresponsive to lustre network traffic
1796 and manifested as a server ptllnd timeout.
1798 Severity : enhancement
1800 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1801 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1802 boolean to control whether to print peer debug info when a
1803 client's RPC times out.
1806 Frequency : only for kernels with patches from Lustre below 1.4.3
1808 Description: Remove old rdonly API
1809 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1812 Frequency : only for devices with external journals
1814 Description: Set external device read-only also
1815 Details : During a commanded failover stop, we set the disk device
1816 read-only while the server shuts down. We now also set any
1817 external journal device read-only at the same time.
1820 Frequency : when upgrading from 1.4 while trying to change parameters
1822 Description: The wrong (new) MDC name was used when setting parameters for
1823 upgraded MDT's. Also allows changing of OSC (and MDC)
1824 parameters if --writeconf is specified at tunefs upgrade time.
1827 Frequency : when setting specific ost indicies
1829 Description: QOS code breaks on skipped indicies
1830 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1831 created with --index need not be sequential.
1833 Severity : enhancement
1835 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1836 Details : The uninit_groups feature works in conjunction with the kernel
1837 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1838 up the pass1 processing of e2fsck. This is a read-only feature
1839 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1840 mount filesystems that have had this feature enabled.
1842 Severity : enhancement
1844 Description: Improve multi-block allocation algorithm to avoid fragmentation
1845 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1846 allocation locality and avoid filesystem fragmentation.
1851 Description: lfs quota fails with deactivated OSTS
1852 Details : With this patch, three improvements are included:
1853 1. detete the softlimit in mds and osts when use "lfs quota".
1854 2. display the inaccurate data in the output of "lfs quota".
1855 3. try to get quota info when "lfs quota" is executed.
1857 ------------------------------------------------------------------------------
1859 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1861 * Support for kernels:
1862 2.4.21-47.0.1.EL (RHEL 3)
1863 2.6.5-7.283 (SLES 9)
1864 2.6.9-42.0.10.EL (RHEL 4)
1865 2.6.12.6 vanilla (kernel.org)
1866 2.6.16.27-0.9 (SLES 10)
1867 * Recommended e2fsprogs version: 1.39.cfs5
1869 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1873 Frequency : occasional, depends on client load and configuration
1874 Bugzilla : 12181, 12203
1875 Description: data loss for recently-modified files
1877 Details : In some cases it is possible that recently written or created
1878 files may not be written to disk in a timely manner (this should
1879 normally be within 30s unless client IO load is very high).
1880 The problem appears as zero-length files or files that are a
1881 multiple of 1MB in size after a client crash or client eviction
1882 that are missing data at the end of the file.
1884 This problem is more likely to be hit on clients where files are
1885 repeatedly created and unlinked in the same directory, clients
1886 have a large amount of RAM, have many CPUs, the filesystem has
1887 many OSTs, the clients are rebooted frequently, and/or the files
1888 are not accessed by other nodes after being written.
1890 The presence of the problem can be detected by looking at
1891 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1892 smaller than the second (nr_unused) then dirty files will not
1893 be flushed automatically to disk. "sync; sleep 10" should be
1894 run several times on the node before unmounting it to update
1895 Lustre (this is also safe to run on nodes without this problem).
1897 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1898 that can cause this same problem, so customers using that kernel
1899 also need to update the kernel in addition to Lustre. In order
1900 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1902 It is normal that files written just before a client crash (less
1903 than 30s) may not yet have been flushed to disk, even for local
1907 Frequency : frequent on thin XT3 nodes
1909 Description: UUID collision on thin XT3 Linux nodes
1910 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1911 randomly, since we relied on an insufficiently-seeded PRNG.
1916 Description: OSS hangs after "All ost request buffers busy"
1917 Details : A deadlock between quota and journal operations caused OSS
1918 hangs after printing "All ost request buffers busy."
1921 Frequency : always on liblustre builds
1923 Description: Cleanup compiler warnings on liblustre
1926 Frequency : always on liblustre builds on XT3
1928 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1932 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1933 > 32000 subdirectories
1934 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1935 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1936 subdirectory count crosses 32000. This will aid e2fsck to
1937 correctly handle more than 32000 subdirectories.
1940 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1942 Description: System hang on startup
1943 Details : This bug allowed the liblustre (e.g. catamount) client to
1944 return to the app before handling all startup RPCs. This
1945 could leave the node unresponsive to lustre network traffic
1946 and manifested as a server ptllnd timeout.
1948 Severity : enhancement
1950 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1951 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1952 boolean to control whether to print peer debug info when a
1953 client's RPC times out.
1958 Description: make O_SYNC working on 2.6 kernels
1959 Details : 2.6 kernels use different method for mark pages for write,
1960 so need add a code to lustre for O_SYNC work.
1965 Description: Failure to close file and release space on NFS
1966 Details : Put inode details into lock acquired in ll_intent_file_open.
1967 Use mdc_intent_lock in ll_intent_open to properly
1968 detect all kind of errors unhandled by mdc_enqueue.
1973 Description: proc file read during shutdown sometimes raced obd removal,
1975 Details : Add lock to prevent obd access after proc file removal.
1978 Frequency : Only for files larger than 4GB on 32-bit clients.
1980 Description: improperly doing page alignment of locks
1981 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1982 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1987 Description: RPCs being resent when they shouldn't be
1988 Details : Some RPCs that should not be resent are being resent. This
1989 can cause inconsistencies in the RPC state machine. Do not
1990 resend such requests.
1993 Frequency : rare, only with NFS export
1995 Description: Crash on NFS re-export node
1996 Details : under very unusual load conditions an assertion is hit in
1997 ll_intent_file_open()
2000 Frequency : only if OST filesystem is corrupted
2002 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2003 Details : for a short time RPCs with bulk IO are in the replay list,
2004 but replay of bulk IOs is unimplemented. If the OST filesystem
2005 is corrupted due to disk cache incoherency and then replay is
2006 started it is possible to trip an assertion. Avoid putting
2007 committed RPCs into the replay list at all to avoid this issue.
2012 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2013 Details : Large single O_DIRECT read and write calls can fail to allocate
2014 a sufficiently large buffer to process the request. In case of
2015 allocation failure the allocation is retried with a smaller
2016 buffer and broken into smaller requests.
2018 Severity : enhancement
2020 Description: Add -o localflock option to simulate old noflock behaviour.
2021 Details : This will achieve local-only flock/fcntl locks coherentness.
2026 Description: versioning check is incomplete
2027 Details : Checking the version difference of client vs. server, report
2028 error if the gap is too big.
2033 Description: add support PG_writeback bit
2034 Details : add support for PG_writeback bit for Lustre, for more carefull
2035 work with page cache in 2.6 kernel. This also fix some deadlocks
2036 and remove hack for work O_SYNC with 2.6 kernel.
2038 Severity : enhancement
2040 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2041 Details : The uninit_groups feature works in conjunction with the kernel
2042 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2043 up the pass1 processing of e2fsck. This is a read-only feature
2044 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2045 mount filesystems that have had this feature enabled.
2047 Severity : enhancement
2049 Description: Improve multi-block allocation algorithm to avoid fragmentation
2050 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2051 allocation locality and avoid filesystem fragmentation.
2053 ------------------------------------------------------------------------------
2055 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2057 * Support for kernels:
2058 2.6.9-42.0.3.EL (RHEL 4)
2059 2.6.5-7.276 (SLES 9)
2060 2.4.21-47.0.1.EL (RHEL 3)
2061 2.6.12.6 vanilla (kernel.org)
2062 2.6.16.21-0.8 (SLES10)
2063 * Recommended e2fsprogs version: 1.39.cfs2-0
2065 * The backwards-compatible /proc/sys/portals symlink has been removed
2066 in this release. Before upgrading, please ensure that you change
2067 any configuration scripts or /etc/sysctl.conf files that access
2068 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2069 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2070 in advance of the upgrade on any system running Lustre 1.4.6 or
2071 newer, since /proc/sys/lnet was added in that version.
2072 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2076 Frequency : only when quota is used
2078 Description: avoid scanning export list for quota master
2079 Details : Change the algorithms to avoid scanning export list in order
2080 to improve the efficiency.
2083 Frequency : MDS failover only, very rarely
2085 Description: "went back in time" messages on mds failover
2086 Details : The greatest transno may be lost when the current operation
2087 finishes with an error (transno==0) and the client's last_rcvd
2088 record is over-written. Save the greatest transno in the
2089 mds_last_transno for this case.
2092 Frequency : always for specific kernels and striping counts
2094 Description: client may get "Matching packet too big" without ACL support
2095 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2096 when trying to access files in certain configurations. The
2097 clients should in fact be denied when mounting because they do
2098 not understand ACLs.
2101 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2103 Description: many clients connecting with IO in progress causes connect timeouts
2104 Details : Avoid synchronous journal commits to avoid delays caused by many
2105 clients connecting/disconnecting when bulk IO is in progress.
2106 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2107 OST_IO_PORTAL to avoid delays behind potentially many pending
2111 Frequency : occasionally with multiple writers to a single file
2113 Description: shared writes to file may result in wrong size reported by stat()
2114 Details : Allow growing of kms when extent lock is cancelled
2117 Frequency : always with random mmap IO to multi-striped file
2119 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2120 Details : Only if the hole is at the end of OST object so that kms is too
2121 small. Fix is to increase kms accordingly in ll_nopage.
2124 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2126 Description: writes to a missing object would leak memory on the OST
2127 Details : If there is an inconsistency between the MDS and OST filesystems,
2128 such that the MDS references an object that doesn't exist, writes
2129 to that object will leak memory due to incorrect cleanup in the
2130 error handling path, eventually running out of memory on the OST.
2135 Description: Creating too long symlink causes lustre errors
2136 Details : Check symlink and name lengths before sending requests to MDS.
2139 Frequency : only if flock is enabled (not on by default)
2141 Description: posix locks not released on fd closure on 2.6.9+
2142 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2143 kernels, this caused such locks not to be released on fd close and
2144 then assertions on fs unmount about still used locks.
2147 Frequency : MDS failover only, very rarely
2149 Description: clients may get ASSERTION(granted_lock != NULL)
2150 Details : When request was taking a long time, and a client was resending
2151 a getattr by name lock request. The were multiple lock requests
2152 with the same client lock handle and
2153 mds_getattr_name->fixup_handle_for_resent_request found one of the
2154 lock handles but later failed with ASSERTION(granted_lock != NULL).
2159 Description: handle->h_buffer_credits > 0, assertion failure
2160 Details : h_buffer_credits is zero after truncate, causing assertion
2161 failure. This patch extends the transaction or creates a new
2165 Frequency : NFS re-export or patchless client
2166 Bugzilla : 11179, 10796
2167 Description: Crash on NFS re-export node (__d_move)
2168 Details : We do not want to hash the dentry if we don't have a lock.
2169 But if this dentry is later used in d_move, we'd hit uninitialised
2170 list head d_hash, so we just do this to init d_hash field but
2171 leave dentry unhashed.
2174 Frequency : NFS re-export or patchless client
2176 Description: NFS exports has problem with symbolic link
2177 Details : lustre client didn't properly install dentry when re-exported
2178 to NFS or running patchless client.
2181 Frequency : NFS re-export or patchless client
2183 Description: Various nfs/patchless fixes.
2184 Details : fixes reuse disconected alias for lookup process - this fixes
2185 warning "find_exported_dentry: npd != pd",
2186 fix permission error with open files at nfs.
2187 fix apply umask when do revalidate.
2190 Frequency : occasional
2192 Description: Crash on NFS re-export node
2193 Details : calling clear_page() on the wrong pointer triggered oops in
2194 generic_mapping_read().
2197 Frequency : rarely, using O_DIRECT IO
2199 Description: unaligned directio crashes client with LASSERT
2200 Details : check for unaligned buffers before trying any requests.
2203 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2205 Description: stale data returned from RAID cache
2206 Details : If only a small amount of IO is done to the RAID device before
2207 reading it again it is possible to get stale data from the RAID
2208 cache instead of reading it from disk.
2211 Frequency : always for sles10 kernel
2213 Description: sles10 support
2214 Details : ll_follow_link: compile fixes and using of nd_set_link
2215 under newer kernels.
2218 Frequency : depends on arch, kernel and compiler version, always on sles10
2221 Description: recursive or deep enough symlinks cause stack overflow
2222 Details : getting rid of large stack-allocated variable in
2226 Frequency : depends on hardware
2228 Description: lustre write performance loss in the SLES10 kernel
2229 Details : the performance loss is caused by using of write barriers in the
2230 ext3 code. The SLES10 kernel turns barrier support on by
2231 default. The fix is to undo that change for ldiskfs.
2233 ------------------------------------------------------------------------------
2235 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2237 * Support for kernels:
2238 2.6.9-42.0.3EL (RHEL 4)
2239 2.6.5-7.276 (SLES 9)
2240 2.4.21-47.0.1.EL (RHEL 3)
2241 2.6.12.6 vanilla (kernel.org)
2245 Frequency : quota enabled and large files being deleted
2247 Description: releasing more than 4GB of quota at once hangs OST
2248 Details : If a user deletes more than 4GB of files on a single OST it
2249 will cause the OST to spin in an infinite loop. Release
2250 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2255 Description: statfs data retrieved from /proc may be stale or zero
2256 Details : When reading per-device statfs data from /proc, in the
2257 {kbytes,files}_{total,free,avail} files, it may appear
2258 as zero or be out of date.
2261 Frequency : systems with MD RAID1 external journal devices
2263 Description: lconf's call to blkid is confused by RAID1 journal devices
2264 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2265 of returning all block devices that match the journal UUID.
2268 Frequency : always, for aggregate stripe size over 4GB
2270 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2271 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2272 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2275 Frequency : always if "lfs find" used on a local file/directory
2277 Description: "lfs find" segfaults if used on a local file/directory
2278 Details : The case where a directory component was not specified wasn't
2279 handled correctly. Handle this properly.
2282 Frequency : always on ppc64
2284 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2285 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2286 perform endianness conversion.
2289 Frequency : rarely (truncate to non-zero file size after write under load)
2290 Bugzilla : 10730, 10687
2291 Description: Files padded with zeros to next 4K multiple
2292 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2293 it is possible that files that are truncated to a non-zero size
2294 immediately after being written are filled with zero bytes beyond
2295 the truncated size. No file data is lost.
2297 Severity : enhancement
2299 Description: Allow recovery/failover for liblustre clients.
2300 Details : liblustre clients were unaware of failover configurations until
2303 Severity : enhancement
2305 Description: user file locks should fail when not mounting with flock option
2306 Details : Set up an error-returning stub in ll_file_operations.lock field
2307 to prevent incorrect behaviour when client is mounted without
2308 flock option. Also, set up properly f_op->flock field for
2312 Frequency : always on ia64
2314 Description: "lfs df" loops on printing out MDS statfs information
2315 Details : The obd_ioctl_data was not initialized and in some systems
2316 this caused a failure during the ioctl that did not return
2317 an error. Initialize the struct and return an error on failure.
2320 Frequency : SLES 9 only
2322 Description: Error of copying files with lustre special EAs as root
2323 Details : Client side always return success for setxattr call for lustre
2324 special xattr (currently only "trusted.lov").
2327 Frequency : rarely on clusters with both ia64+i386 clients
2329 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2330 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2331 clients are doing concurrent writes to the same file it is
2332 possible that the ia64 clients may LASSERT because the OST
2333 extent locks are not PAGE_SIZE aligned. Ensure that grown
2334 locks are always aligned on the request boundary.
2337 Frequency : specific use, occasional
2339 Description: Overwriting in use executable truncates on-disk binary image
2340 Details : If one node attempts to overwrite an executable in use by
2341 another node, we now correctly return ETXTBSY instead of
2342 truncating the file.
2344 Severity : enhancement
2346 Description: Async OSC create to avoid the blocking unnecessarily.
2347 Details : If a OST has no remain object, system will block on the creating
2348 when need to create a new object on this OST. Now, ways use
2349 pre-created objects when available, instead of blocking on an
2350 empty osc while others are not empty. If we must block, we block
2351 for the shortest possible period of time.
2356 Description: chmod on Lustre root is propagated to other clients
2357 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2358 invalid by the follow_mount time.
2361 Frequency : liblustre clients only
2363 Description: Race in 'instant cancel' lock handling could lead to such locks
2364 never to be granted in case of SMP MDS
2365 Details : Do not destroy not yet granted but cbpending locks in
2369 Frequency : replay/resend of open
2371 Description: non null lock assetion failure in mds_intent_policy
2372 Details : Trying to replay/resend lockless open requests resulted in
2373 mds_open() returning 0 with no lock. Now it sets a flag if
2374 a lock is going to be returned.
2376 Severity : enhancement
2378 Description: Checksum enhancements
2379 Details : New checksum enhancements allow for resending RPCs that failed
2382 Severity : enhancement
2384 Description: Tunables on number of dirty pages in cacche
2385 Details : Allow to set limit on number of dirty pages cached.
2390 Description: client crash on unmount - lock still has references
2391 Details : In some error handling cases it was possible to leak a lock
2392 reference on a client while accessing a file. This was not
2393 harmful to the client during operation, but would cause the
2394 client to crash when the filesystem is unmounted.
2397 Frequency : specific case, rare
2399 Description: ETXTBSY on mds though file not in use by client
2400 Details : ETXTBSY is no longer incorrectly returned when attempting to
2401 chmod or chown a directory that the user previously tried to
2402 execute or a currently-executing binary.
2405 Frequency : extremely rare except on liblustre-based clients
2407 Description: Lustre space not freed when files are deleted
2408 Details : Clean up open-unlinked files after client eviction. Previously
2409 the unlink was skipped and the files remained as orphans.
2414 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2415 Details : In some cases it was possible to send a blocking callback to
2416 a client doing a glimpse, even though that client didn't get
2417 a lock granted. When the glimpse lock is cancelled on the OST
2418 the freed lock is left on the waiting list and corrupted the list.
2421 Frequency : all core dumps
2423 Description: Broke core dumps to lustre
2424 Details : Negative dentry may be unhashed if parent does not have UPDATE
2425 lock, but some callers, e.g. do_coredump, expect dentry to be
2426 hashed after successful create, hash it in ll_create_it.
2428 ------------------------------------------------------------------------------
2430 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2432 * Support for kernels:
2433 2.6.9-42.0.2.EL (RHEL 4)
2434 2.6.5-7.276 (SLES 9)
2435 2.4.21-40.EL (RHEL 3)
2436 2.6.12.6 vanilla (kernel.org)
2440 Frequency : always on RHEL 3
2442 Description: Number of open files grows over time
2443 Details : The number of open files grows over time, whether or not
2444 Lustre is started. This was due to a filp leak introduced
2445 by one of our kernel patches.
2447 ------------------------------------------------------------------------------
2449 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
2451 * Support for kernels:
2452 2.6.9-42.EL (RHEL 4)
2453 2.6.5-7.267 (SLES 9)
2454 2.4.21-40.EL (RHEL 3)
2455 2.6.12.6 vanilla (kernel.org)
2460 Bugzilla : 5719, 9635, 9792, 9684
2461 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2462 Details : If a server is under heavy load and cannot reply to new
2463 connection requests before the client resends the (re)connect,
2464 the connection handling code can behave badly if two service
2465 threads are concurrently handing separate (re)connections from
2466 the same client. Add better locking to the connection handling
2467 code, and ensure that only a single connection will be processed
2468 for a given client UUID, even if the lock is dropped.
2470 Severity : enhancement
2472 Description: add TCP zero-copy support to kernel
2473 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2474 sends if the hardware supports scatter-gather and checksumming.
2475 This allows socklnd to do client-write and server-read more
2476 efficiently and reduce CPU utilization from skbuf copying.
2479 Frequency : only if NFS exporting from client
2481 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2482 Details : If setting attributes on a file created under NFS that had
2483 never been opened it would be possible to oops the client
2484 if the file had no objects.
2488 Bugzilla : 9326, 10402, 10897
2489 Description: client crash in ptlrpcd_wake() thread when sending async RPC
2490 Details : It is possible that ptlrpcd_wake() dereferences a freed async
2491 RPC. In rare cases the ptlrpcd thread alread processed the RPC
2492 before ptlrpcd_wake() was called and the request was freed.
2495 Frequency : always for liblustre
2497 Description: liblustre client does MDS+OSTs setattr RPC for each write
2498 Details : When doing a write from a liblustre client, the client
2499 incorrectly issued an RPC to the MDS and each OST the file was
2500 striped over in order to update the timestamps. When writing
2501 with small chunks and many clients this could overwhelm the MDS
2502 with RPCs. In all cases it would slow down the write because
2503 these RPCs are unnecessary.
2505 Severity : enhancement
2507 Description: allow number of MDS service threads to be changed at module load
2508 Details : It is now possible to change the number of MDS service threads
2509 running. Adding "options mds mds_num_threads={N}" to the MDS's
2510 /etc/modprobe.conf will set the number of threads for the next
2511 time Lustre is restarted (assuming the "mds" module is also
2512 reloaded at that time). The default number of threads will
2513 stay the same, 32 for most systems.
2518 Description: OST crash if filesystem is unformatted or corrupt
2519 Details : If an OST is started on a device that has never been formatted
2520 or if the filesystem is corrupt and cannot even mount then the
2521 error handling cleanup routines would dereference a NULL pointer.
2526 Description: NULL pointer deref in llap_from_page.
2527 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2528 incorrect) mapping if the page was truncated/reclaimed while it was
2529 searched for. Check for this condition and skip such pages when
2530 doing readahead. Introduce extra check to llap_from_page() to
2531 verify page->mapping->host is non-NULL (so page is not anonymous).
2534 Frequency : Sometimes when using sys_sendfile
2536 Description: "page not covered by a lock" warnings from ll_readpage
2537 Details : sendfile called ll_readpage without right page locks present.
2538 Now we introduced ll_file_sendfile that does necessary locking
2539 around call to generic_file_sendfile() much like we do in
2543 Frequency : with certain MDS communication failures at client mount time
2545 Description: NULL pointer deref after failed client mount
2546 Details : a client connection request may delayed by the network layer
2547 and not be sent until after the PTLRPC layer has timed out the
2548 request. If the client fails the mount immediately it will try
2549 to clean up before the network times out the request. Add a
2550 reference from the request import to the obd device and delay
2551 the cleanup until the network drops the request.
2554 Frequency : occasionally during client (re)connect
2556 Description: assertion failure during client (re)connect
2557 Details : processing a client connection request may be delayed by the
2558 client or server longer than the client connect timeout. This
2559 causes the client to resend the connection request. If the
2560 original connection request is replied in this interval, the
2561 client may trip an assertion failure in ptlrpc_connect_interpret()
2562 which thought it would be the only running connect process.
2565 Frequency : only with obd_echo servers and clients that are rebooted
2567 Description: kernel BUG accessing uninitialized data structure
2568 Details : When running an obd_echo server it did not start the ping_evictor
2569 thread, and when a client was evicted an uninitialized data
2570 structure was accessed. Start the ping_evictor in the RPC
2571 service startup instead of the OBD startup.
2573 Severity : enhancement
2574 Bugzilla : 10193 (patchless)
2575 Description: Remove dependency on various unexported kernel interfaces.
2576 Details : No longer need reparent_to_init, exit_mm, exit_files,
2577 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2580 Frequency : rare (only users of deprecated and unsupported LDAP config)
2582 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2583 Details : LDAP apparently contains 'lustreName' attributes instead of
2584 'name'. A simple remapping of the name is sufficient.
2587 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2589 Description: waiting_locks_callback trips kernel BUG if client is evicted
2590 Details : Running with the dump_on_timeout debug flag turned on makes
2591 it possible that the waiting_locks_callback() can try to dump
2592 the Lustre kernel debug logs from an interrupt handler. Defer
2593 this log dumping to the expired_lock_main() thread.
2595 Severity : enhancement
2597 Description: Support NFS exporting on 2.6 kernels.
2598 Details : Implement non-rawops metadata methods for NFS server to use without
2599 changing NFS server code.
2602 Frequency : very rare (synthetic metadata workload only)
2604 Description: two racing renames might cause an MDS thread to deadlock
2605 Details : Running the "racer" program may cause one MDS thread to rename
2606 a file from being the source of a rename to being the target of
2607 a rename at exactly the same time that another thread is doing
2608 so, and the second thread has already enqueued these locks after
2609 doing a lookup of the target and is trying to relock them in
2610 order. Ensure that we don't try to re-lock the same resource.
2613 Frequency : only very large systems with liblustre clients
2615 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2616 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2617 all OSTs in parallel. This allows the eviction of stale liblustre
2618 clients to proceed much faster than if they were done in series,
2619 and also offers similar improvements for other set_info RPCs.
2624 Description: excessive CPU usage during initial read phase on client
2625 Details : During the initial read phase on a client, it would agressively
2626 retry readahead on the file, consuming too much CPU and impacting
2627 performance (since 1.4.5.8). Improve the readahead algorithm
2628 to avoid this, and also improve some other common cases (read
2629 of small files in particular, where "small" is files smaller than
2630 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2635 Description: MDS crash when receiving packet with unknown intent.
2636 Details : Do not LBUG in unknown intent case, just return -EFAULT
2638 Severity : enhancement
2639 Bugzilla : 9293, 9385
2640 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2641 Details : Do not serialize getattr (non-intent version) and statfs.
2644 Frequency : occasional, when OST network is overloaded/intermittent
2646 Description: client evicted by OST after bulk IO timeout
2647 Details : If a client sends a bulk IO request (read or write) the OST
2648 may evict the client if it is unresposive to its data GET/PUT
2649 request. This is incorrect if the network is overloaded (takes
2650 too long to transfer the RPC data) or dropped the OST GET/PUT
2651 request. There is no need to evict the client at all, since
2652 the pinger and/or lock callbacks will handle this, and the
2653 client can restart the bulk request.
2656 Frequency : Always when mmapping file with no objects
2658 Description: client crashes when mmapping file with no objects
2659 Details : Check that we actually have objects in a file before doing any
2660 operations on objects in ll_vm_open, ll_vm_close and
2666 Description: Request leak when working with deleted CWD
2667 Details : Introduce advanced request refcount tracking for requests
2668 referenced from lustre intent.
2670 Severity : Enhancement
2672 Description: Cache open file handles on client.
2673 Details : MDS now will return special lock along with openhandle, if
2674 requested and client is allowed to hold openhandle, even if unused,
2675 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2676 closing files for every read/write openration.
2678 Severity : Enhancement
2680 Description: Cache open negative dentries on client when possible.
2681 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2682 negative dentries on lock revocation.
2687 Description: Remounting a client read-only wasn't possible with a zconf mount
2688 Details : It wasn't possible to remount a client read-only with llmount.
2690 Severity : enhancement
2691 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2692 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2693 MPICH is included to add Lustre-specific ADIO interfaces.
2694 This is based closely on the UFS ADIO layer and only differs
2695 in file creation, in order to allow the OST striping to be set.
2696 This is user-contributed code and not supported by CFS.
2701 Description: extended inode attributes (immutable, append-only) work improperly
2702 when 2.4 and 2.6 kernels are used on client/server or vice versa
2703 Details : Introduce kernel-independent values for these flags.
2705 Severity : enhancement
2708 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2709 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2710 granularity. Now, Fractional MB granularity is be supported,
2711 this is very useful for low memory system.
2713 Severity : enhancement
2715 Description: Getattr by fid
2716 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2717 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2718 have locks and avoids getattr rpc requests for it on every stat.
2721 Frequency : Always, for filesystems larger than 2TB
2723 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2724 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2725 filesystems larger than 2TB to be mounted with mballoc enabled.
2728 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2730 Description: filesystem corruption for non-standard kernels and very large OSTs
2731 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2732 filesystems larger than 2TB is mounted then the kernel will
2733 silently corrupt the start of the filesystem. CONFIG_LBD is
2734 enabled for all CFS-supported kernels, but the possibility of
2735 this happening with a modified kernel config exists.
2737 Severity : enhancement
2739 Description: add client O_DIRECT support for 2.6 kernels
2740 Details : It is now possible to do O_DIRECT reads and writes to files
2741 in the Lustre client mountpoint on 2.6 kernel clients.
2743 Severity : enhancement
2745 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2746 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2747 parallel, not waiting for response from every OST before sending
2748 a rpc to the next OST.
2753 Description: setattr vs write race when updating file timestamps
2754 Details : Client processes that update a file timestamp into the past
2755 right after writing to the file (e.g. tar) it is possible that
2756 the updated file modification time can be reset to the current
2757 time due to a race between processing the setattr and write RPC.
2759 Severity : enhancement
2761 Description: Bring 'lfs find' closer in line with regular Linux find.
2762 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2763 -print0 options and obtains all the needed info through the lustre
2766 Severity : enhancement
2768 Description: support up to 1024 configured devices on one node
2769 Details : change obd_dev array from statically allocated to dynamically
2770 allocated structs as they are first used to reduce memory usage
2775 Description: Flush dirty partially truncated pages during truncate
2776 Details : Immediatelly flush partially truncated pages in filter_setattr,
2777 this way we completely avoid having any pages in page cache on OST
2778 and can retire ugly workarounds during writes to flush such pages.
2783 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2784 Details : protect inode from truncation within vfs_unlink() context
2785 just take a reference before calling vfs_unlink() and release it
2786 when parent's i_sem is free.
2789 Frequency : always, if extents are used on OSTs
2791 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2792 Details : OSTs using the extents format would not zero the high 16 bits of
2793 the index physical block number. This is not a problem for any
2794 OST filesystems smaller than 16TB, and no kernels support ext3
2795 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2796 new/modified files) and can be fixed for existing filesystems
2797 with e2fsprogs-1.39-cfs1.
2802 Description: import connection selection may be incorrect if timer wraps
2803 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2804 import connections to be ignored if the 32-bit jiffies counter
2805 wraps. Use a 64-bit jiffies counter.
2808 Frequency : during server recovery
2810 Description: crash after server is denying duplicate export
2811 Details : If clients are resending connect requests to the server, the
2812 server refuses to allow a client to connect multiple times.
2813 Fixed a bug in the handling of this case.
2816 Frequency : very large clusters immediately after boot
2818 Description: LNET request buffers exhausted under heavy short-term load
2819 Details : If a large number of client requests are generated on a service
2820 that has previously never seen so many requests it is possible
2821 that the request buffer growth cannot keep up with the spike in
2822 demand. Instead of dropping incoming requests, they are held in
2823 the LND until the RPC service can accept more requests.
2826 Frequency : Sometimes during replay
2828 Description: Assertion failure in ll_local_open after replay.
2829 Details : If replay happened on an open request reply before we were able
2830 to set replay handler, reply will become not swabbed tripping the
2831 assertion in ll_local_open. Now we set the handler right after
2832 recognising of open request
2835 Frequency : very rare
2837 Description: kernel reports "badness in vsnprintf"
2838 Details : Reading from the "recovery_status" /proc file in small chunks
2839 may cause a negative length in lprocfs_obd_rd_recovery_status()
2840 call to vsnprintf() (which is otherwise harmless). Exit early
2841 if there is no more space in the output buffer.
2843 Severity : enhancement
2845 Description: clear OBD RPC statistics by writing to them
2846 Details : It is now possible to clear the OBD RPC statistics by writing
2847 to the "stats" file.
2852 Description: Client mtime is not the same on different clients after utimes
2853 Details : In some cases, the client was using the utimes() syscall on
2854 a file cached on another node. The clients now validate the
2855 ctime from the MDS + OSTs to determine which one is right.
2860 Description: Inability to activate failout mode
2861 Details : lconf script incorrectly assumed that in python string's numeric
2862 value is used in comparisons.
2865 Frequency : always with multiple stripes per file
2867 Description: Inefficient object allocation for mutli-stripe files
2868 Details : When selecting which OSTs to stripe files over, for files with
2869 a stripe count that divides evenly into the number of OSTs,
2870 the MDS is always picking the same starting OST for each file.
2871 Return the OST selection heuristic to the original design.
2876 Description: mount failures may take full timeout to return an error
2877 Details : Under some heavy load conditions it is possible that a
2878 failed mount can wait for the full obd_timeout interval,
2879 possibly several minutes, before reporting an error.
2880 Instead return an error as soon as the status is known.
2882 ------------------------------------------------------------------------------
2884 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
2886 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2887 INTEROPERATE with older versions automatically. Please read the
2888 user documentation before upgrading any part of a live system.
2889 * WARNING: Lustre networking configuration changes are required with
2890 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2893 * Support for kernels:
2894 2.6.9-22.0.2.EL (RHEL 4)
2895 2.6.5-7.244 (SLES 9)
2896 2.6.12.6 vanilla (kernel.org)
2899 Severity : enhancement
2900 Bugzilla : 7981/8208
2901 Description: Introduced Lustre Networking (LNET)
2902 Details : LNET is new networking infrastructure for Lustre, it includes
2903 a reorganized network configuration mode (see the user
2904 documentation for full details) as well as support for routing
2905 between different network fabrics. Lustre Networking Devices
2906 (LNDs) for the supported network fabrics have also been
2907 created for this new infrastructure.
2909 Severity : enhancement
2910 Description: Introduced Access control lists
2911 Details : clients can set ACLs on files and directories in order to have
2912 more fine-grained permissions than the standard Unix UGO+RWX.
2913 The MDS must be started with the "-o acl" mount option.
2915 Severity : enhancement
2916 Description: Introduced filesystem quotas
2917 Details : Administrators may now establish per-user quotas on the
2920 Severity : enhancement
2922 Description: Configuration change for the XT3
2923 The PTLLND is now used to run Lustre over Portals on the XT3
2924 The configure option(s) --with-cray-portals are no longer used.
2925 Rather --with-portals=<path-to-portals-includes> is used to
2926 enable building on the XT3. In addition to enable XT3 specific
2927 features the option --enable-cray-xt3 must be used.
2932 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2933 Details : A race between allocating a new llcd and re-getting the llcd_lock
2934 allowed another thread to grab newly-allocated llcd.
2936 Severity : enhancement
2938 Description: 2.6 OST async journal commit and locking fix to improve performance
2939 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2940 2.6 kernels are now async as they already were in 2.4 kernels so
2941 that they can commit concurrently with the network bulk transfer.
2942 For block-allocated files the filter allocation semaphore is held
2943 to avoid filesystem fragmentation during allocation. BKL lock
2944 removed for 2.6 xattr operations where it is no longer needed.
2949 Description: lconf incorrectly determined whether two IP networks could talk
2950 Details : In some more complicated routing and multiple-network
2951 configurations, lconf will avoid trying to make a network
2952 connection to a disjoint part of the IP space. It was doing the
2953 math incorrectly for one set of cases.
2958 Description: Fix for potential infinite loop processing records in an llog.
2959 Details : If an llog record is corrupted/zeroed, it is possible to loop
2960 forever in llog_process(). Validate the llog record length
2961 and skip the remainder of the block on error.
2964 Frequency : occasional (liblustre only)
2966 Description: liblustre could not open files whose last component is a symlink
2967 Details : sysio_path_walk() would incorrectly pass the open intent to
2968 intermediate path components.
2971 Frequency : rare (liblustre only with non-standard tuning)
2972 Bugzilla : 7201 (7350)
2973 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2974 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2975 after it has been released, only do this for referenced locks
2977 Severity : enhancement
2979 Description: specifying an (invalid) directory default stripe_size of -1
2980 would reset the directory default striping
2981 Details : stripe_size -1 was used internally to signal directory stripe
2982 removal, now use "all default" to signal dir stripe removal
2983 as a directory striping of "all default" is not useful
2986 Frequency : common for large clusters running liblustre clients
2988 Description: doing an ls when liblustre clients are running is slow
2989 Details : sending a glimpse AST to a liblustre client waits for every AST
2990 to time out, as liblustre clients will not respond. Since they
2991 cannot cache data we refresh the OST lock LVB from disk instead.
2993 Severity : enhancement
2995 Description: doing an ls at the same time as file IO can be slow
2996 Details : enqueue and other "small" requests can be blocked behind many
2997 large IO requests. Create a new OST IO portal for non-IO
2998 requests so they can be processed faster.
3001 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3003 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3004 created file with mode 000
3005 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3006 allow an ftruncate() as a non-root user to a file with mode 000.
3007 Lustre now respects this flag to disable mode checks when
3008 truncating a file owned by the user
3011 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3014 Description: Revoking locks from clients that went dead or catatonic might take
3016 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3017 cancellation of such locks instant on servers without waiting for
3018 any reply from clients. Clients drops these locks when cancel
3019 notification from server is received without replying.
3022 Frequency : liblustre-only, when liblustre client dies or becomes busy
3024 Description: Doing ls on Linux clients can take a long time with active
3026 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3027 granting such locks to it in the first place if possible. Locks
3028 are taken by proxy on the OST during the read or write and
3029 dropped immediately afterward. Add connect flags handling, do
3030 not grant locks to liblustre clients for glimpse ASTs.
3032 Severity : enhancement
3034 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3035 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3036 sequential reads and behaves badly with applications doing random
3037 reads. Improve it by reading ahead at least read region, and
3038 avoiding excessive large RPC for small reads.
3040 Severity : enhancement
3042 Description: Creating more than 1000 files for a single job may cause a load
3043 imbalance on the OSTs if there are also a large number of OSTs.
3044 Details : qos_prep_create() uses an OST index reseed value that is an
3045 even multiple of the number of available OSTs so that if the
3046 reseed happens in the middle of the object allocation it will
3047 still utilize the OSTs as uniformly as possible.
3052 Description: OST or MDS may oops in ping_evictor_main()
3053 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3054 but doesn't restart at beginning of obd_exports_timed list
3057 Severity : enhancement
3059 Description: improve by-nid export eviction on the MDS and OST
3060 Details : allow multiple exports with the same NID to be evicted at one
3061 time without re-searching the exports list.
3064 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3066 Description: MDS may oops in groups_free()
3067 Details : in rare race conditions a newly allocated group_info struct is
3068 freed again, and this can be NULL. The 2.4 compatibility code
3069 for groups_free() checked for a NULL pointer, but 2.6 did not.
3072 Frequency : common for liblustre clients doing little filesystem IO
3073 Bugzilla : 9352, 7313
3074 Description: server may evict liblustre clients accessing contended locks
3075 Details : if a client is granted a lock or receives a completion AST
3076 with a blocking AST already set it would not reply to the AST
3077 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3080 Frequency : lfs setstripe, only systems with more than 160 OSTs
3082 Description: unable to set striping with a starting offset beyond OST 160
3083 Details : llapi_create_file() incorrectly limited the starting stripe
3084 index to the maximum single-file stripe count.
3087 Frequency : LDAP users only
3089 Description: lconf did not handle in-kernel recovery with LDAP properly
3090 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3092 Severity : enhancement
3094 Description: bind OST threads to NUMA nodes to improve performance
3095 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3096 node and do their allocations there to localize memory access
3098 Severity : enhancement
3100 Description: llmount can determine client NID directly from Myrinet (GM)
3101 Details : the client NID code from gmnalnid was moved directly into
3102 llmount, removing the need to use this or specifying the
3103 client NID explicitly when mounting GM clients with zeroconf
3106 Frequency : if client is started with down MDS
3108 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3109 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3110 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3111 ptlrpc_import_delayed_req() because we didn't check if the
3112 request was interrupted, and we also didn't break out of the
3113 event loop if there was no timeout
3118 Description: data loss during non-page-aligned writes to a single file from
3119 both multiple nodes and multiple threads on one node at same time
3120 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3121 inconsistency led to false short-reads, that were cached and later
3122 used by ->prepare_write() to fill in partially written page,
3123 leading to data loss.
3126 Frequency : always, if lconf --abort_recovery used
3128 Description: lconf --abort_recovery fails with 'Operation not supported'
3129 Details : lconf was attempting to abort recovery on the MDT device and not
3132 Severity : enhancement
3134 Description: remove cleanup logs
3135 Details : replace lconf-generated cleanup logs with lustre internal
3136 cleanup routines. Eliminates the need for client-cleanup and
3139 Severity : enhancement
3141 Description: add support for EAs (user and system) on lustre filesystems
3142 Details : it is now possible to store extended attributes in the Lustre
3143 client filesystem, and with the user_xattr mount option it
3144 is possible to allow users to store EAs on their files also
3146 Severity : enhancement
3148 Description: Add possibility (config option) to show minimal available OST free
3150 Details : When compiled with --enable-mindf configure option, statfs(2)
3151 (and so, df) will return least minimal free space available from
3152 all OSTs as amount of free space on FS, instead of summary of
3153 free spaces of all OSTs.
3155 Severity : enhancement
3157 Description: do not expand extent locks acquired on OST-side
3158 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3159 by server: they are not cached anyway.
3162 Frequency : when mmap is used/binaries executed from Lustre
3164 Description: Unmmap pages before throwing them away from read cache.
3165 Details : llap_shrink cache now attempts to unmap pages before discarding
3166 them (if unmapping failed - do not discard). SLES9 kernel has
3167 extra checks that trigger if this unmapping is not done first.
3172 Description: lconf didn't resolve symlinks before checking to see whether a
3173 given mountpoint was already in use
3176 Frequency : when migrating failover services
3177 Bugzilla : 6395, 9514
3178 Description: When migrating a subset of services from a node (e.g. failback
3179 from a failover service node) the remaining services would
3180 time out and evict clients.
3181 Details : lconf --force (implied by --failover) sets the global obd_timeout
3182 to 5 seconds in order to quickly disconnect, but this caused
3183 other RPCs to time out too quickly. Do not change the global
3184 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3186 Severity : enhancement
3187 Frequency : if MDS is started with down OST
3188 Bugzilla : 9439,5706
3189 Description: Allow startup/shutdown of an MDS without depending on the
3190 availability of the OSTs.
3191 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3192 Add appropriate locking and lov-osc refcounts for safe
3193 cleaning. Add osc abort_inflight calls in case the
3194 synchronize never started.
3197 Frequency : occasional (Cray XT3 only)
3199 Description: root not authorized to access files in CRAY_PORTALS environment
3200 Details : The client process capabilities were not honoured on the MDS in
3201 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3202 been accessed by an authorized user then root was able to access
3203 the file on the local client also. The root user capabilities
3204 are now allowed on the MDS, as this environment has secure UID.
3207 Frequency : occasional
3209 Description: ldiskfs "too long searching" message happens too often
3210 Details : A debugging message (otherwise harmless) prints too often on
3211 the OST console. This has been reduced to only happen when
3212 there are fragmentation problems on the filesystem.
3217 Description: Division by zero in statfs when all OSCs are inactive
3218 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3219 lov_statfs divides by value returned by lov_get_stripecnt().
3223 Bugzilla : 9489, 3273
3224 Description: First write from each client to each OST was only 4kB in size,
3225 to initialize client writeback cache, which caused sub-optimal
3226 RPCs and poor layout on disk for the first writen file.
3227 Details : Clients now request an initial cache grant at (re)connect time
3228 and so that they can start streaming writes to the cache right
3229 away and always do full-sized RPCs if there is enough data.
3230 If the OST is rebooted the client also re-establishes its grant
3231 so that client cached writes will be honoured under the grant.
3236 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3237 Details : Now I/O RPCs go to different portal number and (presumably) fast
3238 lock requests (and glimses) and other RPCs get their own service
3239 threads pool that should be able to service those RPCs
3242 Severity : enhancement
3244 Description: Ability to exchange lustre version between client and servers and
3245 issue warnings at client side if client is too old. Also for
3246 liblustre clients there is ability to refuse connection of too old
3248 Details : New 'version' field is added to connect data structure that is
3249 filled with version info. That info is later checked by server and
3253 Frequency : rare, liblustre only.
3254 Bugzilla : 9296, 9581
3255 Description: Two simultaneous writes from liblustre at offset within same page
3256 might proceed at the same time overwriting eachother with stale
3258 Details : I/O lock withing llu_file_prwv was released too early, before data
3259 actually was hitting the wire. Extended lock-holding time until
3260 server acknowledges receiving data.
3263 Frequency : extremely rare. Never observed in practice.
3265 Description: avoid generating lustre_handle cookie of 0.
3266 Details : class_handle_hash() generates handle cookies by incrementing
3267 global counter, and can hit 0 occasionaly (this is unlikely, but
3268 not impossible, because initial value of cookie counter is
3269 selected randonly). Value of 0 is used as a sentinel meaning
3270 "unassigned handle" --- avoid it. Also coalesce two critical
3271 sections in this function into one.
3273 Severity : enhancement
3275 Description: allow liblustre clients to delegate truncate locking to OST
3276 Details : To avoid overhead of locking, liblustre client instructs OST to
3277 take extent lock in ost_punch() on client's behalf. New connection
3278 flag is added to handle backward compatibility.
3280 Severity : enhancement
3281 Bugzilla : 4928, 7341, 9758
3282 Description: allow number of OST service threads to be specified
3283 Details : a module parameter allows the number of OST service threads
3284 to be specified via "options ost ost_num_threads={N}" in the
3285 OSS's /etc/modules.conf or /etc/modprobe.conf.
3289 Bugzilla : 6146, 9635, 9895
3290 Description: servers crash with bad pointer in target_handle_connect()
3291 Details : In rare cases when a client is reconnecting it was possible that
3292 the connection request was the last reference for that export.
3293 We would temporarily drop the export reference and get a new
3294 one, but this may have been the last reference and the export
3295 was just destroyed. Get new reference before dropping old one.
3297 Severity : enhancement
3298 Frequency : if client is started with failover MDS
3300 Description: Allow multiple MDS hostnames in the mount command
3301 Details : Try to read the configuration from all specified MDS
3302 hostnames during a client mount in case the "primary"
3305 Severity : enhancement
3307 Description: Stop sending data to evicted clients as soon as possible.
3308 Details : Check if the client we are about to send or are sending data to
3309 was evicted already. (Check is done every second of waiting,
3310 for which l_wait_event interface was extended to allow checking
3311 of exit condition at specified intervals).
3314 Frequency : rare, normally only when NFS exporting is done from client
3316 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3318 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3321 Frequency : always, if srand() is called before liblustre initialization
3323 Description: Liblustre uses system PRNG disturbing its usage by user application
3324 Details : Introduce internal to lustre fast and high-quality PRNG for
3325 lustre usage and make liblustre and some other places in generic
3326 lustre code to use it.
3328 Severity : enhancement
3329 Bugzilla : 9477, 9557, 9870
3330 Description: Verify that the MDS configuration logs are updated when xml is
3331 Details : Check if the .xml configuration logs are newer than the config
3332 logs stored on the MDS and report an error if this is the case.
3333 Request --write-conf, or allow starting with --old_conf.
3335 Severity : enhancement
3337 Description: Handle symlinks in the path when checking if Lustre is mounted.
3338 Details : Resolve intermediate symlinks when checking if a client has
3339 mounted a filesystem to avoid duplicate client mounts.
3344 Description: lconf can hit an error exception but still return success.
3345 Details : The lconf command catches the Command error exception at the top
3346 level script context and will exit with the associated exit
3347 status, but doesn't ensure that this exit status is non-zero.
3352 Description: failure of ptlrpc thread startup can cause oops
3353 Details : Starting a ptlrpc service thread can fail if there are a large
3354 number of threads or the server memory is very fragmented.
3355 Handle this without oopsing.
3358 Frequency : always, only if liblustre and non-default acceptor port was used
3360 Description: liblustre cannot connect to servers with non-default acceptor port
3361 Details : tcpnal_set_default_params() was not called and was therefore
3362 ignoring the environment varaible TCPNAL_PORT, as well as other
3363 TCPNAL_ environment variables
3368 Description: two objects could be created on the same OST for a single file
3369 Details : If an OST is down, in some cases it was possible to create two
3370 objects on a single OST for a single file. No problems other
3371 than potential performance impact and spurious error messages.
3375 Bugzilla : 5681, 9562
3376 Description: Client may oops in ll_unhash_aliases
3377 Details : Client dcache may become inconsistent in race condition.
3378 In some cases "getcwd" can fail if the current directory is
3384 Description: Inode refcounting problems in NFS export code
3385 Details : link_raw functions used to call d_instantiate without obtaining
3386 extra inode reference first.
3390 Bugzilla : 9942, 9903
3391 Description: Referencing freed requests leading to crash, memleaks with NFS.
3392 Details : We used to require that call to ll_revalidate_it was always
3393 followed by ll_lookup_it. Also with revalidate_special() it is
3394 possible to call ll_revalidate_it() twice for the same dentry
3395 even if first occurence returned success. This fix changes semantic
3396 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3397 reference on a request referred from the intent.
3398 ll_intent_release() then releases such a request.
3401 Frequency : rare, normally benchmark loads only
3403 Description: unlinked inodes were kept in memory on the client
3404 Details : If a client is repeatedly creating and unlinking files it
3405 can accumulate a lot of stale inodes in the inode slab cache.
3406 If there is no other client load running this can cause the
3407 client node to run out of memory. Instead flush old inodes
3408 from client cache that have the same inode number as a new inode.
3411 Frequency : SLES9 2.6.5 kernel and long filenames only
3412 Bugzilla : 9969, 10379
3413 Description: utime reports stale NFS file handle
3414 Details : SLES9 uses out-of-dentry names in some cases, which confused
3415 the lustre dentry revalidation. Change it to always use the
3419 Frequency : rare, unless heavy write-truncate concurrency is continuous
3420 Bugzilla : 4180, 6984, 7171, 9963, 9331
3421 Description: OST becomes very slow and/or deadlocked during object unlink
3422 Details : filter_destroy() was holding onto the parent directory lock
3423 while truncating+unlinking objects. For very large objects this
3424 may block other threads for a long time and slow overall OST
3425 responsiveness. It may also be possible to get a lock ordering
3426 deadlock in this case, or run out of journal credits because of
3427 the combined truncate+unlink. Solution is to do object truncate
3428 first in one transaction without parent lock, and then do the
3429 final unlink in a new transaction with the parent lock. This
3430 reduces the lock hold time dramatically.
3433 Frequency : rare, 2.4 kernels only
3435 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3436 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3437 before unmounting their filesystem in order to avoid deadlock.
3438 The kernel_locked() function in 2.4 kernels only checks whether
3439 the kernel lock is held, not whether it is this process that is
3440 holding it as 2.6 kernels do.
3445 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3446 Details : The client ptlrpc code may be trying to reconnect to a down
3447 server before a previous connection attempt has timed out.
3448 Increase the reconnect interval to be longer than the connection
3449 timeout interval to avoid sending duplicate connections to
3453 Frequency : echo_client brw_test command
3455 Description: fix echo_client to work with OST preallocated code
3456 Details : OST preallocation code (5137) didn't take echo_client IO path
3457 into account: echo_client calls filter methods outside of any
3458 OST thread and, hence, there is no per-thread preallocated
3459 pages and buffers to use. Solution: hijack pga pages for IO. As
3460 a byproduct, this avoids unnecessary data copying.
3464 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3465 Description: Client can oops in mdc_commit_close() after open replay
3466 Details : It was possible for the MDS to return an open request with no
3467 transaction number in mds_finish_transno() if the client was
3468 evicted, but without actually returning an error. Clients
3469 would later try to replay that open and may trip an assertion
3470 Simplify the client close codepath, and always return an error
3471 from the MDS in case the open is not successful.
3474 Frequency : rare, 2.6 OSTs only
3476 Description: OST may deadlock under high load on fragmented files
3477 Details : If there was a heavy load and highly-fragmented OST filesystems
3478 it was possible to have all the OST threads deadlock waiting on
3479 allocation of biovecs, because the biovecs were not released
3480 until the entire RPC IO was completed. Instead, release biovecs
3481 as soon as they are complete to ensure forward IO progress.
3483 Severity : enhancement
3485 Description: Support for specifying external journal device at mount
3486 Details : If an OST or MDS device is formatted with an external journal
3487 device, this device major/minor is stored in the ext3 superblock
3488 and may not be valid for failover. Allow detecting and
3489 specifying the external journal at mount time.
3494 Description: Mounting an MDS with pending unlinked files may cause oops
3495 Details : target_finish_recovery() calls mds_postrecov() which returned
3496 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3497 considers this an error and immediately begins cleaning up the
3498 lov, just after starting the mds_lov process
3500 Severity : enhancement
3502 Description: Implement 'lfs df' to report actual free space on per-OST basis
3503 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3504 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3505 usage of objects; '-h' to report in human readable format.
3507 ------------------------------------------------------------------------------
3509 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
3516 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3517 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3518 at startup during orphan unlinking. Instead, skip buddy update
3519 before setup, as it will be regenerated after recovery is complete.
3524 Description: If an OST is inactive, its locks might reference stale inodes.
3525 Details : lov_change_cbdata() must iterate over all namespaces, even if
3526 they are inactive to clear inode references from the lock.
3528 Severity : enhancement
3529 Frequency : occasional, if non-standard max_dirty_mb used
3531 Description: Client will block write RPCs if not enough grant
3532 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3533 then the client will block writes while waiting for another RPC
3534 to complete instead of consuming its dirty limit. With change
3535 we get improved performance when max_dirty_mb is small.
3537 Severity : enhancement
3538 Bugzilla : 3389, 6253
3539 Description: Add support for supplementary groups on the MDS.
3540 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3541 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3542 lookups for user supplementary groups into a cache.
3546 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3547 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3548 for liblustre O_EXCL is enforced by MDS.
3552 Description: readdir never returns NULL in liblustre.
3553 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3554 dentry in struct dirent.
3558 Frequency : liblustre only, depends on application IO pattern
3559 Description: liblustre clients evicted if not contacting servers
3560 Details : Don't put liblustre clients into the ping_evictor list, so
3561 they will not be evicted by the pinger ever.
3563 Severity : enhancement
3565 Description: Add ability to evict clients by NID from MDS.
3566 Details : By echoing "nid:$NID" string into
3567 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3568 $NID would be instantly evicted from this MDS and from all active
3569 OSTs connected to it.
3573 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3574 Details : lookup_it_finish() used to query file size from OSTs that was not
3579 Description: service threads change working directory to that of init
3580 Details : Starting lustre service threads may pin the working directory
3581 of the parent thread, making that filesystem busy. Threads
3582 now change to the working directory of init to avoid this.
3586 Frequency : during shutdown only
3587 Description: shutdown with a failed MDS or OST can cause unmount to hang
3588 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3593 Frequency : 2.6 only
3594 Description: chmod/chown may include an extra supplementary group
3595 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3596 supplementary group and if none is specified this is used.
3599 Bugzilla : 5479 (6816)
3601 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3602 Details : If lookup is in progress on a file that is unlinked we might try
3603 to revalidate the inode and fail in revalidate after lookup is
3604 complete and ll_file_open() enqueues the open again but
3605 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3608 Frequency : always, if lconf --abort_recovery used
3610 Description: lconf --abort_recovery fails with 'Operation not supported'
3611 Details : lconf was attempting to abort recovery on the MDT device and not
3614 ------------------------------------------------------------------------------
3616 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3621 Frequency : rare (only unsupported configurations with a node running as an
3623 Bugzilla : 6514, 5137
3624 Description: Mounting a Lustre file system on a node running as an OST could
3626 Details : OSTs now preallocates memory needed to write out data at
3627 startup, instead of when needed, to avoid having to
3628 allocate memory in possibly low memory situations.
3629 Specifically, if the file system is mounted on on OST,
3630 memory pressure could force it to try to write out data,
3631 which it needed to allocate memory to do. Due to the low
3632 memory, it would be unable to do so and the node would
3633 become unresponsive.
3635 Severity : enhancement
3637 Description: Addition of lconf --service command line option
3638 Details : lconf now accepts a '--service <arg>' option, which is
3639 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3641 Severity : enhancement
3643 Description: Failover mode is now the default for OSTs.
3644 Details : By default, OSTs will now run in failover mode. To return to
3645 the old behaviour, add '--failout' to the lmc line for OSTs.
3647 Severity : enhancement
3649 Description: Health checks are now provided for MDS and OSTs
3650 Details : Additional detailed health check information on MSD and OSTs
3651 is now provided through the procfs health_check value.
3654 Frequency : occasional, depends on IO load
3656 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3657 after numerous create/delete cycles
3658 Details : The ext3 inode allocation policy would not allocate new inodes
3659 very well on the OSTs because there are no new directories
3660 being created. Instead we look for groups with free space if
3661 the parent directories are nearly full.
3666 Description: Network or server problems during mount may cause partially
3667 mounted clients instead of returning an error.
3668 Details : The config llog parsing code may overwrite the error return
3669 code during mount error handling, returning success instead
3675 Description: MDS can fail to allocate large reply buffers
3676 Details : After long uptimes the MDS can fail to allocate large reply
3677 buffers (e.g. zconf client mount config records) due to memory
3678 fragmentation or consumption by the buffer cache. Preallocate
3679 some large reply buffers so that these replies can be sent even
3680 under memory pressure.
3684 Frequency : rare (liblustre)
3685 Description: fsx running with liblustre complained that using truncate() to
3686 extend the file doesn't work. This patch corrects that issue.
3687 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3688 ATTR_SIZE and lsm use in llu_setattr_raw.
3692 Frequency : rare, only 2.6 kernels
3693 Description: Unusual file access patterns on the MDS may result in inode
3694 data being lost in very rare circumstances.
3695 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3696 code in the 2.6 kernel caused the inode and/or EA data not to
3697 be read from disk, causing single-file corruption.
3701 Frequency : rare, only 2.6 filesystems using extents
3702 Description: Heavy concurrent write and delete load may cause data corruption.
3703 Details : It was possible under high-load situations to have an extent
3704 metadata block in the block device cache from a just-unlinked
3705 file overwrite a newly-allocated data block. We now unmap any
3706 metadata buffers that alias just-allocated data blocks.
3710 Frequency : filesystems with default stripe_count larger than 77
3711 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3712 Details : lconf specifies an inode size of 4096 bytes when the default
3713 stripe_count is larger than 77. This conflicts with the default
3714 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3715 this case to avoid pinning too much memory for large EAs.
3717 ------------------------------------------------------------------------------
3719 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3724 Frequency : rare (extremely heavy IO load with hundreds of clients)
3726 Description: Client is evicted, gets IO error writing to file
3727 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3728 caused clients to be evicted for AST timeouts. The fixes in
3729 bug 5192 mean we no longer need to have such short AST timeouts
3730 so ldlm_timeout has been increased.
3733 Frequency : occasional during --force or --failover shutdown under load
3734 Bugzilla : 5949, 4834
3735 Description: Server oops/LBUG if stopped with --force or --failover under load
3736 Details : a collection of import/export refcount and cleanup ordering
3737 issues fixed for safer force cleanup
3740 Frequency : only filesystems larger than 120 OSTs
3741 Bugzilla : 5990, 6223
3742 Description: lfs getstripe would oops on a very large filesystem
3743 Details : lov_getconfig used kfree on vmalloc'd memory
3746 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3747 Bugzilla : 6242, 6243
3748 Description: reading from files that had been truncated to a non-zero size
3749 but never opened returned no data
3750 Details : ll_file_read() reads zeros from no-object files to EOF
3755 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3756 Details : There's little harm in aborting MDS/OSS recovery and letting it
3757 try again, so I removed the LASSERT and return an error instead.
3759 Severity : enhancement
3761 Description: New debugging infrastructure for tracking down data corruption
3762 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3763 (b) cover more of the client-side code path, and (c) try to narrow
3764 down where problems occurred
3768 Bugzilla : 3819, 4364, 4397, 6313
3769 Description: Racing close and eviction MDS could cause assertion in mds_close
3770 Details : It was possible to get multiple mfd references during close and
3771 client eviction, leading to one thread referencing a freed mfd.
3773 Severity: : enhancement
3774 Bugzilla : 3262, 6359
3775 Description: Attempts to reconnect to servers are now more aggressive.
3776 Details : This builds on the enhanced upcall-less recovery that was added
3777 in 1.4.2. When trying to reconnect to servers, clients will
3778 now try each server in the failover group every 10 seconds. By
3779 default, clients would previously try one server every 25 seconds.
3784 Description: After recovery, certain operations trigger a failed
3785 assertion on a client.
3786 Details : Failing over an mds, using lconf -d --failover, while a
3787 client was doing a readdir() call would cause the client to
3788 LBUG after recovery completed and the readdir() was resent.
3790 Severity : enhancement
3792 Description: Default groups are now added by lconf
3793 Details : You can now run lconf --group <servicename> without having to
3794 manually add groups with lmc.
3797 Frequency : occasional
3799 Description: Nodes with an elan id of 0 trigger a failed assertion
3802 Frequency : always when accessing e.g. tty/console device nodes
3804 Description: tty and some other devices nodes cannot be used on lustre
3805 Details : file's private_data field is used by device data and lustre
3806 values in there got lost. New field was added to struct file to
3807 store fs-specific private data.
3810 Frequency : when exporting Lustre via NFS
3812 Description: NFSD failed occasionally when looking up a path component
3813 Details : NFSD is looking up ".." which was broken in ext3 directories
3814 that had grown large enough to become hashed.
3817 Frequency : Clusters with multiple interfaces not on the same subnet
3819 Description: Nodes will repeatedly try to reconnect to an interface which it
3820 cannot reach and report an error to the log.
3821 Details : Extra peer list entries will be created by lconf with some peers
3822 unreachable. lconf now validates the peer before adding it.
3825 Frequency : Only if a default stripe is set on the filesystem root.
3827 Description: Setting a default stripe on the filesystem root prevented the
3828 filesystem from being remounted.
3829 Details : The client was sending extra request flags in the root getattr
3830 request and did not allocate a reply buffer for the dir EA.
3833 Frequency : occasional, higher if lots of files are accessed by one client
3834 Bugzilla : 6159, 6097
3835 Description: Client trips assertion regarding lsm mismatch/magic
3836 Details : While revalidating inodes the VFS looks up inodes with ifind()
3837 and in rare cases can find an inode that is being freed.
3838 The ll_test_inode() code will free the lsm during ifind()
3839 when it finds an existing inode and then the VFS later attaches
3840 this free lsm to a new inode.
3844 Bugzilla : 6422, 7030
3845 Description: MDS deadlock between mkdir and client eviction
3846 Details : Creating a new file via mkdir or mknod (starting a transaction
3847 and getting the ns lock) can deadlock with client eviction
3848 (gets ns lock and trying to finish a synchronous transaction).
3851 Frequency : occasional
3852 Description: While starting a server, the fsfilt_ext3 module could not be
3854 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3855 kernels. Previously, lconf would still use the ext3 name
3856 when trying to load modules. Now, it will correctly use
3857 ext3 on 2.4 and ldiskfs on 2.6.
3859 Severity : enhancement
3860 Description: The default stripe count has been changed to 1
3861 Details : The interpretation of the default stripe count (0, to lfs
3862 or lmc) has been changed to mean striping across a single
3863 OST, rather than all available. For general usage we have
3864 found a stripe count of 1 or 2 works best.
3866 Severity : enhancement
3867 Description: Add support for compiling against Cray portals.
3868 Details : Conditional compiling for some areas that are different
3872 Frequency : occasional
3873 Bugzilla : 6409, 6834
3874 Description: Creating files with an explicit stripe count may lead to
3875 a failed assertion on the MDS
3876 Details : If some OSTs are full or unavailable, creating files may
3877 trigger a failed assertion on the MDS. Now, Lustre will
3878 try to use other servers or return an error to the
3882 Frequency : occasional
3884 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3885 caused lock timeout during readahead (since 1.4.2).
3886 Details : Processes doing readahead might match a lock that hasn't been
3887 granted yet if there are overlapping and conflicting lock
3888 requests. The readahead process waits on ungranted lock
3889 (original lock is CBPENDING), while OST waits for that process
3890 to cancel CBPENDING read lock and eventually evicts client.
3892 Severity : enhancement
3894 Description: Initial enabling of flock support for clients
3895 Details : Implements fcntl advisory locking and file status functions.
3896 This feature is provided as an optional mount flag (default
3897 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3898 locking are implemented yet, and those that are are not guaranteed
3899 to be completely correct in production environments.
3900 mount -t lustre -o [flock|noflock] ...
3903 Frequency : occasional
3905 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3906 assertion in the ext3 JBD (journaling) layer.
3907 Details : The b_committed_data struct is protected by the big kernel lock
3908 in 2.4 kernels, serializing journal_commit_transaction() and
3909 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3910 access to this struct. In 2.6 kernels there is finer grained
3911 locking to improve SMP performance of the JBD layer.
3915 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3917 -----------------------------------------------------------------------------
3919 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3921 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3922 versions of Lustre. Please update all servers and clients to
3923 version 1.4.2 or later at the same time. You must also run
3924 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3925 to update the configuration logs.
3927 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3928 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3929 - swab llog records of type '0' so we get proper header size/idx (5861)
3930 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3931 - fix rename of one directory over another leaking an inode (5953)
3932 - avoid SetPageDirty on 2.6 (5981)
3933 - don't re-add just-being-destroyed locks to the waiting list (5653)
3934 - when creating new directories, inherit the parent's custom
3935 striping settings if present parent (3048)
3936 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3937 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3938 - don't hold client locks on temporary worklist from l_lru (5666)
3939 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3940 - automatically evict dead clients (5921)
3941 - Update file size properly in create+truncate+fstat case (6196)
3942 - Do not unhash mountpoint dentries, do not allow removal of
3944 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3945 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3946 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3947 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3948 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3949 - Correctly handle reads of files with no objects (6243)
3950 - lctl recover will also mark a device active if deactivate used (5933)
3952 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3953 - allow --write-conf on an MDS with different nettype than client (5619)
3954 - don't write config llogs to MDS for mounts not from that MDS (5617)
3955 - lconf should create multiple TCP connections from a client (5201)
3956 - init scripts are now turned off by default; run chkconfig --on
3957 lustre and chkconfig --on lustrefs to use them
3958 - upcalls are no longer needed for clients to recover to failover
3960 - add --abort-recovery option to lconf to abort recovery on device
3962 - add support for an arbitrary number of OSTs (3026)
3963 - Quota support protocol changes.
3964 - forward compatibility changes to wire structs (6007)
3965 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3966 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3967 - improved "lustre status" script
3968 - initialize blocksize for non-regular files (6062)
3969 - added --disable-server and --disable-client configure options (5782)
3970 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3971 - Vanilla 2.4.29 support
3972 - increase maximum number of obd devices to 520 (6242)
3973 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3974 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3975 SLES 9 2.6.5 kernels
3976 - limit stripes per file to 160 (the maximum EA size) (6093)
3978 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3981 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3982 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3983 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3984 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3985 - small ext3 extents cleanups and fixes (5733)
3986 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3987 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3988 - avoid needless client->OST connect, fix handle mismatch (5317)
3989 - fix DLM error path that led to out-of-sync client, long delays (5779)
3990 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3991 - fix several locking issues related to i_size (5492,5624,5654,5672)
3992 - don't move pending lock onto export if it is already evicted (5683)
3993 - fix kernel oops when creating .foo in unlinked directory (5548)
3994 - fix deadlock in obdfilter statistics vs. object create (5811)
3995 - use time_{before,after} to avoid timer jiffies wrap (5882)
3996 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3997 - Do not leak request if server was not able to process it (5154)
3998 - If mds_open unable to find parent dir, make that negative lookup(5154)
3999 - don't create new directories with extent-mapping (5909, 5936)
4001 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4002 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4003 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4004 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4005 - support for RHEL4 kernel on i686 (5773)
4006 - provide error messages when incompatible logs are encountered (5898)
4008 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4009 * version 1.4.0.10 (1.4.1 release candidate 1)
4011 - don't keep a lock reference when lock is not granted (4238)
4012 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4013 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4014 - drop import inflight refcount on signal_completed_replay error (5255)
4015 - unlock page after async write error during send (3677)
4016 - handle missing objects in filter_preprw_read properly (5265)
4017 - no transno return for symlink open, don't save no-trasno open (3440)
4018 - don't try to complete elan receive that already failed (4012)
4019 - free RPC server reply state on error (5406)
4020 - clean up thread from ptlrpc_start_thread() on error (5160)
4021 - readahead could read extra page into cache that wasn't ejected (5388)
4022 - prevent races in class_attach/setup/cleanup/detach (5260)
4023 - don't dereference de->d_inode after l_dput of de (5458)
4024 - use "int" for stripe value returned from lock_to_stripe (5544)
4025 - mballoc allocation and error-checking fixes in 2.6 (5504)
4026 - block device patches to fix I/O request sizes in 2.6 (5482)
4027 - look up hostnames for IB nals (5602)
4028 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4029 - don't start multiple acceptors for the same port (5277)
4030 - fix incorrect LASSERT in mds_getattr_name (5635)
4031 - export a proc file for general "ping" checking (5628)
4032 - fix "lfs check" to not block when the MDS is down (5628)
4034 - service request history (4965)
4035 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4036 - create an "evict_client" /proc entry on OSTs, like the MDS has
4037 - fix mount usage message, return errors per mount(8) (5168)
4038 - change grep [] to grep "[]" in tests so they work in more UMLs
4039 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4040 - remove ancient LOV_MAGIC_V0 EA support (5047)
4041 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4042 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4043 - allow readahead allocations to fail when low on memory (5383)
4044 - mmap locking landed again, after considerable improvement (2828)
4045 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4047 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4050 - send OST transaction number in read/write reply to free req (4966)
4051 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4052 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4053 - the watchdog thread now runs as interruptible (5246)
4054 - flock/lockf fixes (but it's still disabled, pending 5135)
4055 - don't use EXT3 constants in llite code (5094)
4056 - memory shortage at startup could cause assertion (5176)
4058 - reorganization of lov code
4059 - single portals codebase
4061 - add extents/mballoc support (5025)
4062 - direct I/O reads in the obdfilter (4048)
4063 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4065 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4068 - send OST transaction number in read/write reply to free req (4966)
4069 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4070 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4071 - the watchdog thread now runs as interruptible (5246)
4072 - handle missing objects in filter_preprw_read properly (5265)
4073 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4074 - drop import inflight refcount on signal_completed_replay error (5255)
4075 - unlock page after async write error during send (3677)
4076 - return original error code on reconstructed replies (3761)
4077 - no transno return for symlink open, don't save no-trasno open (3440)
4079 - add pid to ldlm debugging output (4922)
4080 - bump the watchdog timeouts -- we can't handle 30sec yet
4081 - extra debugging for orphan dentry/inode bug (5259)
4083 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4086 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4087 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4088 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4089 - fix NULL dereference in /proc/sys/portals/routes (4827)
4090 - allow failed mdc_close() operations to be interrupted (4561)
4091 - stop precreate on OST before MDS would time out on it (4778)
4092 - don't send partial-page writes before EOF from client (4410)
4093 - discard client grant for sub-page writes on large-page clients (4520)
4094 - don't free dentries not owned by NFS code, check generation (4806)
4095 - fix lsm leak if mds_create_objects() fails (4801)
4096 - limit debug_daemon file size, always print CERROR messages (4789)
4097 - use transno after validating reply (3892)
4098 - process timed out requests if import state changes (3754)
4099 - update mtime on OST during writes, return in glimpse (4829)
4100 - add mkfsoptions to LDAP (4679)
4101 - use ->max_readahead method instead of zapping global ra (5039)
4102 - don't interrupt __l_wait_event() during strace
4104 - add software watchdogs to catch hung threads quickly (4941)
4105 - make lustrefs init script start after nfs is mounted
4106 - fix CWARN/ERROR duplication (4930)
4107 - return async write errors to application if possible (2248)
4108 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4109 - print NAL number in %x format (4645)
4110 - update barely-supported suse-2.4.21-171 series (4842)
4111 - support for sles 9 %post scripts
4112 - support for building 2.6 kernel-source packages
4113 - support for sles km_* packages
4115 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4118 - ignore -ENOENT errors in osc_destroy (3639)
4119 - notify osc create thread that OSC is being cleaned up (4600)
4120 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4121 - reconstruct ost_handle() like mds_handle() (4657)
4122 - create a new thread to do import eviction to avoid deadlock (3969)
4123 - let lconf resolve symlinked-to devices (4629)
4124 - don't unlink "objects" from directory with default EA (4554)
4125 - hold socknal file ref over connect in case target is down (4394)
4126 - allow more than 32000 subdirectories in a single directory (3244)
4127 - fix blocks count for O_DIRECT writes (3751)
4128 - OST returns ENOSPC from object create when no space left (4539)
4129 - don't send truncate RPC if file size isn't changing (4410)
4130 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4131 - bind to privileged port in socknal and tcpnal (3689)
4133 - rate limit CERROR/CWARN console message to avoid overload (4519)
4134 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4135 - basic mmap support (3918)
4136 - kernel patch series update from b1_4 (4711)
4138 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4141 - avoid crash during MDS cleanup with OST shut down (2775)
4142 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4143 - don't use bad inodes on the MDS (3744)
4144 - dynamic object preallocation to improve recovery speed (4236)
4145 - don't hold spinlock over lock dumping or change debug flags (4401)
4146 - don't zero obd_dev when it is force cleaned (3651)
4147 - print grants to console if they go negative (4431)
4148 - "lctl deactivate" will stop automatic recovery attempts (3406)
4149 - look for existing locks in ldlm_handle_enqueue() (3764)
4150 - don't resolve lock handle twice in recovery avoiding race (4401)
4151 - revalidate should check working dir is a directory (4134)
4153 - don't always mark "slow" obdfilter messages as errors (4418)
4155 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4158 - don't close LustreDB during write_conf until it is done (3860)
4159 - fix typo in lconf for_each_profile (3821)
4160 - allow dumping logs from multiple threads at one time (3820)
4161 - don't allow multiple threads in OSC recovery (3812)
4162 - fix debug_size parameters (3864)
4163 - fix mds_postrecov to initialize import for llog ctxt (3121)
4164 - replace config semaphore with spinlock (3306)
4165 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4166 - don't allow enqueue to complete on a destroyed export (3822)
4167 - down write_lock before checking llog header bitmap (3825)
4168 - recover from lock replay timeout (3764)
4169 - up llog sem before sending rpc (3652)
4170 - reduce ns lock hold times when setting kms (3267)
4171 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4172 - fix NULL deref and obd_dev leak on setup error (3312)
4173 - replace some LBUG about llog ops with error handling (3841)
4174 - don't match INVALID dentries from d_lookup and spin (3784)
4175 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4176 - fix invalid assertion in ptlrpc_set_wait (3880)
4178 - add libwrap support for the TCP acceptor (3996)
4179 - add /proc/sys/portals/routes for non-root route listing (3994)
4180 - allow setting MDS UUID in .xml (2580)
4181 - print the stack of a process that LBUGs (4228)
4183 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4186 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4187 - make sure to unset replay flag from failed open requests (3440)
4188 - if default stripe count is 0, use OST count for inode size (3636)
4189 - update parent mtime/ctime on client for create/unlink (2611)
4190 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4191 - free recovery state on server during a forced cleanup (3571)
4192 - unregister_reply for resent reqs (3063)
4193 - loop back devices mounting and status check on 2.6 (3563)
4194 - fix resource-creation race that can provoke i_size == 0 (3513)
4195 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4196 - more debugging for page-accounting assertion (3746)
4197 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4198 - avoid deadlock after precreation failure (3758)
4199 - fix race and lock order deadlock in orphan handling (3450, 3750)
4200 - add validity checks when grabbing inodes from l_ast_data (3599)
4202 - add /proc/.../recovery_status to obdfilter (3428)
4203 - lightweight CDEBUG infrastructure, debug daemon (3668)
4204 - change default OSC RPC parameters to be better on small clusters
4205 - turn off OST read cache for files smaller than 32MB
4206 - install man pages and include them in rpms (3100)
4207 - add new init script for (un)mounting lustre filesystems (2593)
4208 - run chkconfig in %post for init scripts (3701)
4209 - drop scimac NAL (unmaintained)
4211 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4214 - clean kiobufs before and after use (3485)
4215 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4216 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4217 - change init script to fail more clearly if not run as root (1528)
4218 - allow clients to reconnect during replay (1742)
4219 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4220 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4221 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4222 - fix cond_resched (3554)
4223 - only evict unfinished clients after recovery (3515)
4224 - allow bulk resend, prevent data loss (3570)
4225 - dynamic ptlrpc request buffer allocation (2102)
4226 - don't allow unlinking open directory if it isn't empty (2904)
4227 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4228 - remove extraneous obd dereference causing LASSERT failure (3334)
4229 - don't use get_cycles() when creating temp. files on the mds (3156)
4230 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4231 - handle EEXIST for set-stripe, set proper directory name (3336)
4233 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4234 - fix ksocknal_fmb_callback() error messages (2918)
4236 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4239 - don't copy lvb into (possibly NULL) reply on error (2983)
4240 - don't deref dentry after dput, don't free lvb on error (2922)
4241 - use the kms to determine writeback rpc length (2947)
4242 - increment oti_logcookies when osc is inactive (2948)
4243 - update client's i_blocks count via lvb messages (2543)
4244 - handle intent open/close of special files properly (1557)
4245 - mount MDS with errors=remount-ro, like obdfilter (2009)
4246 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4247 - don't use cancelling-locks' kms values (2947)
4248 - use highest lock extent for kms, not last one (2925)
4249 - don't dereference ERR_PTR() dentry in error handling path (3107)
4250 - fix thread race in portals_debug_dumplog() (3122)
4251 - create lprocfs device entries at setup instead of at attach (1519)
4252 - common AST error handler, don't evict client on completion race (3145)
4253 - zero nameidata in detach_mnt in 2.6 (3118)
4254 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4255 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4256 - initialize RPC timeout timer earlier for 2.6 (3219)
4257 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4258 - print nal/nid for unknown nid (3258)
4259 - additional checks for oscc recovery before doing precreate (3284)
4260 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4261 - don't crash in mdc_close for bad permissions on open (3285)
4262 - zero i_rdev for non-device files (3147)
4263 - clear page->private before handing to FS, better assertion (3119)
4264 - tune the read pipeline (3236)
4265 - fix incorrect decref of invalidated dentry (2350)
4266 - provide read-ahead stats and refine rpc in flight stats (3328)
4267 - don't hold journal transaction open across create RPC (3313)
4268 - update atime on MDS at close time (3265)
4269 - close LDAP connection when recovering to avoid server load (3315)
4270 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4271 - don't leak open file on MDS after open resend (3325)
4272 - serialize filter_precreate and filter_destroy_precreated (3329)
4273 - loop device shouldn't call sync_dev() for nul device (3092)
4274 - clear page cache after eviction (2766)
4275 - resynchronize MDS->OST in background (2824)
4276 - refuse to mount the same filesystem twice on same mountpoint (3394)
4277 - allow llmount to create routes for mounting behind routers (3320)
4278 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4279 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4280 - fix rare problem with rename on htree directories (3417)
4282 - allow default OST striping configuration per directory (1414)
4283 - fix compilation for qswnal for 2.6 kernels (3125)
4284 - increase maximum number of MDS request buffers for large systems
4285 - change liblustreapi to be useful for external progs like lfsck (3098)
4286 - increase local configuration timeout for slow disks (3353)
4287 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4289 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4292 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4293 - don't overwrite extent policy data in reply if lock was blocked (2901)
4294 - drop filter export grants atomically with removal from device (2663)
4295 - del obd_self_export from work_list in class_disconnect_exports (2908)
4296 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4297 - swab reply message in mdc_close, other PPC fixes (2464)
4298 - fix destroying of named logs (2325)
4299 - overwrite old logs when running lconf --write_conf (2264)
4300 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4301 - fix race in target_handle_connect (2898)
4302 - mds_reint_create() should take same inode create lock (2926)
4303 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4304 - don't close files for self_export to avoid uninitialized obd (2936)
4305 - allow MDS with the same name as client node (2939)
4306 - hold dentry reference for closed log files for unlink (2325)
4307 - reserve space for all logs during transactions (2059)
4308 - don't evict page beyond end of stripe extent (2925)
4309 - don't oops on a deleted current working directory (2399)
4310 - handle hard links to targets without a parent properly (2517)
4311 - don't dereference NULL lock when racing during eviction (2867)
4312 - don't grow lock extents when lots of conflicting locks (2919)
4314 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4317 - account for cache space usage on clients to avoid data loss (974)
4318 - lfsck support in lustre kernel code (2349)
4319 - reduce journal credits needed for BRW writes (2370)
4320 - orphan handling to avoid losing space on client/server crashes
4321 - ptlrpcd can be blocked, stopping ALL progress (2477)
4322 - use lock value blocks to assist in proper KMS, faster stat (1021)
4323 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4324 - recovery for initial connections (2355)
4325 - fixes for mds_cleanup_orphans (1934)
4326 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4327 - block all file creations until orphan recovery completes (1901)
4328 - client remove rq_connection from request struct (2423)
4329 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4330 - recovery timer race (2670)
4331 - mdc_close recovey bug (2532)
4332 - ptlrpc cleanup bug (2710)
4333 - mds timeout on local locks (2588)
4334 - namespace lock held during RPCs (2431)
4335 - handle interrupted sync write properly (2503)
4336 - don't try to handle a message that hasn't been replied to (2699)
4337 - client assert failure during cleanup after abort recovery (2701)
4338 - leak mdc device after failed mount (2712)
4339 - ptlrpc_check_set allows timedout requests to complete (2714)
4340 - wait for inflight reqs when ptlrpcd finishes (2710)
4341 - make sure unregistered services are removed from the srv_list
4342 - reset bulk XID's when resending them (caught by 1138 test)
4343 - unregister_bulk after timeout
4344 - fix lconf error (2694)
4345 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4346 - readahead locks pages, leaves pending causing memory pressure (2673)
4347 - increase OST request buffers to 4096 on large machines (2729)
4348 - fix up permission of existing directories in simple_mkdir (2661)
4349 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4350 - don't assign transno to errored transactions (2742)
4351 - don't delete objects on OST if given a bogus objid from MDS (2751)
4352 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4353 - if rq_no_resend, then timeout request after recovery (2432)
4354 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4355 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4356 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4357 - MDS recovery shouldn't delete objects that it has given out (2730)
4358 - if enqueue arrives after completion, don't clobber LVB (2819)
4359 - don't unlock pages twice when trigger_group_io returns error (2814)
4360 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4361 - don't write pages to disk if there was an error (1450)
4362 - don't ping imports that have recovery disabled (2676)
4363 - take buffered bytes into account when balancing socknal conn (2817)
4364 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4365 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4366 - remove little-endian swabbing of llog records (1987)
4367 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4368 - retry reposting request buffers if they fail (1191)
4369 - grow extent at grant time to avoid granting a revoked lock (2809)
4370 - lock revoke doesn't evict page if covered by a second lock (2765)
4371 - disable VM readahead to avoid reading outside lock extents (2805)
4373 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4374 - updated kernel patches for hp-2.4.20 kernel (2681)
4376 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4379 - fix truncated write corruption (2366)
4380 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4382 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4383 - ensure len(uuid) < 37 in lmc (1171)
4384 - fix ia64 OOPS in llog_test (2255)
4385 - zero end of page at obdfilter for partial page writes (2648)
4386 - don't leave stale dentries around after renames (bug 2428)
4387 - fix timeouts when evicting a client with a single lock held (2642)
4388 - set deadline for the initial HELLO message to drain (2634)
4389 - print out dotted-quad IP addresses in the socknal (2302)
4391 - additional debugging for MDS client eviction problem (2443)
4392 - fix mkfsoptions support for osts (2603, 2604)
4394 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4397 - add series for the vanilla 2.6.0 kernel
4398 - add series for the vanilla 2.4.24 kernel
4399 - add series for a cray x86/64 UL kernel drop
4400 - fix xattr patches for the vanilla 2.4.19 series
4402 - generate true UUIDs in lmc (1171)
4403 - have portals stack dumping break in UML (2466)
4404 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4405 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4406 - wake callback waiting threads on client eviction (2460)
4407 - Add --ptldebug and --subsystem to lmc (1719)
4408 - update assertion to allow safe interrupt allocation
4409 - set rq_no_resend for cancel requests (2432)
4410 - recalculate ptlrpcd timeout after resend (2494)
4411 - call vfs_rmdir when removing pending directories (2368)
4412 - fix renaming a file to itself (2429)
4413 - lmc creates a default one-stripe lov (2454)
4414 - expand procfs space to handle large clusters (2326)
4415 - increase UML stack to avoid overflow
4416 - update lconf's list of debug and subsystem masks
4417 - fix lfs find --obd (2510)
4418 - /proc tunable for disabling filter read caching (2591)
4419 - stop rpm packages from altering slapd.conf (2301)
4420 - disable nagle in the socknal under 0conf (2578)
4421 - choose mds inode size based on stripe count (2572)
4422 - fix kernel-source rpm problems (2516)
4424 - add --disable-doc to avoid pdf generation (2421)
4425 - update documentation, tests, type-os, comments
4426 - avoid format warnings on ia64
4427 - remove the TOE NAL
4428 - tiny code cleanups by removing unused fields
4430 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4433 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4434 - remove the most blatant lies from BUILDING file (2371)
4435 - change default debug level to reasonable production setting
4436 - reduce client side cache size to reduce cache flush time
4437 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4438 - make TCP zerocopy and pinger support enabled by default (2476)
4439 - sync writes completed after process exits caused crashes (2319)
4440 - maintain correct mount count on the MDS (2356)
4441 - backout 1557, because 2316 wasn't really fixed
4442 - better file I/O statistics gathering in /proc
4443 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4444 - another kernel patch to fix zero-copy TCP function export
4445 - don't take duplicate lock when processing re-sent getattr (2420)
4446 - lctl uses obd_self_export instead of creating new conn (2353)
4447 - MDS/OST recovery case which requires object creation asserted (2425)
4448 - move lfs from /usr/sbin to /usr/bin in packages
4449 - fix race between mds_client_add and mds_client_free (2417)
4450 - use kmalloc instead of slabs in portals (2430)
4451 - don't create duplicate records when a failover MDS is present (2442)
4452 - remove unnecessary mount age check (2332)
4453 - don't remove directory inodes from locks prematurely (2451)
4454 - don't break if MDS service name is the same as hostname (2103)
4455 - fix races in client write RPC generation when cache full (2482)
4457 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4460 - remove now-unused request->rq_obd (278)
4461 - if an allocation fails, print out how much memory we've used (1933)
4462 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4463 - add the "configurable stack size" patch to most series files (1256)
4464 - ability to write large log records, for 100+ OST configs (2306)
4465 - fix NULL deref when filter_prep fails (2314)
4466 - fix operator precedence error in filter_sync
4467 - dynamic allocation of socknal TX descriptors (2315)
4468 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4469 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4470 - ensure that CWARN messages in Portals always get to the syslog
4471 - __init/__exit are not for prototype decls (ldlm_init/exit)
4472 - x86-64 compile warning fixes
4473 - fix gateway LMC keyword conflict (2318)
4474 - fix MDS lock inversions in getattr/reint paths (1844)
4475 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4476 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4477 - DLM race condition prevented some lock evictions (2328)
4478 - ENOMEM detection and retry on socknal sends (2230)
4479 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4480 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4481 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4482 - better error messages when a client is rejected during recovery (1505)
4483 - avoid cancelling locks which were never granted, after failure (2330)
4484 - fix i_sem/journal inversion in mds_client_add (2333)
4485 - fix truncate/getattr lock cycle deadlock (2334)
4486 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4487 - fix two rare exit paths which could leak an l_lock() ref (2321)
4488 - fencepost error in MDS/OST orphan recovery (2226)
4489 - make log record alignment 8 bytes (1988)
4490 - lstripe now fails when requested offset > ost_count (2237)
4491 - ensure that all kernel series have a complete list.h (1607)
4492 - fix crashes in special-file operations (2316)
4493 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4495 - allow configurable automake binary, for testing new versions
4496 - small update to the lfs documentation
4498 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4500 * fix negative export reference count in fsfilt_sync (2312)
4502 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4503 * release candidate 0.9.1
4505 - orphans are moved into the PENDING directory for possible recovery
4506 - replayed opens now open by fid for orphan/rename safety (1042)
4507 - last close of an orphan inode generates a transno (683)
4508 - chdir() and mount() now pin the directory entry (1020)
4509 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4510 - discard very old requests without processing them (1502)
4511 - remove some common, well-understood CERRORs (1505)
4512 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4513 - clear "grant" flags in OST replies until OST grant code lands (1644)
4514 - fix read performance by not clobbering i_blksize on client (1598)
4515 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4516 - make LRU size a /proc tunable, clears locks when reduced (707)
4517 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4518 - server locks take references on exports now (1558)
4519 - build fixes for 2.4.20-rh trees (1663)
4520 - return an error from lov_create if all OSCs are inactive (1751)
4521 - fix import levels when a reconnect happens without a timeout (1597)
4522 - exit early from mds_open if we get a lookup error (1749)
4523 - partial page read at EOF wouldn't wait for disk before sending (1642)
4524 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4525 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4526 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4527 - write/truncate lock inversion (1639)
4528 - Don't auto-load obdclass, portals modules during cleanup (1495)
4529 - fix timestamps from jumping to "now" (1763)
4530 - extra journal assertions (1648)
4531 - add an extra multiunlink test (1771)
4532 - fix read_record/write_record API (1776)
4533 - fix leak of offset_extent, possible incorrect i_size later (1772)
4534 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4535 - Debugging for the kqswnal_get_idle_tx problems (1820)
4536 - Allow recovery to be attempted multiple times (1536)
4537 - Write out MDS last_rcvd file after it is first created (1600)
4538 - Fix tx_descriptor leak in failed transmit situations (1827)
4539 - ext3 journaling fixes for assertion failure after IO error (1871)
4540 - class_export_put() on freed export after completion AST error (1896)
4541 - Fix revalidate looping in VFS (1322)
4542 - Don't access a freed export during MDS_REINT timeout (1521)
4543 - Add open-unlink recovery support on the MDS (1673,1764)
4544 - Return an error if no MDS data was read from last_rcvd (1946)
4545 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4546 - Don't setup a disk device that doesn't match exported UUID (317)
4547 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4548 - avoid committing NULL handle in force close
4549 - local.sh is now a one-stripe LOV configuration
4550 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4551 - fix ext3 htree duplicate directory entry corruption (1516)
4552 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4553 - update to new LOV EA format (2097)
4554 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4556 2003-06-15 Phil Schwan <phil@clusterfs.com>
4559 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4560 - per-import/export recovery handling (958, 931, 959)
4561 - multiple last-rcvd slots, for serving multiple FSes (949)
4562 - connections are again shared between multiple imp/exports (963, 964)
4563 - "umount -f" would hang if any requests needed to be sent (393, 978)
4564 - avoid pinning large req buffer by copying for queued messages (989)
4565 - add "uuid" to "lctl device" command to help upcalls (991)
4566 - "open" RPCs with transnos would confuse recovery counters (1037)
4567 - do proper endian conversion of all wire messages (288, 340, 891)
4568 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4569 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4570 - fix ptlrpc_connection leak in target_handle_connect (1174)
4571 - fix import refcounting bug in OST and MDS cleanup (1134)
4572 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4573 - fix very unlikely obd_types race condition (501)
4574 - remove osc_open hack for echo_client (1187)
4575 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4576 - a failure in read_inode2 leads to deadlock (1139)
4577 - cancel ack-locks as soon as transaction is committed (1072)
4578 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4579 - make sure to commitrw after any preprw to avoid deadlock (1162)
4580 - failing to execute a file in a lustre FS would lock inode (1203)
4581 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4582 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4583 - obd_uuid2tgt would walk off the end of the list (1255)
4584 - on IA64 the osc would give portals incorrect bulk size (1258)
4585 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4586 - fix lock inversion caused by new llite matching code (1282)
4587 - limit the number of dirty pages on a client to 10MB (1286)
4588 - timed out locks were not being corrected cancelled (1289)
4589 - fix O_DIRECT above 4GB on IA-32 (1292)
4590 * major user-visible changes
4591 - fail out/fail over policy now controlled by the upcall (993)
4593 - add OBD_PING to check server availability and failure (954)
4594 - lustre messages are now sent in sending host order (288, 340, 891)
4595 - add eadatalen to MDS getattr reply (340)
4596 - OST read replies may contain second buffer, with per-page status (593)
4598 2003-03-11 Phil Schwan <phil@clusterfs.com>
4601 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4602 - failed open()s could cause deadlock; fixed (867, 869)
4603 - stop cancelling OST locks when files are closed (481)
4604 - overlapping XID spaces caused network corruption (851, 853)
4605 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4606 - setattr_raw added, to do single-RPC, server-side setattrs
4607 - lmc/lconf syntax change for OST UUIDs
4608 - fix crashy race condition between ptlrpc_free_req and osc_close
4609 - don't use request in mdc_enqueue if we hit a timeout (889)
4610 - don't set the inode i_size for regular files from the MDS (896)
4611 - handle out of order completion AST (842)
4612 - don't LBUG if a lock request times out after receiving AST (913)
4613 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4614 - if a bad lock AST arrives, send an error instead of dropping entirely
4615 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4616 - fix leak in bulk IO when only partially completed (899, 900, 926)
4617 - fix O_DIRECT for ia64 (55)
4618 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4619 - C-z after timeout could hang a process forever; fixed (977)
4621 - client-side I/O cache (678, 924, 929, 941, 970)
4623 - READPAGE and SETATTRs which don't take server-side locks get
4626 2003-02-11 Phil Schwan <phil@clusterfs.com>
4629 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4630 - Fix fsfilt_extN_readpage() to read a full page of directory
4631 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4632 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4633 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4634 - Most intent operations were removed, in favour of a new RPC mode
4635 that does a single RPC to the server and bypasses most of the VFS
4636 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4637 - Aggressively cancel local locks on DLM servers
4638 - mds_reint_unlink sends EA to the client if it's the last nlink.
4639 client uses that EA to unlink OST objects.
4640 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4641 - recursive symlinks were fixed (439)
4642 - fixed NULL deref in DEBUG_REQ
4643 - filter_update_lastobjid no longer calls sync, which annoyed extN
4644 - fixed multi-client small-writes to a single file problem (445)
4645 - fixed mtime updates during file writes (607)
4646 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4647 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4648 - fixed Linux OST size reporting problem (444, 656)
4649 - OST now updates object mtime with writes or setattr (607, 619)
4650 - client verifies file size before zeroing page past EOF (445)
4651 - OST now writes last allocated objid to disk with allocation (108)
4652 - LOV on echo now works (409)
4654 - mds_reint_unlink sends a new buffer, with the EA included. this
4655 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4656 set if a regular file was being unlinked, and it was the last link
4657 - use PtlGet from the target for bulk writes (315)
4658 - OST now updates object mtime with writes or setattr (607, 619)
4659 - LDLM now has a grant-time callback to revalidate locked items, if
4661 - Many MDS operations were reorganized to combat race conditions
4663 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4664 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4666 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4669 - Fully reactivate OST imports after reconnection (512, others)
4670 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4671 - More graceful error handling for truncating on dead OST (515)
4672 - Don't error out unless we're actually accessing dead stripes (474)
4673 - Fix garbage sizes when stripes are missing (410)
4674 - LRU counters were broken, causing constant lock purge (433, 432)
4675 - garbage on read from stripes with failed OSTs (441)
4676 - mark OSCs as active before reconnecting during recovery (438)
4677 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4678 - lfind did not preserve OST order in output (443)
4679 - symlinks cause hung clients, incorrect data (439)
4680 - stop dereferencing request after dropping refcount (457)
4681 - don't LASSERT(spin_is_locked) on non-SMP (455)
4682 - fixes for many rename() bugs
4683 - fstat didn't correctly synchronize attributes (399)
4684 - server must handle lock cancellation during blocking AST prep (487)
4685 - bulk descriptors were free()d too soon (511)
4686 - fix paths in lconf, which would load incorrect modules (451, 507)
4687 - fix confusing lconf 'host not found' error message (386)
4688 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4689 - fix race condition in mdc_blocking_ast() for inode access (526)
4690 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4691 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4692 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4693 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4694 - fix lstripe in several ways (kernel side) (527)
4695 - fix request leak in ldlm_cli_enqueue (262)
4696 - incorrect OSC was marked inactive after OST failure
4697 - call mds_fs_cleanup before unmounting filesystem (524)
4698 - fix races between taking ns_lock and ldlm_lock_change_resource
4699 - fix races updating LOV export open file list
4700 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4701 - fix recovery NULL deref in ldlm_cli_cancel_unused
4702 - fix some DLM races by using new hash table for lock handles (419)
4703 - permit the client to specify desired inodes, at replay
4704 - duplicate requests when we queue them for replay reintegration
4705 - fix last_rcvd offset calculation
4706 - sync after each recovered transaction, so we always make progress
4707 - never, not always, ERESTART requests without transnos
4708 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4710 - skip replay if the MDS says that the client is already connected
4711 - don't check for a recovery-enabled export to match lctl's UUID
4712 - don't INC_USE_COUNT for phantom exports
4713 - don't crash when cleaning up phantom exports (567)
4714 - don't double-finish or set replay data for errored mdc_open requests
4715 - abort requests when they time out, so we don't get old replies
4716 - send/receive replies for AST messages again
4717 - if the client says that it doesn't have the lock, cancel it on the
4719 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4720 mark it as destroyed, it will all work out when decref is called
4721 - fix module use counts (22, 581)
4723 - ASTs now expect a reply (server cancels lock on error reply)
4725 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4728 - fix many simultaneous client startup (392)
4729 - fix dentry->d_it clobbering
4730 - credentials weren't being shipped for readdir/getattr operations
4731 - remove invalid assertions triggered during some concurrent MD
4733 - proper Lustre versions added (336, 389)
4734 - fix memory leak for create error case (398)
4735 - fix LOV locking bug that would get cli/srv out of sync
4736 - fix echo client over LOV (409)
4737 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4738 - fix double-O_EXCL intent crash (424)
4739 - avoid sending multiple lock CANCELs (352)
4741 - MDS can do multi-client recovery (modulo bugs in new code)
4743 - many updates, edits, cleanups
4745 2002-11-18 Phil Schwan <phil@clusterfs.com>
4748 - fix null d_it dereference (346)
4749 - fix full OST/dbench hang (333)
4750 - fix permission problem with file removal (286)
4751 - fix removal of OSCs from LOV when they fail
4752 - fix NULL deref during bulk timeout (214)
4753 - fix problems related to multiple filesystems on one MDS (241)
4754 - fixed serious subtle metadata locking bugs
4755 - free locks on clients when inodes are removed due to memory
4757 - fix inode pointer in lock data (285)
4758 - partial support for multiple MDS on a single host (241)
4759 - data locks weren't cancelled at clear_inode time (290, 311)
4760 - intent locks could lead to unbounded lock growth (205)
4761 - added a maximum lock count, an LRU list, and a flusher
4762 - fix multiple rename (365)
4763 - properly abstracted the echo client
4764 - OSC locked 1 byte too many; fixed
4765 - rewrote brw callback code:
4766 - fixed recovery bugs related to LOVs (306)
4767 - fixed too-many-pages-in-one-write crash (191)
4768 - fixed (again) crash in sync_io_timeout (214)
4769 - probably fixed callback-related race (385)
4771 - Add capability to MDS protocol
4772 - LDLM cancellations and callbacks on different portals
4774 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4777 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4778 - defer open object destruction to close time (601981, 138)
4779 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4780 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4781 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4782 - don't call obd_setattr in ll_file_release for destroyed objects
4783 * protocol change to lustre_msg: move |version| and add |flags|
4784 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4785 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4786 * added replay of create, unlink, link and rename operations during
4787 MDS failover; recovery should be much more robust now
4788 * remove failed OSCs from LOVs (only lov_create uses this so far)
4789 * the lustre-HOWTO was brought (more) up to date (582544)
4791 2002-10-23 Phil Schwan <phil@clusterfs.com>
4794 - in-use dentries weren't being reused properly (617851)
4795 - prevent multiple LDLM setup (599178)
4796 - fix LOV size calculations for truncate (617853)
4797 - fix client handling of MDS intent errors (POSIX)
4798 - fix permission bug in lovstripe.c test (624321)
4799 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4800 - truncate past end of file could corrupt data
4801 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4802 - a race in recovery could return ETIMEDOUT to apps (623947)
4803 - building outside the source directory was fixed
4804 * the lustre-HOWTO was brought (more) up to date (582544)
4805 * major progress was made on recovery functionality
4807 2002-10-10 Phil Schwan <phil@clusterfs.com>
4810 - recovery deadlock fix
4811 - rm -rf causes LBUG fix (617817)
4812 - file open by multiple tasks fix (618962)
4813 - directory permissions bugs (602707 and 620007)
4814 - journal_stop fixed with locking (611313)
4815 - O_APPEND failures resolved (618273, perhaps 614459)
4816 - lconf PATH fix (619770)
4817 - IA64 build fix (621450)
4818 - RPC buffer sizes scale with amount of memory
4820 2002-10-01 Phil Schwan <phil@clusterfs.com>
4823 - locks would be cancelled without throwing away data pages,
4824 resulting in inconsistent data (605627)
4825 - inode attributes were not always being refreshed (605627, 612449)
4826 - lconf now continues to cleanup after lctl reports an error
4827 - MDS now enforces user permissions (602707)
4828 - lprocfs cleanup fixed, but not yet enabled (614157)
4829 - fixed infinite server hang, should a client not respond to an AST
4830 - avoid going into recovery if user calls readlink() with a buffer
4831 that's too small (613941)
4832 - AST RPCs no longer require replies (614867) -- this may be changed
4833 - don't crash server if client sends an IOV that's too big (611336)
4834 - fixed lock conversion deadlock (611892)
4835 - fixed the following of symlinks (614622)
4836 * recovery: the server can remove locks from a client that dies, other
4837 clients can make progress
4838 * more extN patch fixes
4839 * compile-time configurable ptlrpc buffer allocations
4841 - collaborative read cache document
4842 - Lustre Lite Performance CDR document-in-progress
4844 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4847 - fix typo in patch-2.4.18
4849 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4852 - clear ptlrpc request each time in handle_incoming_request()
4853 - unlink of files now destroys the object on the OST
4855 2002-09-19 Peter Braam <braam@clusterfs.com>
4857 * add hard link support
4858 * change obdfile creation method
4859 * kernel patch changed
4861 2002-09-19 Peter Braam <braam@clusterfs.com>
4864 - stack overflow bug in extN fixed
4866 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4868 * documentation updates
4869 - add man pages for config tools
4870 - update tests/README to describe testing with new config tools
4871 - finish metadata API descriptions
4872 * bug fixes and cleanups
4873 - statfs workaround for 16TB limit
4874 - LOV stripe allocation improved, can stripe on subset of OSTs
4875 - LOV file size/IO offset was wrong for files > 4GB in size
4876 - object EA data was being dropped, caused files to be unreadable
4877 - memory overflow with non-LOV OST caused memory corruption
4878 - fixed regression tests to work with new config tools, obdfilter
4879 - fixed bug when directory size became larger than 1 block
4880 - fixed bug (for single client case) when PWD was deleted
4881 - invalidate local directory pages when doing intent-based ops
4882 - avoid LDLM oops when lock callback contained bad data
4884 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4886 * documentation updates
4887 * bug fixes and cleanups
4888 - configuration tools
4891 - 64-bit compile warnings
4892 - 64-bit internal statfs data
4894 * test_brw on persistent OST devices
4896 * lprocfs (disabled)
4898 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4900 * documentation updates
4901 * bug fixes and cleanups
4902 * configuration tools
4904 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4907 * many small fixes to 0_5_4
4908 * io/network handling
4909 * thinkos in MDS operations
4911 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4914 * crucial basic fixes to 0.5.3
4915 * IOR, Iozone work over Elan
4918 2002-08-07 Phil Schwan <phil@clusterfs.com>
4919 * version 0_5_3, our first alpha
4920 * we use the new Portals iovs
4921 * documentation updates
4922 * bug fixes and cleanups
4923 * small changes in the DLM wire protocol
4925 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4926 * version 0_5_1 with some initial stability,
4927 * locking on MD and file I/O.
4928 * documentation updates
4929 * several bug fixes since 0.5.0
4930 * small changes in wire protocol
4932 2002-07-18 Phil Schwan <phil@clusterfs.com>
4934 * delivered as Lustre Light Alpha
4935 * fixed a crash after handling invalid MDS requests
4936 * fixed directory pages for architectures with non-4k pages sizes
4938 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4939 * release version v0_4_4
4940 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4942 * Fixes a number of interruption problems with OST operations.
4943 * Update documentation for portals header changes
4944 * Move all wire protocol structs/defines to lustre_idl.h
4945 * Fixes symlink length bug.
4946 * Add tcpdump to repository.
4948 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4949 * release version v0_4_3
4950 * Fixes statfs for inodes on extN.
4951 * Fixes bug in runtests which would delete /etc/hosts.
4952 * Use 64-bit object IDs wherever possible (not into VFS though)
4953 Remove ost_get_info, which is unused by lustre, and out of date.
4955 2002-07-03 Peter Braam <braam@clusterfs.com>
4956 * release version v0_4_2 Fixes a lookup error (type not passed)
4957 * move forward to head of Portals
4958 * move forward to latest Lustre kernel
4960 2002-06-25 Peter Braam <braam@clusterfs.com>
4961 * release version v0_4_1. Hopefully stable on single node use.