4 2.6.9-55.0.9.EL (RHEL 4),
5 2.6.16.53-0.8 (SLES 10),
6 2.6.18-8.1.14.EL5 (RHEL 5).
7 2.6.18.8 vanilla (kernel.org)
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.9-42.0.10.EL (RHEL 4),
11 2.6.16 - 2.6.21 vanilla (kernel.org)
12 * Recommended e2fsprogs version: 1.40.2-cfs1
13 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
17 Description: Remove 2.4 kernel checks lustre 1.8
18 Details : Remove 2.5.0 check from quota_check.c, quota_ctl.c,
19 quota_interface.c in quota subsystem.
20 Remove 2.4 kernel check from obd_mount.c, linux/linux-sysctl.c in
22 Remove lock_24kernel from fsfilt_ext3.c fsfilt_reiserfs.c
23 lvfs_linux.c upcall_cache.c in lvfs subsystem.
24 Remove 2.4 kernel check from lproc_osc.c, osc_request.c in osc
27 Severity : enhancement
29 Description: Build SLES10 patchless client fails
30 Details : The configure was broken by run ./configure with
31 --with-linux-obj=.... argument for patchless client. When the
32 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
33 can't search header adequately. Use absolute path such as
34 -I($LINUX)/include instead.
38 Description: interrupt oig_wait produce painc on resend.
39 Details : brw_redo_request can be used for resend requests from ptlrpcd and
40 private set, and this produce situation when rq_ptlrpcd_data not
41 copyed to new allocated request and triggered LBUG on assert
42 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
43 ptlrpcd set if request is changed and can be safety changed to use
48 Description: LASSERT_{REQ,REP}SWAB macros are buggy
49 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
50 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
51 these macros and replace them with their difinitions instead.
55 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
56 Details : Update which_patch & target file for SLES10 latest kernel.
59 Description: Remove client patches from SLES 10 kernel.
60 Details : This causes SLES 10 clients to behave as patchless clients
61 even on a Lustre-patched (server) kernel.
63 Severity : enhancement
65 Description: self-adjustable client's lru lists
66 Details : use adaptive algorithm for managing client cached locks lru
67 lists according to current server load, other client's work
68 pattern, memory activities, etc. Both, server and client
69 side namespaces provide number of proc tunables for controlling
72 Severity : enhancement
74 Description: light-weight GSS support
75 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
76 and reduce performance overhead.
78 Severity : enhancement
80 Description: Linux keyring support
81 Details : Support using service of Linux keyring for Lustre GSS internal
82 context refresh/cache mechanism.
86 Description: Fix errors in lfs documentation
87 Details : Fixes man pages
91 Description: don't use GFP_* in generic Lustre code.
92 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
96 Description: obdclass is limited by single OBD_ALLOC(idarray)
97 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
101 Description: warnings with build patchless client with vanila 2.6.19 and up
102 Details : change old ctl_table style and replace ctl_table/ctl_table_header
103 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
107 Frequency : Only for SLES
108 Description: sanity_quota fail test_1
109 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
110 file, which makes TSTID[2] inunique.
114 Frequency : Only for SLES9
115 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
116 Details : Update target/ChangeLog/which_patch .
120 Frequency : Only for test_10 in sanity-quota.sh
121 Description: a bug in quota test code
122 Details : checking a test flag in a wrong place
126 Frequency : Only for test_13 in sanity-quota.sh
127 Description: a bug in quota test script
128 Details : shouldn't check overfull allocated quota
132 Description: jbd statistics
133 Details : Port older jbd statistics patch for sles10
135 Severity : enhancement
137 Description: Add printing inode info into message about error in writepage.
142 Description: reply_lock_interpret crash due to race with it and lock cancel.
143 Details : Do not replay locks that are being cancelled. Do not reference
144 locks by their address during replay, just by their handle.
148 Frequency : When flocks are used.
149 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
150 Details : Flock locks might destroy just granted lock if it could be merged
151 with another existing flock, this is done in completion handler,
152 so teach ldlm_cli_enquque_fini that this is a valid case for
158 Description: Oops in read and write path when failing to allocate lock.
159 Details : Check if lock allocation failed and return error back.
163 Description: lstripe command fails for valid OST index
164 Details : The stripe offset is compared to lov->desc.ld_tgt_count
165 instead of lov->desc.ld_active_tgt_count.
169 Description: sanity.sh failed test 103
170 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
175 Description: df doesn't work properly if diskfs blocksize != 4K
176 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
180 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
181 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
185 Description: obd_health_check_timeout too short
186 Details : set obd_health_check_timeout as 1.5x of obd_timeout
190 Description: llapi_file_create() does not allow some changes
191 Details : add llapi_file_open() that allows specifying the mode and
192 open flags, and also returns an open file handle.
196 Description: merge and cleanup kernel patches.
197 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
201 Description: Add journal checksum support.(Kernel part)
202 Details : The journal checksum feature adds two new flags i.e
203 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
204 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
205 indicates that the commit block contains the checksum for
206 the blocks described by the descriptor blocks. Now commit
207 record can be sent to disk without waiting for descriptor
208 blocks to be written to disk. This behavior is controlled
209 using JBD2_FEATURE_ASYNC_COMMIT flag.
213 Description: OSS needs mutliple precreate threads
214 Details : Add ability to start more than one create thread per OSS.
218 Description: Sanity.sh test_65k failed: setstripe should have succeeded
219 Details : execute lfs setstripe on client
223 Description: mds_obd_create error creating tmp object
224 Details : When the user sets quota on root, llog will be affected and can't
225 create files and write files.
228 Frequency : Always on ia64 patchless client, and possibly others.
230 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
231 Details : This allows the patchless client to be loaded on architectures
236 Description: RedHat Update kernel for RHEL5
237 Details : Add the kernel config file for RHEL5.
241 Description: RedHat Update kernel for RHEL5
242 Details : Modify the kernel config file more closer RHEL5.
246 Description: Build failure against Centos5 (RHEL5)
247 Details : Define PAGE_SIZE when it isn't present.
251 Description: test_7 on sanity-quota.sh can't be used on separate machines
252 Details : add facet to handle it
256 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
257 Details : with NFS, the anon dentry's parent was set to itself in
258 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
259 corresponding dentry other than use rec->ur_name.
261 Severity : enhancement
263 Description: lfs setstripe enhancement
264 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
268 Description: testing performance impact of enabling checksumming
269 Details : enable checksum by default, allow --disable-checksum
270 configure option and "-o nochecksum" mount option
274 Description: lustre support for RHEL5
275 Details : Add support for RHEL5.
279 Description: Client eviction due to failover config
280 Details : after a connection loss, the lustre client should attempt to
281 reconnect to the last active server first before trying the
282 other potential connections.
286 Description: when mds and osts use different quota unit(32bit and 64bit),
287 quota will be released repeatly.
288 Details : void sending multiple quota reqs to mds, which will keep the
289 status between the reqs.
293 Description: osts not allocated evenly to files
294 Details : change the condition to increase offset_idx
297 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
299 Description: Data corruption for OSTs that are formatted larger than 2TB
301 Details : When generating the bio request for lustre file writes the
302 sector number would overflow a temporary variable before being
303 used for the IO. The data reads correctly from Lustre (which
304 will overflow in a similar manner) but other file data or
305 filesystem metadata may be corrupted in some cases.
309 Description: Tune the kernel for good SCSI performance.
310 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
311 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
316 Description: rewrite ext2-derived code in obdclass/uuid.c
317 Details : rewrite inherited code (uuid parsing code from ext2 utils)
318 from scratch preserving functionality.
322 Description: "lfs find -obd UUID" prints directories
323 Details : "lfs find -obd UUID" will return all directory names instead
324 of just file names. It is incorrect because the directories
325 do not reside on the OSTs.
329 Description: use i_size_read and i_size_write in 2.6 port
330 Details : replace inode->i_size access with i_size_read/write()
333 Frequency : when removing large files
335 Description: scheduling issue during removal of large Lustre files
336 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
337 It causes scheduling issues when removing large files (17TB in the
341 Frequency : only with liblustre clients on XT3
343 Description: evictions taking too long
344 Details : allow llrd to evict clients directly on OSTs
347 Frequency : only on ppc
349 Description: /proc/fs/lustre/devices broken on ppc
350 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
351 We should make sure the type of 'index' is loff_t and then cast
352 explicitly as needed below. Do not assign an explicitly cast
356 Frequency : only for rhel5
358 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
359 Details : Modify the target file & which_kernel.
361 Severity : enhancement
363 Description: omit set fsid for export NFS
364 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
365 as FS_REQUIRES_DEV for avoid problems with generate fsid.
369 Frequency : Always, for kernels after 2.6.16
370 Description: Fix warning idr_remove called for id=.. which is not
372 Details : Last kernels save old s_dev before kill super and not allow
373 to restore from callback - restore it before call kill_anon_super.
377 Description: buffer overruns could theoretically occur
378 Details : llapi_semantic_traverse() modifies the "path" argument by
379 appending values to the end of the origin string, and a
380 overrun may occur. Adding buffer overrun check in liblustreapi.
384 Description: Fix error on 'ls .' at the top of the Lustre mount.
385 Details : Don't revalidate dentry if it is a root dentry.
389 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
390 Details : Modify vm-tunables-rhel4.patch.
394 Description: Kernel config for 2.6.18-vanilla.
395 Details : Modify targets/2.6-vanilla.target.in.
396 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
397 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
398 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
399 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
403 Description: improve handling recoverable errors
404 Details : if request processig with error which can be recoverable on server
405 request should be resend, otherwise page released from cache and
410 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
411 Details : Modify target file & which_patch.
412 A flaw was found in the IA32 system call emulation provided
413 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
414 value could be stored in the %RAX register, which could trigger an
415 out-of-bounds system call table access. An untrusted local user
416 could exploit this flaw to run code in the kernel
417 (ie a root privilege escalation). (CVE-2007-4573).
421 Description: O_DIRECT bypasses client statistics.
422 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
423 read_ahead_stats were updated. copied the stats section from
424 osc_send_oap_rpc() into async_internal().
428 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
432 Description: change order of libsysio includes
433 Details : '#include sysio.h' should always come before '#include xtio.h'
437 Description: handle "serious error: objid * already exists" more gracefully
438 Details : If LAST_ID value on disk is smaller than the objects existing in
439 the O/0/d* directories, it indicates disk corruption and causes an
440 LBUG(). If the object is 0-length, then we should use the existing
441 object. This will help to avoid a full fsck in most cases.
443 Severity : enhancement
445 Description: adapt the lustre_config script to support the upgrade case
446 Details : Add "-u" option for lustre_config script to support upgrading 1.4
447 server targets to 1.6 in parallel.
451 Description: To avoid grant space > avaible space when the disk is almost
452 full. Without this patch you might see the error "grant XXXX >
453 available" or some LBUG about grant, when the disk is almost
455 Details : In filter_check_grant, for non_grant cache write, we should
456 check the left space by if (*left > ungranted + bytes), instead
457 of (*left > ungranted), because only we are sure the left space
458 is enough for another "bytes", then the ungrant space should be
459 increase. In client, we should update cl_avail_grant only there
460 is OBD_MD_FLGRANT in the reply.
464 Description: Update RHEL 4 kernel to fix local root privilege escalation.
465 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
466 described in CVE-2007-4573. This problem could allow untrusted
467 local users to gain root access.
471 Description: Remove 2.4 kernel checks lustre 1.8
472 Details : Remove 2.5.0 check from dcache.c dir.c file.c llite_internal.h
473 llite_lib.c llite_mmap.c lproc_llite.c namei.c rw.c symlink.c
475 Remove 2.5.0 check from handler.c lproc_mds.c mds_fs.c
476 mds_internal.h mds_lib.c mds_open.c mds_reint.c mds_xattr.c
480 Frequency : when using O_DIRECT and quotas
482 Description: Incorrect file ownership on O_DIRECT output files
483 Details : block usage reported by 'lfs quota' does not take into account
484 files that have been written with O_DIRECT.
486 --------------------------------------------------------------------------------
488 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
490 * Support for kernels:
491 2.6.5-7.283 (SLES 9),
492 2.6.9-55.EL (RHEL 4),
493 2.6.16.46-0.14 (SLES 10),
494 2.6.18.8 vanilla (kernel.org)
495 * Client support for unpatched kernels:
496 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
497 2.6.16 - 2.6.22 vanilla (kernel.org)
498 * Due to recently discovered recovery problems, we do not recommend
499 using patchless RHEL 4 clients with this or any earlier release.
500 * Recommended e2fsprogs version: 1.39.cfs8
501 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
502 * Starting with this release, the ldiskfs backing filesystem required
503 by Lustre is now in its own package, lustre-ldiskfs. This package
504 should be installed. It is versioned separately from Lustre and
505 may be released separately in future.
509 Description: block reactivating mgc import until all deactivates complete
510 Details : Fix race when failing back MDT/MGS to itself (testing)
513 Frequency : at statup only
515 Description: mds_lov_synchronize race leads to various problems
516 Details : simultaneous MDT->OST connections at startup can cause the
517 sync to abort, leaving the OSC in a bad state.
519 Severity : enhancement
521 Description: add optional extra BUILD_VERSION info
522 Details : add a new environment variable (namely LUSTRE_VERS) which allows
523 to override the lustre version.
526 Frequency : 2.6.18 servers only
528 Description: ll_kern_mount() doesn't release the module reference
529 Details : The ldiskfs module reference count never drops down to 0
530 because ll_kern_mount() doesn't release the module reference.
535 Description: server LBUG when using old ost_num_threads parameter
536 Details : Accept the old ost_num_threads parameter but warn that it
537 is deprecated, and fix an off-by-one error that caused an LBUG.
542 Description: Transient SCSI error results in persistent IO issue
543 Details : iobuf->dr_error is not reinitialized to 0 between two
547 Frequency : sometimes when underlying device returns I/O errors
549 Description: OSTs not going read-only during write failures
550 Details : OSTs are not remounted read-only when the journal commit threads
551 get I/O errors because fsfilt_ext3 calls journal_start/stop()
552 instead of the ext3 wrappers.
556 Description: poor connect scaling with increasing client count
557 Details : Don't run filter_grant_sanity_check for more than 100 exports
558 to improve scaling for large numbers of clients.
561 Frequency : SLES10 only
563 Description: sanity-quota.sh quotacheck failed: rc = -22
564 Details : Quotas cannot be enabled on SLES10.
567 Frequency : liblustre clients only
569 Description: getdirentries does not give error when run on compute nodes
570 Details : getdirentries does not fail when the size specified as an argument
571 is too small to contain at least one entry
573 Severity : enhancement
575 Description: Add LNET router traceability for debug purposes
576 Details : If a checksum failure occurs with a router as part of the
577 IO path, the NID of the last router that forwarded the bulk data
578 is printed so it can be identified.
583 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
584 Details : Due to a race condition, liblustre clients were occasionally
587 Severity : enhancement
589 Description: lfs setstripe use optional parameters instead of postional
592 Severity : enhancement
594 Description: Nanosecond timestamp support for ldiskfs
595 Details : The on-disk ldiskfs filesystem has added support for nanosecond
596 resolution timestamps. There is not yet support for this at
597 the Lustre filesystem level.
600 Frequency : during server recovery
602 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
603 Details : request with rq_no_resend flag not awake l_wait_event if they get a
607 Frequency : nfs export on patchless client
609 Description: connectathon hang when test nfs export over patchless client
610 Details : Disconnected dentry cannot be found with lookup, so we do not need
611 to unhash it or make it invalid
614 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
615 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
616 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
621 Description: open req refcounting wrong on reconnect
622 Details : If reconnect happened between getting open reply from server and
623 call to mdc_set_replay_data in ll_file_open, we will schedule
624 replay for unreferenced request that we are about to free.
625 Subsequent close will crash in variety of ways.
626 Check that request is still eligible for replay in
627 mdc_set_replay_data().
632 Description: disable writes to filesystem when reading health_check file
633 Details : the default for reading the health_check proc file has changed
634 to NOT do a journal transaction and write to disk, because this
635 can cause reads of the /proc file to hang and block HA state
636 checking on a healthy but otherwise heavily loaded system. It
637 is possible to return to the previous behaviour during configure
638 with --enable-health-write.
640 Severity : enhancement
642 Description: 64-bit inode version
643 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
644 to the inode. This will be required for version-based recovery.
649 Description: MDS fails to start if a duplicate client export is detected
650 Details : in some rare cases it was possible for a client to connect to
651 an MDS multiple times. Upon recovery the MDS would detect this
652 and fail during startup. Handle this more gracefully.
654 Severity : enhancement
656 Description: Add -o localflock option to simulate old noflock
658 Details : This will achieve local-only flock/fcntl locks
664 Description: log_commit_thread vs filter_destroy race leads to crash
665 Details : Take import reference before releasing llog record semaphore
670 Description: Wrong request locking in request set processing
671 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
672 imp_delayed_list, in this place should be used imp_lock.
675 Frequency : when reconnection
677 Description: Grant Leak when osc reconnect to OST
678 Details : When osc reconnect ost, OST(filter) should check whether it
679 should grant more space to client by comparing fed_grant and
680 cl_avail_grant, and return the granted space to client instead
681 of "new granted" space, because client will call osc_init_grant
682 to update the client grant space info.
685 Frequency : when client reconnect to OST
687 Description: Grant Leak when osc do resend and replay bulk write
688 Details : When osc reconnect to OST, OST(filter)should clear grant info of
689 bulk write request, because the grant info while be sync between
690 OSC and OST when reconnect, and we should ignore the grant info
691 these of resend/replay write req.
696 Description: Grant space more than avaiable left space sometimes.
697 Details : When then OST is about to be full, if two bulk writing from
698 different clients came to OST. Accord the avaliable space of the
699 OST, the first req should be permitted, and the second one
700 should be denied by ENOSPC. But if the seconde arrived before
701 the first one is commited. The OST might wrongly permit second
702 writing, which will cause grant space > avaiable space.
705 Frequency : when client is evicted
707 Description: Grant might be wrongly erased when osc is evicted by OST
708 Details : when the import is evicted by server, it will fork another
709 thread ptlrpc_invalidate_import_thread to invalidate the
710 import, where the grant will be set to 0. While the original
711 thread will update the grant it got when connecting. So if
712 the former happened latter, the grant will be wrongly errased
713 because of this race.
718 Description: Checking Stale with correct fid
719 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
720 to check whether it is stale, because sometimes, we want the
721 enqueue happened anyway, and op_data.fid2 will not be initialized.
723 Severity : enhancement
725 Description: update patchless client
726 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
729 Frequency : only with 2.4 kernel
731 Description: random memory corruption
732 Details : size of struct ll_inode_info is to big for union inode.u and this
733 can be cause of random memory corruption.
738 Description: Memory leak in recovery
739 Details : Lov_mds_md was not free in an error handler in mds_create_object.
740 It should also check obd_fail before fsfilt_start, otherwise if
741 fsfilt_start return -EROFS,(failover mds during mds recovery).
742 then the req will return with repmsg->transno = 0 and rc = EROFS.
743 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
744 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
745 be freed no matter whether fsfilt_commit success or not.
748 Frequency : only with huge count clients
750 Description: Prevents from taking the superblock lock in llap_from_page for
752 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
753 for llap_from_page call in ll_removepage prevents from taking the
754 superblock lock for a soon died page.
759 Description: Not check open intent error before release open handle
760 Details : in some rare cases, the open intent error is not checked before
761 release open handle, which may cause
762 ASSERTION(open_req->rq_transno != 0), because it tries to release
763 the failed open handle.
768 Description: Set cat log bitmap only after create log success.
769 Details : in some rare cases, the cat log bitmap is set too early. and it
770 should be set only after create log success.
774 Description: Accessing a block bevice can re-enable I/O when Lustre is
775 tearing down a device.
776 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
781 Description: service threads may hog cpus when there are a lot of requests
783 Details : Insert cond_resched to give other threads a chance to use some of
789 Description: the cat log was not initialized in recovery
790 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
791 unlink log on mds will not be initialized until mds post
792 recovery. And also in mds post recovery, the unlink log will
793 initialization will be done asynchronausly, so there will be race
794 between add unlink log and unlink log initialization.
798 Description: brw_stats were being printed incorrectly
799 Details : brw_stats were being printed as log2 but all of them were not
800 recorded as log2. Also remove some code duplication arising from
801 filter_tally_{read,write}.
805 Frequency : rare, only in recovery.
806 Description: ASSERTION(req->rq_type != LI_POISON) failed
807 Details : imp_lock should be held while iterating over imp_sending_list for
808 prevent destroy request after get timeout in ptlrpc_queue_wait.
812 Description: replay-single.sh test 52 fails
813 Details : A lock's skiplist need to be cleanup when it being unlinked
814 from its resource list.
818 Description: Short directio read returns full requested size rather than
820 Details : Direct I/O operations should return actual amount of bytes
821 transferred rather than requested size.
823 Severity : enhancement
825 Description: metadata RPC reduction (e.g. for rm performance)
826 Details : decrease the amount of synchronous RPC between clients and servers
827 by canceling conflicing lock before the operation on the client side
828 and packing thier handles into the main operation RPC to server.
830 Severity : enhancement
832 Description: Async OSC create to avoid the blocking unnecessarily.
833 Details : If a OST has no remain object, system will block on the creating
834 when need to create a new object on this OST. Now, ways use
835 pre-created objects when available, instead of blocking on an
836 empty osc while others are not empty. If we must block, we block
837 for the shortest possible period of time.
839 --------------------------------------------------------------------------------
841 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
846 Frequency : on some architectures
848 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
849 Details : Uninitialized flags sometimes cause configuration commands to
853 Frequency : patchless clients only
855 Description: missing __iget() symbol export
856 Details : The __iget() symbol export is missing. To avoid the need for
857 this on patchless clients the deathrow inode reaper is turned
858 off, and we depend on the VM to clean up old inodes. This
859 dependency was during via the fix for bug 12181.
861 --------------------------------------------------------------------------------
863 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
865 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
866 INTEROPERATE with older versions automatically. In many cases a
867 special upgrade step is needed. Please read the
868 user documentation before upgrading any part of a live system.
869 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
870 version will not interoperate with 1.6 betas before beta5 (1.5.95).
871 * WARNING: Lustre configuration and startup changes are required with
872 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
878 Severity : enhancement
880 Description: MountConf
881 Details : Lustre configuration is now managed via mkfs and mount
882 commands instead of lmc and lconf. New obd types (MGS, MGC)
883 are added for dynamic configuration management. See
884 https://mail.clusterfs.com/wikis/lustre/MountConf for
887 Severity : enhancement
889 Description: dynamic OST addition
890 Details : OSTs can now be added to a live filesystem
892 Severity : enhancement
894 Description: startup order invariance
895 Details : MDTs and OSTs can be started in any order. Clients only
896 require the MDT to complete startup.
898 Severity : enhancement
900 Description: parallel, asynchronous orphan cleanup
901 Details : orphan cleanup is now performed in separate threads for each
902 OST, allowing parallel non-blocking operation.
904 Severity : enhancement
906 Description: optimized stripe assignment
907 Details : stripe assignments are now made based on ost space available,
908 ost previous usage, and OSS previous usage, in order to try
909 to optimize storage space and networking resources.
911 Severity : enhancement
913 Description: Permanently set tunables
914 Details : All writable /proc/fs/lustre tunables can now be permanently
915 set on a per-server basis, at mkfs time or on a live
918 Severity : enhancement
920 Description: Lustre message v2
921 Details : Add lustre message format v2.
923 Severity : enhancement
925 Description: client OST exclusion list
926 Details : Clients can be started with a list of OSTs that should be
927 declared "inactive" for known non-responsive OSTs.
931 Description: SPEC SFS validation failure on NFS v2 over lustre.
932 Details : Changes the blocksize for regular files to be 2x RPC size,
933 and not depend on stripe size.
935 Severity : enhancement
937 Description: Multiple MD RPCs in flight.
938 Details : Further unserialise some read-only MDS RPCs - learn about intents.
939 To avoid overly-overloading MDS, introduce a limit on number of
940 MDS RPCs in flight for a single client and add /proc controls
941 to adjust this limit.
943 Severity : enhancement
945 Description: client read/write statistics
946 Details : Add client read/write call usage stats for performance
947 analysis of user processes.
948 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
949 file access. extents_stats shows chunk size distribution.
950 extents_stats_per_process show chunk size distribution per
953 Severity : enhancement
955 Description: mds statistics
956 Details : Add detailed mds operations statistics in
957 /proc/fs/lustre/mds/*/stats.
961 Description: Failure of copying files with lustre special EAs.
962 Details : Client side always return success for setxattr call for lustre
963 special xattr (currently only "trusted.lov").
968 Description: Refcount LNET uuids
969 Details : The global LNET uuid list grew linearly with every startup;
970 refcount repeated list entries instead of always adding to
973 Severity : enhancement
975 Description: Dynamic service threads
976 Details : Within a small range, start extra service threads
977 automatically when the request queue builds up.
980 Frequency : mixed-endian client/server environments
982 Description: mixed-endian crashes
983 Details : The new msg_v2 system had some failures in mixed-endian
986 Severity : enhancement
988 Description: Easy OST removal
989 Details : OSTs can be permanently deactivated with e.g. 'lctl
990 conf_param lustre-OST0001.osc.active=0'
992 Severity : enhancement
994 Description: MGS proc entries
995 Details : Added basic proc entries for the MGS showing what filesystems
998 Severity : enhancement
1000 Description: provide MGS failover
1001 Details : Added config lock reacquisition after MGS server failover.
1003 Severity : enhancement
1005 Description: add Linux 2.4 support
1006 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1010 Description: a large application tries to do I/O to the same resource and dies
1011 in the middle of it.
1012 Details : Check the req->rq_arrival time after the call to
1013 ost_brw_lock_get(), but before we do anything about
1014 processing it & sending the BULK transfer request. This
1015 should help move old stale pending locks off the queue as
1016 quickly as obd_timeout.
1019 Frequency : when an incorrect nid is specified during startup
1021 Description: ptlrpc connect to non-existant node causes kernel crash
1022 Details : LNET can't be re-entered from an event callback, which
1023 happened when we expire a message after the export has been
1024 cleaned up. Instead, hand the zombie cleanup off to another
1027 Severity : enhancement
1029 Description: plain/inodebits lock performance improvement
1030 Details : Grouping plain/inodebits in granted list by their request modes
1031 and bits policy, thus improving the performance of search through
1035 Frequency : only if OST filesystem is corrupted
1037 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1038 Details : for a short time RPCs with bulk IO are in the replay list,
1039 but replay of bulk IOs is unimplemented. If the OST filesystem
1040 is corrupted due to disk cache incoherency and then replay is
1041 started it is possible to trip an assertion. Avoid putting
1042 committed RPCs into the replay list at all to avoid this issue.
1045 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1047 Description: System hang on startup
1048 Details : This bug allowed the liblustre (e.g. catamount) client to
1049 return to the app before handling all startup RPCs. This
1050 could leave the node unresponsive to lustre network traffic
1051 and manifested as a server ptllnd timeout.
1053 Severity : enhancement
1055 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1056 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1057 boolean to control whether to print peer debug info when a
1058 client's RPC times out.
1061 Frequency : only for kernels with patches from Lustre below 1.4.3
1063 Description: Remove old rdonly API
1064 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1067 Frequency : only for devices with external journals
1069 Description: Set external device read-only also
1070 Details : During a commanded failover stop, we set the disk device
1071 read-only while the server shuts down. We now also set any
1072 external journal device read-only at the same time.
1075 Frequency : when upgrading from 1.4 while trying to change parameters
1077 Description: The wrong (new) MDC name was used when setting parameters for
1078 upgraded MDT's. Also allows changing of OSC (and MDC)
1079 parameters if --writeconf is specified at tunefs upgrade time.
1082 Frequency : when setting specific ost indicies
1084 Description: QOS code breaks on skipped indicies
1085 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1086 created with --index need not be sequential.
1088 Severity : enhancement
1090 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1091 Details : The uninit_groups feature works in conjunction with the kernel
1092 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1093 up the pass1 processing of e2fsck. This is a read-only feature
1094 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1095 mount filesystems that have had this feature enabled.
1097 Severity : enhancement
1099 Description: Improve multi-block allocation algorithm to avoid fragmentation
1100 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1101 allocation locality and avoid filesystem fragmentation.
1103 ------------------------------------------------------------------------------
1105 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1107 * Support for kernels:
1108 2.4.21-47.0.1.EL (RHEL 3)
1109 2.6.5-7.283 (SLES 9)
1110 2.6.9-42.0.10.EL (RHEL 4)
1111 2.6.12.6 vanilla (kernel.org)
1112 2.6.16.27-0.9 (SLES 10)
1113 * Recommended e2fsprogs version: 1.39.cfs5
1115 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1119 Frequency : occasional, depends on client load and configuration
1120 Bugzilla : 12181, 12203
1121 Description: data loss for recently-modified files
1123 Details : In some cases it is possible that recently written or created
1124 files may not be written to disk in a timely manner (this should
1125 normally be within 30s unless client IO load is very high).
1126 The problem appears as zero-length files or files that are a
1127 multiple of 1MB in size after a client crash or client eviction
1128 that are missing data at the end of the file.
1130 This problem is more likely to be hit on clients where files are
1131 repeatedly created and unlinked in the same directory, clients
1132 have a large amount of RAM, have many CPUs, the filesystem has
1133 many OSTs, the clients are rebooted frequently, and/or the files
1134 are not accessed by other nodes after being written.
1136 The presence of the problem can be detected by looking at
1137 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1138 smaller than the second (nr_unused) then dirty files will not
1139 be flushed automatically to disk. "sync; sleep 10" should be
1140 run several times on the node before unmounting it to update
1141 Lustre (this is also safe to run on nodes without this problem).
1143 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1144 that can cause this same problem, so customers using that kernel
1145 also need to update the kernel in addition to Lustre. In order
1146 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1148 It is normal that files written just before a client crash (less
1149 than 30s) may not yet have been flushed to disk, even for local
1153 Frequency : frequent on thin XT3 nodes
1155 Description: UUID collision on thin XT3 Linux nodes
1156 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1157 randomly, since we relied on an insufficiently-seeded PRNG.
1162 Description: OSS hangs after "All ost request buffers busy"
1163 Details : A deadlock between quota and journal operations caused OSS
1164 hangs after printing "All ost request buffers busy."
1167 Frequency : always on liblustre builds
1169 Description: Cleanup compiler warnings on liblustre
1172 Frequency : always on liblustre builds on XT3
1174 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1178 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1179 > 32000 subdirectories
1180 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1181 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1182 subdirectory count crosses 32000. This will aid e2fsck to
1183 correctly handle more than 32000 subdirectories.
1186 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1188 Description: System hang on startup
1189 Details : This bug allowed the liblustre (e.g. catamount) client to
1190 return to the app before handling all startup RPCs. This
1191 could leave the node unresponsive to lustre network traffic
1192 and manifested as a server ptllnd timeout.
1194 Severity : enhancement
1196 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1197 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1198 boolean to control whether to print peer debug info when a
1199 client's RPC times out.
1204 Description: make O_SYNC working on 2.6 kernels
1205 Details : 2.6 kernels use different method for mark pages for write,
1206 so need add a code to lustre for O_SYNC work.
1211 Description: Failure to close file and release space on NFS
1212 Details : Put inode details into lock acquired in ll_intent_file_open.
1213 Use mdc_intent_lock in ll_intent_open to properly
1214 detect all kind of errors unhandled by mdc_enqueue.
1219 Description: proc file read during shutdown sometimes raced obd removal,
1221 Details : Add lock to prevent obd access after proc file removal.
1224 Frequency : Only for files larger than 4GB on 32-bit clients.
1226 Description: improperly doing page alignment of locks
1227 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1228 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1233 Description: RPCs being resent when they shouldn't be
1234 Details : Some RPCs that should not be resent are being resent. This
1235 can cause inconsistencies in the RPC state machine. Do not
1236 resend such requests.
1239 Frequency : rare, only with NFS export
1241 Description: Crash on NFS re-export node
1242 Details : under very unusual load conditions an assertion is hit in
1243 ll_intent_file_open()
1246 Frequency : only if OST filesystem is corrupted
1248 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1249 Details : for a short time RPCs with bulk IO are in the replay list,
1250 but replay of bulk IOs is unimplemented. If the OST filesystem
1251 is corrupted due to disk cache incoherency and then replay is
1252 started it is possible to trip an assertion. Avoid putting
1253 committed RPCs into the replay list at all to avoid this issue.
1258 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1259 Details : Large single O_DIRECT read and write calls can fail to allocate
1260 a sufficiently large buffer to process the request. In case of
1261 allocation failure the allocation is retried with a smaller
1262 buffer and broken into smaller requests.
1264 Severity : enhancement
1266 Description: Add -o localflock option to simulate old noflock behaviour.
1267 Details : This will achieve local-only flock/fcntl locks coherentness.
1272 Description: versioning check is incomplete
1273 Details : Checking the version difference of client vs. server, report
1274 error if the gap is too big.
1279 Description: add support PG_writeback bit
1280 Details : add support for PG_writeback bit for Lustre, for more carefull
1281 work with page cache in 2.6 kernel. This also fix some deadlocks
1282 and remove hack for work O_SYNC with 2.6 kernel.
1284 Severity : enhancement
1286 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1287 Details : The uninit_groups feature works in conjunction with the kernel
1288 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1289 up the pass1 processing of e2fsck. This is a read-only feature
1290 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1291 mount filesystems that have had this feature enabled.
1293 Severity : enhancement
1295 Description: Improve multi-block allocation algorithm to avoid fragmentation
1296 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1297 allocation locality and avoid filesystem fragmentation.
1299 ------------------------------------------------------------------------------
1301 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1303 * Support for kernels:
1304 2.6.9-42.0.3.EL (RHEL 4)
1305 2.6.5-7.276 (SLES 9)
1306 2.4.21-47.0.1.EL (RHEL 3)
1307 2.6.12.6 vanilla (kernel.org)
1308 2.6.16.21-0.8 (SLES10)
1309 * Recommended e2fsprogs version: 1.39.cfs2-0
1311 * The backwards-compatible /proc/sys/portals symlink has been removed
1312 in this release. Before upgrading, please ensure that you change
1313 any configuration scripts or /etc/sysctl.conf files that access
1314 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1315 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1316 in advance of the upgrade on any system running Lustre 1.4.6 or
1317 newer, since /proc/sys/lnet was added in that version.
1318 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1322 Frequency : only when quota is used
1324 Description: avoid scanning export list for quota master
1325 Details : Change the algorithms to avoid scanning export list in order
1326 to improve the efficiency.
1329 Frequency : MDS failover only, very rarely
1331 Description: "went back in time" messages on mds failover
1332 Details : The greatest transno may be lost when the current operation
1333 finishes with an error (transno==0) and the client's last_rcvd
1334 record is over-written. Save the greatest transno in the
1335 mds_last_transno for this case.
1338 Frequency : always for specific kernels and striping counts
1340 Description: client may get "Matching packet too big" without ACL support
1341 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1342 when trying to access files in certain configurations. The
1343 clients should in fact be denied when mounting because they do
1344 not understand ACLs.
1347 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1349 Description: many clients connecting with IO in progress causes connect timeouts
1350 Details : Avoid synchronous journal commits to avoid delays caused by many
1351 clients connecting/disconnecting when bulk IO is in progress.
1352 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1353 OST_IO_PORTAL to avoid delays behind potentially many pending
1357 Frequency : occasionally with multiple writers to a single file
1359 Description: shared writes to file may result in wrong size reported by stat()
1360 Details : Allow growing of kms when extent lock is cancelled
1363 Frequency : always with random mmap IO to multi-striped file
1365 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1366 Details : Only if the hole is at the end of OST object so that kms is too
1367 small. Fix is to increase kms accordingly in ll_nopage.
1370 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1372 Description: writes to a missing object would leak memory on the OST
1373 Details : If there is an inconsistency between the MDS and OST filesystems,
1374 such that the MDS references an object that doesn't exist, writes
1375 to that object will leak memory due to incorrect cleanup in the
1376 error handling path, eventually running out of memory on the OST.
1381 Description: Creating too long symlink causes lustre errors
1382 Details : Check symlink and name lengths before sending requests to MDS.
1385 Frequency : only if flock is enabled (not on by default)
1387 Description: posix locks not released on fd closure on 2.6.9+
1388 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1389 kernels, this caused such locks not to be released on fd close and
1390 then assertions on fs unmount about still used locks.
1393 Frequency : MDS failover only, very rarely
1395 Description: clients may get ASSERTION(granted_lock != NULL)
1396 Details : When request was taking a long time, and a client was resending
1397 a getattr by name lock request. The were multiple lock requests
1398 with the same client lock handle and
1399 mds_getattr_name->fixup_handle_for_resent_request found one of the
1400 lock handles but later failed with ASSERTION(granted_lock != NULL).
1405 Description: handle->h_buffer_credits > 0, assertion failure
1406 Details : h_buffer_credits is zero after truncate, causing assertion
1407 failure. This patch extends the transaction or creates a new
1411 Frequency : NFS re-export or patchless client
1412 Bugzilla : 11179, 10796
1413 Description: Crash on NFS re-export node (__d_move)
1414 Details : We do not want to hash the dentry if we don't have a lock.
1415 But if this dentry is later used in d_move, we'd hit uninitialised
1416 list head d_hash, so we just do this to init d_hash field but
1417 leave dentry unhashed.
1420 Frequency : NFS re-export or patchless client
1422 Description: NFS exports has problem with symbolic link
1423 Details : lustre client didn't properly install dentry when re-exported
1424 to NFS or running patchless client.
1427 Frequency : NFS re-export or patchless client
1429 Description: Various nfs/patchless fixes.
1430 Details : fixes reuse disconected alias for lookup process - this fixes
1431 warning "find_exported_dentry: npd != pd",
1432 fix permission error with open files at nfs.
1433 fix apply umask when do revalidate.
1436 Frequency : occasional
1438 Description: Crash on NFS re-export node
1439 Details : calling clear_page() on the wrong pointer triggered oops in
1440 generic_mapping_read().
1443 Frequency : rarely, using O_DIRECT IO
1445 Description: unaligned directio crashes client with LASSERT
1446 Details : check for unaligned buffers before trying any requests.
1449 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1451 Description: stale data returned from RAID cache
1452 Details : If only a small amount of IO is done to the RAID device before
1453 reading it again it is possible to get stale data from the RAID
1454 cache instead of reading it from disk.
1457 Frequency : always for sles10 kernel
1459 Description: sles10 support
1460 Details : ll_follow_link: compile fixes and using of nd_set_link
1461 under newer kernels.
1464 Frequency : depends on arch, kernel and compiler version, always on sles10
1467 Description: recursive or deep enough symlinks cause stack overflow
1468 Details : getting rid of large stack-allocated variable in
1472 Frequency : depends on hardware
1474 Description: lustre write performance loss in the SLES10 kernel
1475 Details : the performance loss is caused by using of write barriers in the
1476 ext3 code. The SLES10 kernel turns barrier support on by
1477 default. The fix is to undo that change for ldiskfs.
1479 ------------------------------------------------------------------------------
1481 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1483 * Support for kernels:
1484 2.6.9-42.0.3EL (RHEL 4)
1485 2.6.5-7.276 (SLES 9)
1486 2.4.21-47.0.1.EL (RHEL 3)
1487 2.6.12.6 vanilla (kernel.org)
1491 Frequency : quota enabled and large files being deleted
1493 Description: releasing more than 4GB of quota at once hangs OST
1494 Details : If a user deletes more than 4GB of files on a single OST it
1495 will cause the OST to spin in an infinite loop. Release
1496 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1501 Description: statfs data retrieved from /proc may be stale or zero
1502 Details : When reading per-device statfs data from /proc, in the
1503 {kbytes,files}_{total,free,avail} files, it may appear
1504 as zero or be out of date.
1507 Frequency : systems with MD RAID1 external journal devices
1509 Description: lconf's call to blkid is confused by RAID1 journal devices
1510 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1511 of returning all block devices that match the journal UUID.
1514 Frequency : always, for aggregate stripe size over 4GB
1516 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1517 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1518 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1521 Frequency : always if "lfs find" used on a local file/directory
1523 Description: "lfs find" segfaults if used on a local file/directory
1524 Details : The case where a directory component was not specified wasn't
1525 handled correctly. Handle this properly.
1528 Frequency : always on ppc64
1530 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1531 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1532 perform endianness conversion.
1535 Frequency : rarely (truncate to non-zero file size after write under load)
1536 Bugzilla : 10730, 10687
1537 Description: Files padded with zeros to next 4K multiple
1538 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1539 it is possible that files that are truncated to a non-zero size
1540 immediately after being written are filled with zero bytes beyond
1541 the truncated size. No file data is lost.
1543 Severity : enhancement
1545 Description: Allow recovery/failover for liblustre clients.
1546 Details : liblustre clients were unaware of failover configurations until
1549 Severity : enhancement
1551 Description: user file locks should fail when not mounting with flock option
1552 Details : Set up an error-returning stub in ll_file_operations.lock field
1553 to prevent incorrect behaviour when client is mounted without
1554 flock option. Also, set up properly f_op->flock field for
1558 Frequency : always on ia64
1560 Description: "lfs df" loops on printing out MDS statfs information
1561 Details : The obd_ioctl_data was not initialized and in some systems
1562 this caused a failure during the ioctl that did not return
1563 an error. Initialize the struct and return an error on failure.
1566 Frequency : SLES 9 only
1568 Description: Error of copying files with lustre special EAs as root
1569 Details : Client side always return success for setxattr call for lustre
1570 special xattr (currently only "trusted.lov").
1573 Frequency : rarely on clusters with both ia64+i386 clients
1575 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1576 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1577 clients are doing concurrent writes to the same file it is
1578 possible that the ia64 clients may LASSERT because the OST
1579 extent locks are not PAGE_SIZE aligned. Ensure that grown
1580 locks are always aligned on the request boundary.
1583 Frequency : specific use, occasional
1585 Description: Overwriting in use executable truncates on-disk binary image
1586 Details : If one node attempts to overwrite an executable in use by
1587 another node, we now correctly return ETXTBSY instead of
1588 truncating the file.
1590 Severity : enhancement
1592 Description: Async OSC create to avoid the blocking unnecessarily.
1593 Details : If a OST has no remain object, system will block on the creating
1594 when need to create a new object on this OST. Now, ways use
1595 pre-created objects when available, instead of blocking on an
1596 empty osc while others are not empty. If we must block, we block
1597 for the shortest possible period of time.
1602 Description: chmod on Lustre root is propagated to other clients
1603 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1604 invalid by the follow_mount time.
1607 Frequency : liblustre clients only
1609 Description: Race in 'instant cancel' lock handling could lead to such locks
1610 never to be granted in case of SMP MDS
1611 Details : Do not destroy not yet granted but cbpending locks in
1615 Frequency : replay/resend of open
1617 Description: non null lock assetion failure in mds_intent_policy
1618 Details : Trying to replay/resend lockless open requests resulted in
1619 mds_open() returning 0 with no lock. Now it sets a flag if
1620 a lock is going to be returned.
1622 Severity : enhancement
1624 Description: Checksum enhancements
1625 Details : New checksum enhancements allow for resending RPCs that failed
1628 Severity : enhancement
1630 Description: Tunables on number of dirty pages in cacche
1631 Details : Allow to set limit on number of dirty pages cached.
1636 Description: client crash on unmount - lock still has references
1637 Details : In some error handling cases it was possible to leak a lock
1638 reference on a client while accessing a file. This was not
1639 harmful to the client during operation, but would cause the
1640 client to crash when the filesystem is unmounted.
1643 Frequency : specific case, rare
1645 Description: ETXTBSY on mds though file not in use by client
1646 Details : ETXTBSY is no longer incorrectly returned when attempting to
1647 chmod or chown a directory that the user previously tried to
1648 execute or a currently-executing binary.
1651 Frequency : extremely rare except on liblustre-based clients
1653 Description: Lustre space not freed when files are deleted
1654 Details : Clean up open-unlinked files after client eviction. Previously
1655 the unlink was skipped and the files remained as orphans.
1660 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1661 Details : In some cases it was possible to send a blocking callback to
1662 a client doing a glimpse, even though that client didn't get
1663 a lock granted. When the glimpse lock is cancelled on the OST
1664 the freed lock is left on the waiting list and corrupted the list.
1667 Frequency : all core dumps
1669 Description: Broke core dumps to lustre
1670 Details : Negative dentry may be unhashed if parent does not have UPDATE
1671 lock, but some callers, e.g. do_coredump, expect dentry to be
1672 hashed after successful create, hash it in ll_create_it.
1674 ------------------------------------------------------------------------------
1676 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1678 * Support for kernels:
1679 2.6.9-42.0.2.EL (RHEL 4)
1680 2.6.5-7.276 (SLES 9)
1681 2.4.21-40.EL (RHEL 3)
1682 2.6.12.6 vanilla (kernel.org)
1686 Frequency : always on RHEL 3
1688 Description: Number of open files grows over time
1689 Details : The number of open files grows over time, whether or not
1690 Lustre is started. This was due to a filp leak introduced
1691 by one of our kernel patches.
1693 ------------------------------------------------------------------------------
1695 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1697 * Support for kernels:
1698 2.6.9-42.EL (RHEL 4)
1699 2.6.5-7.276 (SLES 9)
1700 2.4.21-40.EL (RHEL 3)
1701 2.6.12.6 vanilla (kernel.org)
1706 Bugzilla : 5719, 9635, 9792, 9684
1707 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1708 Details : If a server is under heavy load and cannot reply to new
1709 connection requests before the client resends the (re)connect,
1710 the connection handling code can behave badly if two service
1711 threads are concurrently handing separate (re)connections from
1712 the same client. Add better locking to the connection handling
1713 code, and ensure that only a single connection will be processed
1714 for a given client UUID, even if the lock is dropped.
1716 Severity : enhancement
1718 Description: add TCP zero-copy support to kernel
1719 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1720 sends if the hardware supports scatter-gather and checksumming.
1721 This allows socklnd to do client-write and server-read more
1722 efficiently and reduce CPU utilization from skbuf copying.
1725 Frequency : only if NFS exporting from client
1727 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1728 Details : If setting attributes on a file created under NFS that had
1729 never been opened it would be possible to oops the client
1730 if the file had no objects.
1733 Frequency : always for liblustre
1735 Description: liblustre client does MDS+OSTs setattr RPC for each write
1736 Details : When doing a write from a liblustre client, the client
1737 incorrectly issued an RPC to the MDS and each OST the file was
1738 striped over in order to update the timestamps. When writing
1739 with small chunks and many clients this could overwhelm the MDS
1740 with RPCs. In all cases it would slow down the write because
1741 these RPCs are unnecessary.
1743 Severity : enhancement
1745 Description: allow number of MDS service threads to be changed at module load
1746 Details : It is now possible to change the number of MDS service threads
1747 running. Adding "options mds mds_num_threads={N}" to the MDS's
1748 /etc/modprobe.conf will set the number of threads for the next
1749 time Lustre is restarted (assuming the "mds" module is also
1750 reloaded at that time). The default number of threads will
1751 stay the same, 32 for most systems.
1756 Description: OST crash if filesystem is unformatted or corrupt
1757 Details : If an OST is started on a device that has never been formatted
1758 or if the filesystem is corrupt and cannot even mount then the
1759 error handling cleanup routines would dereference a NULL pointer.
1764 Description: NULL pointer deref in llap_from_page.
1765 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1766 incorrect) mapping if the page was truncated/reclaimed while it was
1767 searched for. Check for this condition and skip such pages when
1768 doing readahead. Introduce extra check to llap_from_page() to
1769 verify page->mapping->host is non-NULL (so page is not anonymous).
1772 Frequency : Sometimes when using sys_sendfile
1774 Description: "page not covered by a lock" warnings from ll_readpage
1775 Details : sendfile called ll_readpage without right page locks present.
1776 Now we introduced ll_file_sendfile that does necessary locking
1777 around call to generic_file_sendfile() much like we do in
1781 Frequency : with certain MDS communication failures at client mount time
1783 Description: NULL pointer deref after failed client mount
1784 Details : a client connection request may delayed by the network layer
1785 and not be sent until after the PTLRPC layer has timed out the
1786 request. If the client fails the mount immediately it will try
1787 to clean up before the network times out the request. Add a
1788 reference from the request import to the obd device and delay
1789 the cleanup until the network drops the request.
1792 Frequency : occasionally during client (re)connect
1794 Description: assertion failure during client (re)connect
1795 Details : processing a client connection request may be delayed by the
1796 client or server longer than the client connect timeout. This
1797 causes the client to resend the connection request. If the
1798 original connection request is replied in this interval, the
1799 client may trip an assertion failure in ptlrpc_connect_interpret()
1800 which thought it would be the only running connect process.
1803 Frequency : only with obd_echo servers and clients that are rebooted
1805 Description: kernel BUG accessing uninitialized data structure
1806 Details : When running an obd_echo server it did not start the ping_evictor
1807 thread, and when a client was evicted an uninitialized data
1808 structure was accessed. Start the ping_evictor in the RPC
1809 service startup instead of the OBD startup.
1811 Severity : enhancement
1812 Bugzilla : 10193 (patchless)
1813 Description: Remove dependency on various unexported kernel interfaces.
1814 Details : No longer need reparent_to_init, exit_mm, exit_files,
1815 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1818 Frequency : rare (only users of deprecated and unsupported LDAP config)
1820 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1821 Details : LDAP apparently contains 'lustreName' attributes instead of
1822 'name'. A simple remapping of the name is sufficient.
1825 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1827 Description: waiting_locks_callback trips kernel BUG if client is evicted
1828 Details : Running with the dump_on_timeout debug flag turned on makes
1829 it possible that the waiting_locks_callback() can try to dump
1830 the Lustre kernel debug logs from an interrupt handler. Defer
1831 this log dumping to the expired_lock_main() thread.
1833 Severity : enhancement
1835 Description: Support NFS exporting on 2.6 kernels.
1836 Details : Implement non-rawops metadata methods for NFS server to use without
1837 changing NFS server code.
1840 Frequency : very rare (synthetic metadata workload only)
1842 Description: two racing renames might cause an MDS thread to deadlock
1843 Details : Running the "racer" program may cause one MDS thread to rename
1844 a file from being the source of a rename to being the target of
1845 a rename at exactly the same time that another thread is doing
1846 so, and the second thread has already enqueued these locks after
1847 doing a lookup of the target and is trying to relock them in
1848 order. Ensure that we don't try to re-lock the same resource.
1851 Frequency : only very large systems with liblustre clients
1853 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1854 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1855 all OSTs in parallel. This allows the eviction of stale liblustre
1856 clients to proceed much faster than if they were done in series,
1857 and also offers similar improvements for other set_info RPCs.
1862 Description: excessive CPU usage during initial read phase on client
1863 Details : During the initial read phase on a client, it would agressively
1864 retry readahead on the file, consuming too much CPU and impacting
1865 performance (since 1.4.5.8). Improve the readahead algorithm
1866 to avoid this, and also improve some other common cases (read
1867 of small files in particular, where "small" is files smaller than
1868 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1873 Description: MDS crash when receiving packet with unknown intent.
1874 Details : Do not LBUG in unknown intent case, just return -EFAULT
1876 Severity : enhancement
1877 Bugzilla : 9293, 9385
1878 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1879 Details : Do not serialize getattr (non-intent version) and statfs.
1882 Frequency : occasional, when OST network is overloaded/intermittent
1884 Description: client evicted by OST after bulk IO timeout
1885 Details : If a client sends a bulk IO request (read or write) the OST
1886 may evict the client if it is unresposive to its data GET/PUT
1887 request. This is incorrect if the network is overloaded (takes
1888 too long to transfer the RPC data) or dropped the OST GET/PUT
1889 request. There is no need to evict the client at all, since
1890 the pinger and/or lock callbacks will handle this, and the
1891 client can restart the bulk request.
1894 Frequency : Always when mmapping file with no objects
1896 Description: client crashes when mmapping file with no objects
1897 Details : Check that we actually have objects in a file before doing any
1898 operations on objects in ll_vm_open, ll_vm_close and
1904 Description: Request leak when working with deleted CWD
1905 Details : Introduce advanced request refcount tracking for requests
1906 referenced from lustre intent.
1908 Severity : Enhancement
1910 Description: Cache open file handles on client.
1911 Details : MDS now will return special lock along with openhandle, if
1912 requested and client is allowed to hold openhandle, even if unused,
1913 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1914 closing files for every read/write openration.
1916 Severity : Enhancement
1918 Description: Cache open negative dentries on client when possible.
1919 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1920 negative dentries on lock revocation.
1925 Description: Remounting a client read-only wasn't possible with a zconf mount
1926 Details : It wasn't possible to remount a client read-only with llmount.
1928 Severity : enhancement
1929 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1930 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1931 MPICH is included to add Lustre-specific ADIO interfaces.
1932 This is based closely on the UFS ADIO layer and only differs
1933 in file creation, in order to allow the OST striping to be set.
1934 This is user-contributed code and not supported by CFS.
1939 Description: extended inode attributes (immutable, append-only) work improperly
1940 when 2.4 and 2.6 kernels are used on client/server or vice versa
1941 Details : Introduce kernel-independent values for these flags.
1943 Severity : enhancement
1946 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1947 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1948 granularity. Now, Fractional MB granularity is be supported,
1949 this is very useful for low memory system.
1951 Severity : enhancement
1953 Description: Getattr by fid
1954 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1955 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1956 have locks and avoids getattr rpc requests for it on every stat.
1959 Frequency : Always, for filesystems larger than 2TB
1961 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1962 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1963 filesystems larger than 2TB to be mounted with mballoc enabled.
1966 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1968 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1969 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1970 filesystems larger than 2TB is mounted then the kernel will
1971 silently corrupt the start of the filesystem. CONFIG_LBD is
1972 enabled for all CFS-supported kernels, but the possibility of
1973 this happening with a modified kernel config exists.
1975 Severity : enhancement
1977 Description: add client O_DIRECT support for 2.6 kernels
1978 Details : It is now possible to do O_DIRECT reads and writes to files
1979 in the Lustre client mountpoint on 2.6 kernel clients.
1981 Severity : enhancement
1983 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1984 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1985 parallel, not waiting for response from every OST before sending
1986 a rpc to the next OST.
1991 Description: setattr vs write race when updating file timestamps
1992 Details : Client processes that update a file timestamp into the past
1993 right after writing to the file (e.g. tar) it is possible that
1994 the updated file modification time can be reset to the current
1995 time due to a race between processing the setattr and write RPC.
1997 Severity : enhancement
1999 Description: Bring 'lfs find' closer in line with regular Linux find.
2000 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2001 -print0 options and obtains all the needed info through the lustre
2004 Severity : enhancement
2006 Description: support up to 1024 configured devices on one node
2007 Details : change obd_dev array from statically allocated to dynamically
2008 allocated structs as they are first used to reduce memory usage
2013 Description: Flush dirty partially truncated pages during truncate
2014 Details : Immediatelly flush partially truncated pages in filter_setattr,
2015 this way we completely avoid having any pages in page cache on OST
2016 and can retire ugly workarounds during writes to flush such pages.
2021 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2022 Details : protect inode from truncation within vfs_unlink() context
2023 just take a reference before calling vfs_unlink() and release it
2024 when parent's i_sem is free.
2029 Description: last_id value checked outside lock on OST caused LASSERT failure
2030 Details : If there were multiple MDS->OST object precreate requests in
2031 flight, it was possible that the OST's last object id was checked
2032 outside a lock and incorrectly tripped an assertion. Move checks
2033 inside locks, and discard old precreate requests.
2036 Frequency : always, if extents are used on OSTs
2038 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2039 Details : OSTs using the extents format would not zero the high 16 bits of
2040 the index physical block number. This is not a problem for any
2041 OST filesystems smaller than 16TB, and no kernels support ext3
2042 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2043 new/modified files) and can be fixed for existing filesystems
2044 with e2fsprogs-1.39-cfs1.
2049 Description: import connection selection may be incorrect if timer wraps
2050 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2051 import connections to be ignored if the 32-bit jiffies counter
2052 wraps. Use a 64-bit jiffies counter.
2055 Frequency : very large clusters immediately after boot
2057 Description: LNET request buffers exhausted under heavy short-term load
2058 Details : If a large number of client requests are generated on a service
2059 that has previously never seen so many requests it is possible
2060 that the request buffer growth cannot keep up with the spike in
2061 demand. Instead of dropping incoming requests, they are held in
2062 the LND until the RPC service can accept more requests.
2065 Frequency : Sometimes during replay
2067 Description: Assertion failure in ll_local_open after replay.
2068 Details : If replay happened on an open request reply before we were able
2069 to set replay handler, reply will become not swabbed tripping the
2070 assertion in ll_local_open. Now we set the handler right after
2071 recognising of open request
2074 Frequency : very rare
2076 Description: kernel reports "badness in vsnprintf"
2077 Details : Reading from the "recovery_status" /proc file in small chunks
2078 may cause a negative length in lprocfs_obd_rd_recovery_status()
2079 call to vsnprintf() (which is otherwise harmless). Exit early
2080 if there is no more space in the output buffer.
2082 Severity : enhancement
2084 Description: clear OBD RPC statistics by writing to them
2085 Details : It is now possible to clear the OBD RPC statistics by writing
2086 to the "stats" file.
2091 Description: Client mtime is not the same on different clients after utimes
2092 Details : In some cases, the client was using the utimes() syscall on
2093 a file cached on another node. The clients now validate the
2094 ctime from the MDS + OSTs to determine which one is right.
2099 Description: Inability to activate failout mode
2100 Details : lconf script incorrectly assumed that in pythong string's numeric
2101 value is used in comparisons.
2104 Frequency : always with multiple stripes per file
2106 Description: Inefficient object allocation for mutli-stripe files
2107 Details : When selecting which OSTs to stripe files over, for files with
2108 a stripe count that divides evenly into the number of OSTs,
2109 the MDS is always picking the same starting OST for each file.
2110 Return the OST selection heuristic to the original design.
2115 Description: mount failures may take full timeout to return an error
2116 Details : Under some heavy load conditions it is possible that a
2117 failed mount can wait for the full obd_timeout interval,
2118 possibly several minutes, before reporting an error.
2119 Instead return an error as soon as the status is known.
2121 Frequency : quota enabled and large files being deleted
2123 Description: releasing more than 4GB of quota at once hangs OST
2124 Details : If a user deletes more than 4GB of files on a single OST it
2125 will cause the OST to spin in an infinite loop. Release
2126 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2131 Description: statfs data retrieved from /proc may be stale or zero
2132 Details : When reading per-device statfs data from /proc, in the
2133 {kbytes,files}_{total,free,avail} files, it may appear
2134 as zero or be out of date.
2137 Frequency : systems with MD RAID1 external journal devices
2139 Description: lconf's call to blkid is confused by RAID1 journal devices
2140 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2141 of returning all block devices that match the journal UUID.
2144 Frequency : always, for aggregate stripe size over 4GB
2146 Description: assertion fails when trying to use 4GB stripe size
2147 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2148 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2151 Frequency : always on ppc64
2153 Description: the first write on an ext3 filesystem with mballoc got stuck
2154 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2155 perform endianness conversion.
2157 ------------------------------------------------------------------------------
2159 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2161 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2162 INTEROPERATE with older versions automatically. Please read the
2163 user documentation before upgrading any part of a live system.
2164 * WARNING: Lustre networking configuration changes are required with
2165 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2168 * Support for kernels:
2169 2.6.9-22.0.2.EL (RHEL 4)
2170 2.6.5-7.244 (SLES 9)
2171 2.6.12.6 vanilla (kernel.org)
2174 Severity : enhancement
2175 Bugzilla : 7981/8208
2176 Description: Introduced Lustre Networking (LNET)
2177 Details : LNET is new networking infrastructure for Lustre, it includes
2178 a reorganized network configuration mode (see the user
2179 documentation for full details) as well as support for routing
2180 between different network fabrics. Lustre Networking Devices
2181 (LNDs) for the supported network fabrics have also been
2182 created for this new infrastructure.
2184 Severity : enhancement
2185 Description: Introduced Access control lists
2186 Details : clients can set ACLs on files and directories in order to have
2187 more fine-grained permissions than the standard Unix UGO+RWX.
2188 The MDS must be started with the "-o acl" mount option.
2190 Severity : enhancement
2191 Description: Introduced filesystem quotas
2192 Details : Administrators may now establish per-user quotas on the
2195 Severity : enhancement
2197 Description: Configuration change for the XT3
2198 The PTLLND is now used to run Lustre over Portals on the XT3
2199 The configure option(s) --with-cray-portals are no longer used.
2200 Rather --with-portals=<path-to-portals-includes> is used to
2201 enable building on the XT3. In addition to enable XT3 specific
2202 features the option --enable-cray-xt3 must be used.
2207 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2208 Details : A race between allocating a new llcd and re-getting the llcd_lock
2209 allowed another thread to grab newly-allocated llcd.
2211 Severity : enhancement
2213 Description: 2.6 OST async journal commit and locking fix to improve performance
2214 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2215 2.6 kernels are now async as they already were in 2.4 kernels so
2216 that they can commit concurrently with the network bulk transfer.
2217 For block-allocated files the filter allocation semaphore is held
2218 to avoid filesystem fragmentation during allocation. BKL lock
2219 removed for 2.6 xattr operations where it is no longer needed.
2224 Description: lconf incorrectly determined whether two IP networks could talk
2225 Details : In some more complicated routing and multiple-network
2226 configurations, lconf will avoid trying to make a network
2227 connection to a disjoint part of the IP space. It was doing the
2228 math incorrectly for one set of cases.
2233 Description: Fix for potential infinite loop processing records in an llog.
2234 Details : If an llog record is corrupted/zeroed, it is possible to loop
2235 forever in llog_process(). Validate the llog record length
2236 and skip the remainder of the block on error.
2239 Frequency : occasional (liblustre only)
2241 Description: liblustre could not open files whose last component is a symlink
2242 Details : sysio_path_walk() would incorrectly pass the open intent to
2243 intermediate path components.
2246 Frequency : rare (liblustre only with non-standard tuning)
2247 Bugzilla : 7201 (7350)
2248 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2249 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2250 after it has been released, only do this for referenced locks
2252 Severity : enhancement
2254 Description: specifying an (invalid) directory default stripe_size of -1
2255 would reset the directory default striping
2256 Details : stripe_size -1 was used internally to signal directory stripe
2257 removal, now use "all default" to signal dir stripe removal
2258 as a directory striping of "all default" is not useful
2261 Frequency : common for large clusters running liblustre clients
2263 Description: doing an ls when liblustre clients are running is slow
2264 Details : sending a glimpse AST to a liblustre client waits for every AST
2265 to time out, as liblustre clients will not respond. Since they
2266 cannot cache data we refresh the OST lock LVB from disk instead.
2268 Severity : enhancement
2270 Description: doing an ls at the same time as file IO can be slow
2271 Details : enqueue and other "small" requests can be blocked behind many
2272 large IO requests. Create a new OST IO portal for non-IO
2273 requests so they can be processed faster.
2276 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2278 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2279 created file with mode 000
2280 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2281 allow an ftruncate() as a non-root user to a file with mode 000.
2282 Lustre now respects this flag to disable mode checks when
2283 truncating a file owned by the user
2286 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2289 Description: Revoking locks from clients that went dead or catatonic might take
2291 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2292 cancellation of such locks instant on servers without waiting for
2293 any reply from clients. Clients drops these locks when cancel
2294 notification from server is received without replying.
2297 Frequency : liblustre-only, when liblustre client dies or becomes busy
2299 Description: Doing ls on Linux clients can take a long time with active
2301 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2302 granting such locks to it in the first place if possible. Locks
2303 are taken by proxy on the OST during the read or write and
2304 dropped immediately afterward. Add connect flags handling, do
2305 not grant locks to liblustre clients for glimpse ASTs.
2307 Severity : enhancement
2309 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2310 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2311 sequential reads and behaves badly with applications doing random
2312 reads. Improve it by reading ahead at least read region, and
2313 avoiding excessive large RPC for small reads.
2315 Severity : enhancement
2317 Description: Creating more than 1000 files for a single job may cause a load
2318 imbalance on the OSTs if there are also a large number of OSTs.
2319 Details : qos_prep_create() uses an OST index reseed value that is an
2320 even multiple of the number of available OSTs so that if the
2321 reseed happens in the middle of the object allocation it will
2322 still utilize the OSTs as uniformly as possible.
2327 Description: OST or MDS may oops in ping_evictor_main()
2328 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2329 but doesn't restart at beginning of obd_exports_timed list
2332 Severity : enhancement
2334 Description: improve by-nid export eviction on the MDS and OST
2335 Details : allow multiple exports with the same NID to be evicted at one
2336 time without re-searching the exports list.
2339 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2341 Description: MDS may oops in groups_free()
2342 Details : in rare race conditions a newly allocated group_info struct is
2343 freed again, and this can be NULL. The 2.4 compatibility code
2344 for groups_free() checked for a NULL pointer, but 2.6 did not.
2347 Frequency : common for liblustre clients doing little filesystem IO
2348 Bugzilla : 9352, 7313
2349 Description: server may evict liblustre clients accessing contended locks
2350 Details : if a client is granted a lock or receives a completion AST
2351 with a blocking AST already set it would not reply to the AST
2352 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2355 Frequency : lfs setstripe, only systems with more than 160 OSTs
2357 Description: unable to set striping with a starting offset beyond OST 160
2358 Details : llapi_create_file() incorrectly limited the starting stripe
2359 index to the maximum single-file stripe count.
2362 Frequency : LDAP users only
2364 Description: lconf did not handle in-kernel recovery with LDAP properly
2365 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2367 Severity : enhancement
2369 Description: bind OST threads to NUMA nodes to improve performance
2370 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2371 node and do their allocations there to localize memory access
2373 Severity : enhancement
2375 Description: llmount can determine client NID directly from Myrinet (GM)
2376 Details : the client NID code from gmnalnid was moved directly into
2377 llmount, removing the need to use this or specifying the
2378 client NID explicitly when mounting GM clients with zeroconf
2381 Frequency : if client is started with down MDS
2383 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2384 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2385 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2386 ptlrpc_import_delayed_req() because we didn't check if the
2387 request was interrupted, and we also didn't break out of the
2388 event loop if there was no timeout
2393 Description: data loss during non-page-aligned writes to a single file from
2394 both multiple nodes and multiple threads on one node at same time
2395 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2396 inconsistency led to false short-reads, that were cached and later
2397 used by ->prepare_write() to fill in partially written page,
2398 leading to data loss.
2401 Frequency : always, if lconf --abort_recovery used
2403 Description: lconf --abort_recovery fails with 'Operation not supported'
2404 Details : lconf was attempting to abort recovery on the MDT device and not
2407 Severity : enhancement
2409 Description: remove cleanup logs
2410 Details : replace lconf-generated cleanup logs with lustre internal
2411 cleanup routines. Eliminates the need for client-cleanup and
2414 Severity : enhancement
2416 Description: add support for EAs (user and system) on lustre filesystems
2417 Details : it is now possible to store extended attributes in the Lustre
2418 client filesystem, and with the user_xattr mount option it
2419 is possible to allow users to store EAs on their files also
2421 Severity : enhancement
2423 Description: Add possibility (config option) to show minimal available OST free
2425 Details : When compiled with --enable-mindf configure option, statfs(2)
2426 (and so, df) will return least minimal free space available from
2427 all OSTs as amount of free space on FS, instead of summary of
2428 free spaces of all OSTs.
2430 Severity : enhancement
2432 Description: do not expand extent locks acquired on OST-side
2433 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2434 by server: they are not cached anyway.
2437 Frequency : when mmap is used/binaries executed from Lustre
2439 Description: Unmmap pages before throwing them away from read cache.
2440 Details : llap_shrink cache now attempts to unmap pages before discarding
2441 them (if unmapping failed - do not discard). SLES9 kernel has
2442 extra checks that trigger if this unmapping is not done first.
2447 Description: lconf didn't resolve symlinks before checking to see whether a
2448 given mountpoint was already in use
2451 Frequency : when migrating failover services
2452 Bugzilla : 6395, 9514
2453 Description: When migrating a subset of services from a node (e.g. failback
2454 from a failover service node) the remaining services would
2455 time out and evict clients.
2456 Details : lconf --force (implied by --failover) sets the global obd_timeout
2457 to 5 seconds in order to quickly disconnect, but this caused
2458 other RPCs to time out too quickly. Do not change the global
2459 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2461 Severity : enhancement
2462 Frequency : if MDS is started with down OST
2463 Bugzilla : 9439,5706
2464 Description: Allow startup/shutdown of an MDS without depending on the
2465 availability of the OSTs.
2466 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2467 Add appropriate locking and lov-osc refcounts for safe
2468 cleaning. Add osc abort_inflight calls in case the
2469 synchronize never started.
2472 Frequency : occasional (Cray XT3 only)
2474 Description: root not authorized to access files in CRAY_PORTALS environment
2475 Details : The client process capabilities were not honoured on the MDS in
2476 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2477 been accessed by an authorized user then root was able to access
2478 the file on the local client also. The root user capabilities
2479 are now allowed on the MDS, as this environment has secure UID.
2482 Frequency : occasional
2484 Description: ldiskfs "too long searching" message happens too often
2485 Details : A debugging message (otherwise harmless) prints too often on
2486 the OST console. This has been reduced to only happen when
2487 there are fragmentation problems on the filesystem.
2492 Description: Division by zero in statfs when all OSCs are inactive
2493 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2494 lov_statfs divides by value returned by lov_get_stripecnt().
2498 Bugzilla : 9489, 3273
2499 Description: First write from each client to each OST was only 4kB in size,
2500 to initialize client writeback cache, which caused sub-optimal
2501 RPCs and poor layout on disk for the first writen file.
2502 Details : Clients now request an initial cache grant at (re)connect time
2503 and so that they can start streaming writes to the cache right
2504 away and always do full-sized RPCs if there is enough data.
2505 If the OST is rebooted the client also re-establishes its grant
2506 so that client cached writes will be honoured under the grant.
2511 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2512 Details : Now I/O RPCs go to different portal number and (presumably) fast
2513 lock requests (and glimses) and other RPCs get their own service
2514 threads pool that should be able to service those RPCs
2517 Severity : enhancement
2519 Description: Ability to exchange lustre version between client and servers and
2520 issue warnings at client side if client is too old. Also for
2521 liblustre clients there is ability to refuse connection of too old
2523 Details : New 'version' field is added to connect data structure that is
2524 filled with version info. That info is later checked by server and
2528 Frequency : rare, liblustre only.
2529 Bugzilla : 9296, 9581
2530 Description: Two simultaneous writes from liblustre at offset within same page
2531 might proceed at the same time overwriting eachother with stale
2533 Details : I/O lock withing llu_file_prwv was released too early, before data
2534 actually was hitting the wire. Extended lock-holding time until
2535 server acknowledges receiving data.
2538 Frequency : extremely rare. Never observed in practice.
2540 Description: avoid generating lustre_handle cookie of 0.
2541 Details : class_handle_hash() generates handle cookies by incrementing
2542 global counter, and can hit 0 occasionaly (this is unlikely, but
2543 not impossible, because initial value of cookie counter is
2544 selected randonly). Value of 0 is used as a sentinel meaning
2545 "unassigned handle" --- avoid it. Also coalesce two critical
2546 sections in this function into one.
2548 Severity : enhancement
2550 Description: allow liblustre clients to delegate truncate locking to OST
2551 Details : To avoid overhead of locking, liblustre client instructs OST to
2552 take extent lock in ost_punch() on client's behalf. New connection
2553 flag is added to handle backward compatibility.
2555 Severity : enhancement
2556 Bugzilla : 4928, 7341, 9758
2557 Description: allow number of OST service threads to be specified
2558 Details : a module parameter allows the number of OST service threads
2559 to be specified via "options ost ost_num_threads={N}" in the
2560 OSS's /etc/modules.conf or /etc/modprobe.conf.
2564 Bugzilla : 6146, 9635, 9895
2565 Description: servers crash with bad pointer in target_handle_connect()
2566 Details : In rare cases when a client is reconnecting it was possible that
2567 the connection request was the last reference for that export.
2568 We would temporarily drop the export reference and get a new
2569 one, but this may have been the last reference and the export
2570 was just destroyed. Get new reference before dropping old one.
2572 Severity : enhancement
2573 Frequency : if client is started with failover MDS
2575 Description: Allow multiple MDS hostnames in the mount command
2576 Details : Try to read the configuration from all specified MDS
2577 hostnames during a client mount in case the "primary"
2580 Severity : enhancement
2582 Description: Stop sending data to evicted clients as soon as possible.
2583 Details : Check if the client we are about to send or are sending data to
2584 was evicted already. (Check is done every second of waiting,
2585 for which l_wait_event interface was extended to allow checking
2586 of exit condition at specified intervals).
2589 Frequency : rare, normally only when NFS exporting is done from client
2591 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2593 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2596 Frequency : always, if srand() is called before liblustre initialization
2598 Description: Liblustre uses system PRNG disturbing its usage by user application
2599 Details : Introduce internal to lustre fast and high-quality PRNG for
2600 lustre usage and make liblustre and some other places in generic
2601 lustre code to use it.
2603 Severity : enhancement
2604 Bugzilla : 9477, 9557, 9870
2605 Description: Verify that the MDS configuration logs are updated when xml is
2606 Details : Check if the .xml configuration logs are newer than the config
2607 logs stored on the MDS and report an error if this is the case.
2608 Request --write-conf, or allow starting with --old_conf.
2610 Severity : enhancement
2612 Description: Handle symlinks in the path when checking if Lustre is mounted.
2613 Details : Resolve intermediate symlinks when checking if a client has
2614 mounted a filesystem to avoid duplicate client mounts.
2619 Description: lconf can hit an error exception but still return success.
2620 Details : The lconf command catches the Command error exception at the top
2621 level script context and will exit with the associated exit
2622 status, but doesn't ensure that this exit status is non-zero.
2627 Description: failure of ptlrpc thread startup can cause oops
2628 Details : Starting a ptlrpc service thread can fail if there are a large
2629 number of threads or the server memory is very fragmented.
2630 Handle this without oopsing.
2633 Frequency : always, only if liblustre and non-default acceptor port was used
2635 Description: liblustre cannot connect to servers with non-default acceptor port
2636 Details : tcpnal_set_default_params() was not called and was therefore
2637 ignoring the environment varaible TCPNAL_PORT, as well as other
2638 TCPNAL_ environment variables
2643 Description: two objects could be created on the same OST for a single file
2644 Details : If an OST is down, in some cases it was possible to create two
2645 objects on a single OST for a single file. No problems other
2646 than potential performance impact and spurious error messages.
2650 Bugzilla : 5681, 9562
2651 Description: Client may oops in ll_unhash_aliases
2652 Details : Client dcache may become inconsistent in race condition.
2653 In some cases "getcwd" can fail if the current directory is
2659 Description: Inode refcounting problems in NFS export code
2660 Details : link_raw functions used to call d_instantiate without obtaining
2661 extra inode reference first.
2665 Bugzilla : 9942, 9903
2666 Description: Referencing freed requests leading to crash, memleaks with NFS.
2667 Details : We used to require that call to ll_revalidate_it was always
2668 followed by ll_lookup_it. Also with revalidate_special() it is
2669 possible to call ll_revalidate_it() twice for the same dentry
2670 even if first occurence returned success. This fix changes semantic
2671 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2672 reference on a request referred from the intent.
2673 ll_intent_release() then releases such a request.
2676 Frequency : rare, normally benchmark loads only
2678 Description: unlinked inodes were kept in memory on the client
2679 Details : If a client is repeatedly creating and unlinking files it
2680 can accumulate a lot of stale inodes in the inode slab cache.
2681 If there is no other client load running this can cause the
2682 client node to run out of memory. Instead flush old inodes
2683 from client cache that have the same inode number as a new inode.
2686 Frequency : SLES9 2.6.5 kernel and long filenames only
2687 Bugzilla : 9969, 10379
2688 Description: utime reports stale NFS file handle
2689 Details : SLES9 uses out-of-dentry names in some cases, which confused
2690 the lustre dentry revalidation. Change it to always use the
2694 Frequency : rare, unless heavy write-truncate concurrency is continuous
2695 Bugzilla : 4180, 6984, 7171, 9963, 9331
2696 Description: OST becomes very slow and/or deadlocked during object unlink
2697 Details : filter_destroy() was holding onto the parent directory lock
2698 while truncating+unlinking objects. For very large objects this
2699 may block other threads for a long time and slow overall OST
2700 responsiveness. It may also be possible to get a lock ordering
2701 deadlock in this case, or run out of journal credits because of
2702 the combined truncate+unlink. Solution is to do object truncate
2703 first in one transaction without parent lock, and then do the
2704 final unlink in a new transaction with the parent lock. This
2705 reduces the lock hold time dramatically.
2708 Frequency : rare, 2.4 kernels only
2710 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2711 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2712 before unmounting their filesystem in order to avoid deadlock.
2713 The kernel_locked() function in 2.4 kernels only checks whether
2714 the kernel lock is held, not whether it is this process that is
2715 holding it as 2.6 kernels do.
2720 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2721 Details : The client ptlrpc code may be trying to reconnect to a down
2722 server before a previous connection attempt has timed out.
2723 Increase the reconnect interval to be longer than the connection
2724 timeout interval to avoid sending duplicate connections to
2728 Frequency : echo_client brw_test command
2730 Description: fix echo_client to work with OST preallocated code
2731 Details : OST preallocation code (5137) didn't take echo_client IO path
2732 into account: echo_client calls filter methods outside of any
2733 OST thread and, hence, there is no per-thread preallocated
2734 pages and buffers to use. Solution: hijack pga pages for IO. As
2735 a byproduct, this avoids unnecessary data copying.
2739 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2740 Description: Client can oops in mdc_commit_close() after open replay
2741 Details : It was possible for the MDS to return an open request with no
2742 transaction number in mds_finish_transno() if the client was
2743 evicted, but without actually returning an error. Clients
2744 would later try to replay that open and may trip an assertion
2745 Simplify the client close codepath, and always return an error
2746 from the MDS in case the open is not successful.
2749 Frequency : rare, 2.6 OSTs only
2751 Description: OST may deadlock under high load on fragmented files
2752 Details : If there was a heavy load and highly-fragmented OST filesystems
2753 it was possible to have all the OST threads deadlock waiting on
2754 allocation of biovecs, because the biovecs were not released
2755 until the entire RPC IO was completed. Instead, release biovecs
2756 as soon as they are complete to ensure forward IO progress.
2758 Severity : enhancement
2760 Description: Support for specifying external journal device at mount
2761 Details : If an OST or MDS device is formatted with an external journal
2762 device, this device major/minor is stored in the ext3 superblock
2763 and may not be valid for failover. Allow detecting and
2764 specifying the external journal at mount time.
2769 Description: Mounting an MDS with pending unlinked files may cause oops
2770 Details : target_finish_recovery() calls mds_postrecov() which returned
2771 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2772 considers this an error and immediately begins cleaning up the
2773 lov, just after starting the mds_lov process
2775 Severity : enhancement
2777 Description: Implement 'lfs df' to report actual free space on per-OST basis
2778 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2779 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2780 usage of objects; '-h' to report in human readable format.
2782 ------------------------------------------------------------------------------
2784 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2791 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2792 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2793 at startup during orphan unlinking. Instead, skip buddy update
2794 before setup, as it will be regenerated after recovery is complete.
2799 Description: If an OST is inactive, its locks might reference stale inodes.
2800 Details : lov_change_cbdata() must iterate over all namespaces, even if
2801 they are inactive to clear inode references from the lock.
2803 Severity : enhancement
2804 Frequency : occasional, if non-standard max_dirty_mb used
2806 Description: Client will block write RPCs if not enough grant
2807 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2808 then the client will block writes while waiting for another RPC
2809 to complete instead of consuming its dirty limit. With change
2810 we get improved performance when max_dirty_mb is small.
2812 Severity : enhancement
2813 Bugzilla : 3389, 6253
2814 Description: Add support for supplementary groups on the MDS.
2815 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2816 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2817 lookups for user supplementary groups into a cache.
2821 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2822 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2823 for liblustre O_EXCL is enforced by MDS.
2827 Description: readdir never returns NULL in liblustre.
2828 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2829 dentry in struct dirent.
2833 Frequency : liblustre only, depends on application IO pattern
2834 Description: liblustre clients evicted if not contacting servers
2835 Details : Don't put liblustre clients into the ping_evictor list, so
2836 they will not be evicted by the pinger ever.
2838 Severity : enhancement
2840 Description: Add ability to evict clients by NID from MDS.
2841 Details : By echoing "nid:$NID" string into
2842 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2843 $NID would be instantly evicted from this MDS and from all active
2844 OSTs connected to it.
2848 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2849 Details : lookup_it_finish() used to query file size from OSTs that was not
2854 Description: service threads change working directory to that of init
2855 Details : Starting lustre service threads may pin the working directory
2856 of the parent thread, making that filesystem busy. Threads
2857 now change to the working directory of init to avoid this.
2861 Frequency : during shutdown only
2862 Description: shutdown with a failed MDS or OST can cause unmount to hang
2863 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2868 Frequency : 2.6 only
2869 Description: chmod/chown may include an extra supplementary group
2870 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2871 supplementary group and if none is specified this is used.
2874 Bugzilla : 5479 (6816)
2876 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2877 Details : If lookup is in progress on a file that is unlinked we might try
2878 to revalidate the inode and fail in revalidate after lookup is
2879 complete and ll_file_open() enqueues the open again but
2880 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2883 Frequency : always, if lconf --abort_recovery used
2885 Description: lconf --abort_recovery fails with 'Operation not supported'
2886 Details : lconf was attempting to abort recovery on the MDT device and not
2889 ------------------------------------------------------------------------------
2891 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2896 Frequency : rare (only unsupported configurations with a node running as an
2898 Bugzilla : 6514, 5137
2899 Description: Mounting a Lustre file system on a node running as an OST could
2901 Details : OSTs now preallocates memory needed to write out data at
2902 startup, instead of when needed, to avoid having to
2903 allocate memory in possibly low memory situations.
2904 Specifically, if the file system is mounted on on OST,
2905 memory pressure could force it to try to write out data,
2906 which it needed to allocate memory to do. Due to the low
2907 memory, it would be unable to do so and the node would
2908 become unresponsive.
2910 Severity : enhancement
2912 Description: Addition of lconf --service command line option
2913 Details : lconf now accepts a '--service <arg>' option, which is
2914 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2916 Severity : enhancement
2918 Description: Failover mode is now the default for OSTs.
2919 Details : By default, OSTs will now run in failover mode. To return to
2920 the old behaviour, add '--failout' to the lmc line for OSTs.
2922 Severity : enhancement
2924 Description: Health checks are now provided for MDS and OSTs
2925 Details : Additional detailed health check information on MSD and OSTs
2926 is now provided through the procfs health_check value.
2929 Frequency : occasional, depends on IO load
2931 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2932 after numerous create/delete cycles
2933 Details : The ext3 inode allocation policy would not allocate new inodes
2934 very well on the OSTs because there are no new directories
2935 being created. Instead we look for groups with free space if
2936 the parent directories are nearly full.
2941 Description: Network or server problems during mount may cause partially
2942 mounted clients instead of returning an error.
2943 Details : The config llog parsing code may overwrite the error return
2944 code during mount error handling, returning success instead
2950 Description: MDS can fail to allocate large reply buffers
2951 Details : After long uptimes the MDS can fail to allocate large reply
2952 buffers (e.g. zconf client mount config records) due to memory
2953 fragmentation or consumption by the buffer cache. Preallocate
2954 some large reply buffers so that these replies can be sent even
2955 under memory pressure.
2959 Frequency : rare (liblustre)
2960 Description: fsx running with liblustre complained that using truncate() to
2961 extend the file doesn't work. This patch corrects that issue.
2962 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2963 ATTR_SIZE and lsm use in llu_setattr_raw.
2967 Frequency : rare, only 2.6 kernels
2968 Description: Unusual file access patterns on the MDS may result in inode
2969 data being lost in very rare circumstances.
2970 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2971 code in the 2.6 kernel caused the inode and/or EA data not to
2972 be read from disk, causing single-file corruption.
2976 Frequency : rare, only 2.6 filesystems using extents
2977 Description: Heavy concurrent write and delete load may cause data corruption.
2978 Details : It was possible under high-load situations to have an extent
2979 metadata block in the block device cache from a just-unlinked
2980 file overwrite a newly-allocated data block. We now unmap any
2981 metadata buffers that alias just-allocated data blocks.
2985 Frequency : filesystems with default stripe_count larger than 77
2986 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2987 Details : lconf specifies an inode size of 4096 bytes when the default
2988 stripe_count is larger than 77. This conflicts with the default
2989 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2990 this case to avoid pinning too much memory for large EAs.
2992 ------------------------------------------------------------------------------
2994 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
2999 Frequency : rare (extremely heavy IO load with hundreds of clients)
3001 Description: Client is evicted, gets IO error writing to file
3002 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3003 caused clients to be evicted for AST timeouts. The fixes in
3004 bug 5192 mean we no longer need to have such short AST timeouts
3005 so ldlm_timeout has been increased.
3008 Frequency : occasional during --force or --failover shutdown under load
3009 Bugzilla : 5949, 4834
3010 Description: Server oops/LBUG if stopped with --force or --failover under load
3011 Details : a collection of import/export refcount and cleanup ordering
3012 issues fixed for safer force cleanup
3015 Frequency : only filesystems larger than 120 OSTs
3016 Bugzilla : 5990, 6223
3017 Description: lfs getstripe would oops on a very large filesystem
3018 Details : lov_getconfig used kfree on vmalloc'd memory
3021 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3022 Bugzilla : 6242, 6243
3023 Description: reading from files that had been truncated to a non-zero size
3024 but never opened returned no data
3025 Details : ll_file_read() reads zeros from no-object files to EOF
3030 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3031 Details : There's little harm in aborting MDS/OSS recovery and letting it
3032 try again, so I removed the LASSERT and return an error instead.
3034 Severity : enhancement
3036 Description: New debugging infrastructure for tracking down data corruption
3037 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3038 (b) cover more of the client-side code path, and (c) try to narrow
3039 down where problems occurred
3043 Bugzilla : 3819, 4364, 4397, 6313
3044 Description: Racing close and eviction MDS could cause assertion in mds_close
3045 Details : It was possible to get multiple mfd references during close and
3046 client eviction, leading to one thread referencing a freed mfd.
3048 Severity: : enhancement
3049 Bugzilla : 3262, 6359
3050 Description: Attempts to reconnect to servers are now more aggressive.
3051 Details : This builds on the enhanced upcall-less recovery that was added
3052 in 1.4.2. When trying to reconnect to servers, clients will
3053 now try each server in the failover group every 10 seconds. By
3054 default, clients would previously try one server every 25 seconds.
3059 Description: After recovery, certain operations trigger a failed
3060 assertion on a client.
3061 Details : Failing over an mds, using lconf -d --failover, while a
3062 client was doing a readdir() call would cause the client to
3063 LBUG after recovery completed and the readdir() was resent.
3065 Severity : enhancement
3067 Description: Default groups are now added by lconf
3068 Details : You can now run lconf --group <servicename> without having to
3069 manually add groups with lmc.
3072 Frequency : occasional
3074 Description: Nodes with an elan id of 0 trigger a failed assertion
3077 Frequency : always when accessing e.g. tty/console device nodes
3079 Description: tty and some other devices nodes cannot be used on lustre
3080 Details : file's private_data field is used by device data and lustre
3081 values in there got lost. New field was added to struct file to
3082 store fs-specific private data.
3085 Frequency : when exporting Lustre via NFS
3087 Description: NFSD failed occasionally when looking up a path component
3088 Details : NFSD is looking up ".." which was broken in ext3 directories
3089 that had grown large enough to become hashed.
3092 Frequency : Clusters with multiple interfaces not on the same subnet
3094 Description: Nodes will repeatedly try to reconnect to an interface which it
3095 cannot reach and report an error to the log.
3096 Details : Extra peer list entries will be created by lconf with some peers
3097 unreachable. lconf now validates the peer before adding it.
3100 Frequency : Only if a default stripe is set on the filesystem root.
3102 Description: Setting a default stripe on the filesystem root prevented the
3103 filesystem from being remounted.
3104 Details : The client was sending extra request flags in the root getattr
3105 request and did not allocate a reply buffer for the dir EA.
3108 Frequency : occasional, higher if lots of files are accessed by one client
3109 Bugzilla : 6159, 6097
3110 Description: Client trips assertion regarding lsm mismatch/magic
3111 Details : While revalidating inodes the VFS looks up inodes with ifind()
3112 and in rare cases can find an inode that is being freed.
3113 The ll_test_inode() code will free the lsm during ifind()
3114 when it finds an existing inode and then the VFS later attaches
3115 this free lsm to a new inode.
3119 Bugzilla : 6422, 7030
3120 Description: MDS deadlock between mkdir and client eviction
3121 Details : Creating a new file via mkdir or mknod (starting a transaction
3122 and getting the ns lock) can deadlock with client eviction
3123 (gets ns lock and trying to finish a synchronous transaction).
3126 Frequency : occasional
3127 Description: While starting a server, the fsfilt_ext3 module could not be
3129 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3130 kernels. Previously, lconf would still use the ext3 name
3131 when trying to load modules. Now, it will correctly use
3132 ext3 on 2.4 and ldiskfs on 2.6.
3134 Severity : enhancement
3135 Description: The default stripe count has been changed to 1
3136 Details : The interpretation of the default stripe count (0, to lfs
3137 or lmc) has been changed to mean striping across a single
3138 OST, rather than all available. For general usage we have
3139 found a stripe count of 1 or 2 works best.
3141 Severity : enhancement
3142 Description: Add support for compiling against Cray portals.
3143 Details : Conditional compiling for some areas that are different
3147 Frequency : occasional
3148 Bugzilla : 6409, 6834
3149 Description: Creating files with an explicit stripe count may lead to
3150 a failed assertion on the MDS
3151 Details : If some OSTs are full or unavailable, creating files may
3152 trigger a failed assertion on the MDS. Now, Lustre will
3153 try to use other servers or return an error to the
3157 Frequency : occasional
3159 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3160 caused lock timeout during readahead (since 1.4.2).
3161 Details : Processes doing readahead might match a lock that hasn't been
3162 granted yet if there are overlapping and conflicting lock
3163 requests. The readahead process waits on ungranted lock
3164 (original lock is CBPENDING), while OST waits for that process
3165 to cancel CBPENDING read lock and eventually evicts client.
3167 Severity : enhancement
3169 Description: Initial enabling of flock support for clients
3170 Details : Implements fcntl advisory locking and file status functions.
3171 This feature is provided as an optional mount flag (default
3172 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3173 locking are implemented yet, and those that are are not guaranteed
3174 to be completely correct in production environments.
3175 mount -t lustre -o [flock|noflock] ...
3178 Frequency : occasional
3180 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3181 assertion in the ext3 JBD (journaling) layer.
3182 Details : The b_committed_data struct is protected by the big kernel lock
3183 in 2.4 kernels, serializing journal_commit_transaction() and
3184 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3185 access to this struct. In 2.6 kernels there is finer grained
3186 locking to improve SMP performance of the JBD layer.
3190 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3195 Description: merge and cleanup kernel patches.
3198 -----------------------------------------------------------------------------
3200 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3202 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3203 versions of Lustre. Please update all servers and clients to
3204 version 1.4.2 or later at the same time. You must also run
3205 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3206 to update the configuration logs.
3208 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3209 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3210 - swab llog records of type '0' so we get proper header size/idx (5861)
3211 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3212 - fix rename of one directory over another leaking an inode (5953)
3213 - avoid SetPageDirty on 2.6 (5981)
3214 - don't re-add just-being-destroyed locks to the waiting list (5653)
3215 - when creating new directories, inherit the parent's custom
3216 striping settings if present parent (3048)
3217 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3218 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3219 - don't hold client locks on temporary worklist from l_lru (5666)
3220 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3221 - automatically evict dead clients (5921)
3222 - Update file size properly in create+truncate+fstat case (6196)
3223 - Do not unhash mountpoint dentries, do not allow removal of
3225 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3226 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3227 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3228 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3229 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3230 - Correctly handle reads of files with no objects (6243)
3231 - lctl recover will also mark a device active if deactivate used (5933)
3233 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3234 - allow --write-conf on an MDS with different nettype than client (5619)
3235 - don't write config llogs to MDS for mounts not from that MDS (5617)
3236 - lconf should create multiple TCP connections from a client (5201)
3237 - init scripts are now turned off by default; run chkconfig --on
3238 lustre and chkconfig --on lustrefs to use them
3239 - upcalls are no longer needed for clients to recover to failover
3241 - add --abort-recovery option to lconf to abort recovery on device
3243 - add support for an arbitrary number of OSTs (3026)
3244 - Quota support protocol changes.
3245 - forward compatibility changes to wire structs (6007)
3246 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3247 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3248 - improved "lustre status" script
3249 - initialize blocksize for non-regular files (6062)
3250 - added --disable-server and --disable-client configure options (5782)
3251 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3252 - Vanilla 2.4.29 support
3253 - increase maximum number of obd devices to 520 (6242)
3254 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3255 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3256 SLES 9 2.6.5 kernels
3257 - limit stripes per file to 160 (the maximum EA size) (6093)
3259 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3262 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3263 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3264 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3265 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3266 - small ext3 extents cleanups and fixes (5733)
3267 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3268 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3269 - avoid needless client->OST connect, fix handle mismatch (5317)
3270 - fix DLM error path that led to out-of-sync client, long delays (5779)
3271 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3272 - fix several locking issues related to i_size (5492,5624,5654,5672)
3273 - don't move pending lock onto export if it is already evicted (5683)
3274 - fix kernel oops when creating .foo in unlinked directory (5548)
3275 - fix deadlock in obdfilter statistics vs. object create (5811)
3276 - use time_{before,after} to avoid timer jiffies wrap (5882)
3277 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3278 - Do not leak request if server was not able to process it (5154)
3279 - If mds_open unable to find parent dir, make that negative lookup(5154)
3280 - don't create new directories with extent-mapping (5909, 5936)
3282 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3283 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3284 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3285 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3286 - support for RHEL4 kernel on i686 (5773)
3287 - provide error messages when incompatible logs are encountered (5898)
3289 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3290 * version 1.4.0.10 (1.4.1 release candidate 1)
3292 - don't keep a lock reference when lock is not granted (4238)
3293 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3294 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3295 - drop import inflight refcount on signal_completed_replay error (5255)
3296 - unlock page after async write error during send (3677)
3297 - handle missing objects in filter_preprw_read properly (5265)
3298 - no transno return for symlink open, don't save no-trasno open (3440)
3299 - don't try to complete elan receive that already failed (4012)
3300 - free RPC server reply state on error (5406)
3301 - clean up thread from ptlrpc_start_thread() on error (5160)
3302 - readahead could read extra page into cache that wasn't ejected (5388)
3303 - prevent races in class_attach/setup/cleanup/detach (5260)
3304 - don't dereference de->d_inode after l_dput of de (5458)
3305 - use "int" for stripe value returned from lock_to_stripe (5544)
3306 - mballoc allocation and error-checking fixes in 2.6 (5504)
3307 - block device patches to fix I/O request sizes in 2.6 (5482)
3308 - look up hostnames for IB nals (5602)
3309 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3310 - don't start multiple acceptors for the same port (5277)
3311 - fix incorrect LASSERT in mds_getattr_name (5635)
3312 - export a proc file for general "ping" checking (5628)
3313 - fix "lfs check" to not block when the MDS is down (5628)
3315 - service request history (4965)
3316 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3317 - create an "evict_client" /proc entry on OSTs, like the MDS has
3318 - fix mount usage message, return errors per mount(8) (5168)
3319 - change grep [] to grep "[]" in tests so they work in more UMLs
3320 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3321 - remove ancient LOV_MAGIC_V0 EA support (5047)
3322 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3323 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3324 - allow readahead allocations to fail when low on memory (5383)
3325 - mmap locking landed again, after considerable improvement (2828)
3326 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3328 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3331 - send OST transaction number in read/write reply to free req (4966)
3332 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3333 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3334 - the watchdog thread now runs as interruptible (5246)
3335 - flock/lockf fixes (but it's still disabled, pending 5135)
3336 - don't use EXT3 constants in llite code (5094)
3337 - memory shortage at startup could cause assertion (5176)
3339 - reorganization of lov code
3340 - single portals codebase
3342 - add extents/mballoc support (5025)
3343 - direct I/O reads in the obdfilter (4048)
3344 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3346 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3349 - send OST transaction number in read/write reply to free req (4966)
3350 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3351 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3352 - the watchdog thread now runs as interruptible (5246)
3353 - handle missing objects in filter_preprw_read properly (5265)
3354 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3355 - drop import inflight refcount on signal_completed_replay error (5255)
3356 - unlock page after async write error during send (3677)
3357 - return original error code on reconstructed replies (3761)
3358 - no transno return for symlink open, don't save no-trasno open (3440)
3360 - add pid to ldlm debugging output (4922)
3361 - bump the watchdog timeouts -- we can't handle 30sec yet
3362 - extra debugging for orphan dentry/inode bug (5259)
3364 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3367 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3368 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3369 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3370 - fix NULL dereference in /proc/sys/portals/routes (4827)
3371 - allow failed mdc_close() operations to be interrupted (4561)
3372 - stop precreate on OST before MDS would time out on it (4778)
3373 - don't send partial-page writes before EOF from client (4410)
3374 - discard client grant for sub-page writes on large-page clients (4520)
3375 - don't free dentries not owned by NFS code, check generation (4806)
3376 - fix lsm leak if mds_create_objects() fails (4801)
3377 - limit debug_daemon file size, always print CERROR messages (4789)
3378 - use transno after validating reply (3892)
3379 - process timed out requests if import state changes (3754)
3380 - update mtime on OST during writes, return in glimpse (4829)
3381 - add mkfsoptions to LDAP (4679)
3382 - use ->max_readahead method instead of zapping global ra (5039)
3383 - don't interrupt __l_wait_event() during strace
3385 - add software watchdogs to catch hung threads quickly (4941)
3386 - make lustrefs init script start after nfs is mounted
3387 - fix CWARN/ERROR duplication (4930)
3388 - return async write errors to application if possible (2248)
3389 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3390 - print NAL number in %x format (4645)
3391 - update barely-supported suse-2.4.21-171 series (4842)
3392 - support for sles 9 %post scripts
3393 - support for building 2.6 kernel-source packages
3394 - support for sles km_* packages
3396 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3399 - ignore -ENOENT errors in osc_destroy (3639)
3400 - notify osc create thread that OSC is being cleaned up (4600)
3401 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3402 - reconstruct ost_handle() like mds_handle() (4657)
3403 - create a new thread to do import eviction to avoid deadlock (3969)
3404 - let lconf resolve symlinked-to devices (4629)
3405 - don't unlink "objects" from directory with default EA (4554)
3406 - hold socknal file ref over connect in case target is down (4394)
3407 - allow more than 32000 subdirectories in a single directory (3244)
3408 - fix blocks count for O_DIRECT writes (3751)
3409 - OST returns ENOSPC from object create when no space left (4539)
3410 - don't send truncate RPC if file size isn't changing (4410)
3411 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3412 - bind to privileged port in socknal and tcpnal (3689)
3414 - rate limit CERROR/CWARN console message to avoid overload (4519)
3415 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3416 - basic mmap support (3918)
3417 - kernel patch series update from b1_4 (4711)
3419 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3422 - avoid crash during MDS cleanup with OST shut down (2775)
3423 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3424 - don't use bad inodes on the MDS (3744)
3425 - dynamic object preallocation to improve recovery speed (4236)
3426 - don't hold spinlock over lock dumping or change debug flags (4401)
3427 - don't zero obd_dev when it is force cleaned (3651)
3428 - print grants to console if they go negative (4431)
3429 - "lctl deactivate" will stop automatic recovery attempts (3406)
3430 - look for existing locks in ldlm_handle_enqueue() (3764)
3431 - don't resolve lock handle twice in recovery avoiding race (4401)
3432 - revalidate should check working dir is a directory (4134)
3434 - don't always mark "slow" obdfilter messages as errors (4418)
3436 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3439 - don't close LustreDB during write_conf until it is done (3860)
3440 - fix typo in lconf for_each_profile (3821)
3441 - allow dumping logs from multiple threads at one time (3820)
3442 - don't allow multiple threads in OSC recovery (3812)
3443 - fix debug_size parameters (3864)
3444 - fix mds_postrecov to initialize import for llog ctxt (3121)
3445 - replace config semaphore with spinlock (3306)
3446 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3447 - don't allow enqueue to complete on a destroyed export (3822)
3448 - down write_lock before checking llog header bitmap (3825)
3449 - recover from lock replay timeout (3764)
3450 - up llog sem before sending rpc (3652)
3451 - reduce ns lock hold times when setting kms (3267)
3452 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3453 - fix NULL deref and obd_dev leak on setup error (3312)
3454 - replace some LBUG about llog ops with error handling (3841)
3455 - don't match INVALID dentries from d_lookup and spin (3784)
3456 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3457 - fix invalid assertion in ptlrpc_set_wait (3880)
3459 - add libwrap support for the TCP acceptor (3996)
3460 - add /proc/sys/portals/routes for non-root route listing (3994)
3461 - allow setting MDS UUID in .xml (2580)
3462 - print the stack of a process that LBUGs (4228)
3464 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3467 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3468 - make sure to unset replay flag from failed open requests (3440)
3469 - if default stripe count is 0, use OST count for inode size (3636)
3470 - update parent mtime/ctime on client for create/unlink (2611)
3471 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3472 - free recovery state on server during a forced cleanup (3571)
3473 - unregister_reply for resent reqs (3063)
3474 - loop back devices mounting and status check on 2.6 (3563)
3475 - fix resource-creation race that can provoke i_size == 0 (3513)
3476 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3477 - more debugging for page-accounting assertion (3746)
3478 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3479 - avoid deadlock after precreation failure (3758)
3480 - fix race and lock order deadlock in orphan handling (3450, 3750)
3481 - add validity checks when grabbing inodes from l_ast_data (3599)
3483 - add /proc/.../recovery_status to obdfilter (3428)
3484 - lightweight CDEBUG infrastructure, debug daemon (3668)
3485 - change default OSC RPC parameters to be better on small clusters
3486 - turn off OST read cache for files smaller than 32MB
3487 - install man pages and include them in rpms (3100)
3488 - add new init script for (un)mounting lustre filesystems (2593)
3489 - run chkconfig in %post for init scripts (3701)
3490 - drop scimac NAL (unmaintained)
3492 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3495 - clean kiobufs before and after use (3485)
3496 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3497 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3498 - change init script to fail more clearly if not run as root (1528)
3499 - allow clients to reconnect during replay (1742)
3500 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3501 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3502 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3503 - fix cond_resched (3554)
3504 - only evict unfinished clients after recovery (3515)
3505 - allow bulk resend, prevent data loss (3570)
3506 - dynamic ptlrpc request buffer allocation (2102)
3507 - don't allow unlinking open directory if it isn't empty (2904)
3508 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3509 - remove extraneous obd dereference causing LASSERT failure (3334)
3510 - don't use get_cycles() when creating temp. files on the mds (3156)
3511 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3512 - handle EEXIST for set-stripe, set proper directory name (3336)
3514 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3515 - fix ksocknal_fmb_callback() error messages (2918)
3517 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3520 - don't copy lvb into (possibly NULL) reply on error (2983)
3521 - don't deref dentry after dput, don't free lvb on error (2922)
3522 - use the kms to determine writeback rpc length (2947)
3523 - increment oti_logcookies when osc is inactive (2948)
3524 - update client's i_blocks count via lvb messages (2543)
3525 - handle intent open/close of special files properly (1557)
3526 - mount MDS with errors=remount-ro, like obdfilter (2009)
3527 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3528 - don't use cancelling-locks' kms values (2947)
3529 - use highest lock extent for kms, not last one (2925)
3530 - don't dereference ERR_PTR() dentry in error handling path (3107)
3531 - fix thread race in portals_debug_dumplog() (3122)
3532 - create lprocfs device entries at setup instead of at attach (1519)
3533 - common AST error handler, don't evict client on completion race (3145)
3534 - zero nameidata in detach_mnt in 2.6 (3118)
3535 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3536 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3537 - initialize RPC timeout timer earlier for 2.6 (3219)
3538 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3539 - print nal/nid for unknown nid (3258)
3540 - additional checks for oscc recovery before doing precreate (3284)
3541 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3542 - don't crash in mdc_close for bad permissions on open (3285)
3543 - zero i_rdev for non-device files (3147)
3544 - clear page->private before handing to FS, better assertion (3119)
3545 - tune the read pipeline (3236)
3546 - fix incorrect decref of invalidated dentry (2350)
3547 - provide read-ahead stats and refine rpc in flight stats (3328)
3548 - don't hold journal transaction open across create RPC (3313)
3549 - update atime on MDS at close time (3265)
3550 - close LDAP connection when recovering to avoid server load (3315)
3551 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3552 - don't leak open file on MDS after open resend (3325)
3553 - serialize filter_precreate and filter_destroy_precreated (3329)
3554 - loop device shouldn't call sync_dev() for nul device (3092)
3555 - clear page cache after eviction (2766)
3556 - resynchronize MDS->OST in background (2824)
3557 - refuse to mount the same filesystem twice on same mountpoint (3394)
3558 - allow llmount to create routes for mounting behind routers (3320)
3559 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3560 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3561 - fix rare problem with rename on htree directories (3417)
3563 - allow default OST striping configuration per directory (1414)
3564 - fix compilation for qswnal for 2.6 kernels (3125)
3565 - increase maximum number of MDS request buffers for large systems
3566 - change liblustreapi to be useful for external progs like lfsck (3098)
3567 - increase local configuration timeout for slow disks (3353)
3568 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3570 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3573 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3574 - don't overwrite extent policy data in reply if lock was blocked (2901)
3575 - drop filter export grants atomically with removal from device (2663)
3576 - del obd_self_export from work_list in class_disconnect_exports (2908)
3577 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3578 - swab reply message in mdc_close, other PPC fixes (2464)
3579 - fix destroying of named logs (2325)
3580 - overwrite old logs when running lconf --write_conf (2264)
3581 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3582 - fix race in target_handle_connect (2898)
3583 - mds_reint_create() should take same inode create lock (2926)
3584 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3585 - don't close files for self_export to avoid uninitialized obd (2936)
3586 - allow MDS with the same name as client node (2939)
3587 - hold dentry reference for closed log files for unlink (2325)
3588 - reserve space for all logs during transactions (2059)
3589 - don't evict page beyond end of stripe extent (2925)
3590 - don't oops on a deleted current working directory (2399)
3591 - handle hard links to targets without a parent properly (2517)
3592 - don't dereference NULL lock when racing during eviction (2867)
3593 - don't grow lock extents when lots of conflicting locks (2919)
3595 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3598 - account for cache space usage on clients to avoid data loss (974)
3599 - lfsck support in lustre kernel code (2349)
3600 - reduce journal credits needed for BRW writes (2370)
3601 - orphan handling to avoid losing space on client/server crashes
3602 - ptlrpcd can be blocked, stopping ALL progress (2477)
3603 - use lock value blocks to assist in proper KMS, faster stat (1021)
3604 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3605 - recovery for initial connections (2355)
3606 - fixes for mds_cleanup_orphans (1934)
3607 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3608 - block all file creations until orphan recovery completes (1901)
3609 - client remove rq_connection from request struct (2423)
3610 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3611 - recovery timer race (2670)
3612 - mdc_close recovey bug (2532)
3613 - ptlrpc cleanup bug (2710)
3614 - mds timeout on local locks (2588)
3615 - namespace lock held during RPCs (2431)
3616 - handle interrupted sync write properly (2503)
3617 - don't try to handle a message that hasn't been replied to (2699)
3618 - client assert failure during cleanup after abort recovery (2701)
3619 - leak mdc device after failed mount (2712)
3620 - ptlrpc_check_set allows timedout requests to complete (2714)
3621 - wait for inflight reqs when ptlrpcd finishes (2710)
3622 - make sure unregistered services are removed from the srv_list
3623 - reset bulk XID's when resending them (caught by 1138 test)
3624 - unregister_bulk after timeout
3625 - fix lconf error (2694)
3626 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3627 - readahead locks pages, leaves pending causing memory pressure (2673)
3628 - increase OST request buffers to 4096 on large machines (2729)
3629 - fix up permission of existing directories in simple_mkdir (2661)
3630 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3631 - don't assign transno to errored transactions (2742)
3632 - don't delete objects on OST if given a bogus objid from MDS (2751)
3633 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3634 - if rq_no_resend, then timeout request after recovery (2432)
3635 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3636 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3637 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3638 - MDS recovery shouldn't delete objects that it has given out (2730)
3639 - if enqueue arrives after completion, don't clobber LVB (2819)
3640 - don't unlock pages twice when trigger_group_io returns error (2814)
3641 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3642 - don't write pages to disk if there was an error (1450)
3643 - don't ping imports that have recovery disabled (2676)
3644 - take buffered bytes into account when balancing socknal conn (2817)
3645 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3646 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3647 - remove little-endian swabbing of llog records (1987)
3648 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3649 - retry reposting request buffers if they fail (1191)
3650 - grow extent at grant time to avoid granting a revoked lock (2809)
3651 - lock revoke doesn't evict page if covered by a second lock (2765)
3652 - disable VM readahead to avoid reading outside lock extents (2805)
3654 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3655 - updated kernel patches for hp-2.4.20 kernel (2681)
3657 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3660 - fix truncated write corruption (2366)
3661 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3663 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3664 - ensure len(uuid) < 37 in lmc (1171)
3665 - fix ia64 OOPS in llog_test (2255)
3666 - zero end of page at obdfilter for partial page writes (2648)
3667 - don't leave stale dentries around after renames (bug 2428)
3668 - fix timeouts when evicting a client with a single lock held (2642)
3669 - set deadline for the initial HELLO message to drain (2634)
3670 - print out dotted-quad IP addresses in the socknal (2302)
3672 - additional debugging for MDS client eviction problem (2443)
3673 - fix mkfsoptions support for osts (2603, 2604)
3675 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3678 - add series for the vanilla 2.6.0 kernel
3679 - add series for the vanilla 2.4.24 kernel
3680 - add series for a cray x86/64 UL kernel drop
3681 - fix xattr patches for the vanilla 2.4.19 series
3683 - generate true UUIDs in lmc (1171)
3684 - have portals stack dumping break in UML (2466)
3685 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3686 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3687 - wake callback waiting threads on client eviction (2460)
3688 - Add --ptldebug and --subsystem to lmc (1719)
3689 - update assertion to allow safe interrupt allocation
3690 - set rq_no_resend for cancel requests (2432)
3691 - recalculate ptlrpcd timeout after resend (2494)
3692 - call vfs_rmdir when removing pending directories (2368)
3693 - fix renaming a file to itself (2429)
3694 - lmc creates a default one-stripe lov (2454)
3695 - expand procfs space to handle large clusters (2326)
3696 - increase UML stack to avoid overflow
3697 - update lconf's list of debug and subsystem masks
3698 - fix lfs find --obd (2510)
3699 - /proc tunable for disabling filter read caching (2591)
3700 - stop rpm packages from altering slapd.conf (2301)
3701 - disable nagle in the socknal under 0conf (2578)
3702 - choose mds inode size based on stripe count (2572)
3703 - fix kernel-source rpm problems (2516)
3705 - add --disable-doc to avoid pdf generation (2421)
3706 - update documentation, tests, type-os, comments
3707 - avoid format warnings on ia64
3708 - remove the TOE NAL
3709 - tiny code cleanups by removing unused fields
3711 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3714 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3715 - remove the most blatant lies from BUILDING file (2371)
3716 - change default debug level to reasonable production setting
3717 - reduce client side cache size to reduce cache flush time
3718 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3719 - make TCP zerocopy and pinger support enabled by default (2476)
3720 - sync writes completed after process exits caused crashes (2319)
3721 - maintain correct mount count on the MDS (2356)
3722 - backout 1557, because 2316 wasn't really fixed
3723 - better file I/O statistics gathering in /proc
3724 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3725 - another kernel patch to fix zero-copy TCP function export
3726 - don't take duplicate lock when processing re-sent getattr (2420)
3727 - lctl uses obd_self_export instead of creating new conn (2353)
3728 - MDS/OST recovery case which requires object creation asserted (2425)
3729 - move lfs from /usr/sbin to /usr/bin in packages
3730 - fix race between mds_client_add and mds_client_free (2417)
3731 - use kmalloc instead of slabs in portals (2430)
3732 - don't create duplicate records when a failover MDS is present (2442)
3733 - remove unnecessary mount age check (2332)
3734 - don't remove directory inodes from locks prematurely (2451)
3735 - don't break if MDS service name is the same as hostname (2103)
3736 - fix races in client write RPC generation when cache full (2482)
3738 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3741 - remove now-unused request->rq_obd (278)
3742 - if an allocation fails, print out how much memory we've used (1933)
3743 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3744 - add the "configurable stack size" patch to most series files (1256)
3745 - ability to write large log records, for 100+ OST configs (2306)
3746 - fix NULL deref when filter_prep fails (2314)
3747 - fix operator precedence error in filter_sync
3748 - dynamic allocation of socknal TX descriptors (2315)
3749 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3750 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3751 - ensure that CWARN messages in Portals always get to the syslog
3752 - __init/__exit are not for prototype decls (ldlm_init/exit)
3753 - x86-64 compile warning fixes
3754 - fix gateway LMC keyword conflict (2318)
3755 - fix MDS lock inversions in getattr/reint paths (1844)
3756 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3757 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3758 - DLM race condition prevented some lock evictions (2328)
3759 - ENOMEM detection and retry on socknal sends (2230)
3760 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3761 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3762 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3763 - better error messages when a client is rejected during recovery (1505)
3764 - avoid cancelling locks which were never granted, after failure (2330)
3765 - fix i_sem/journal inversion in mds_client_add (2333)
3766 - fix truncate/getattr lock cycle deadlock (2334)
3767 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3768 - fix two rare exit paths which could leak an l_lock() ref (2321)
3769 - fencepost error in MDS/OST orphan recovery (2226)
3770 - make log record alignment 8 bytes (1988)
3771 - lstripe now fails when requested offset > ost_count (2237)
3772 - ensure that all kernel series have a complete list.h (1607)
3773 - fix crashes in special-file operations (2316)
3774 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3776 - allow configurable automake binary, for testing new versions
3777 - small update to the lfs documentation
3779 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3781 * fix negative export reference count in fsfilt_sync (2312)
3783 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3784 * release candidate 0.9.1
3786 - orphans are moved into the PENDING directory for possible recovery
3787 - replayed opens now open by fid for orphan/rename safety (1042)
3788 - last close of an orphan inode generates a transno (683)
3789 - chdir() and mount() now pin the directory entry (1020)
3790 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3791 - discard very old requests without processing them (1502)
3792 - remove some common, well-understood CERRORs (1505)
3793 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3794 - clear "grant" flags in OST replies until OST grant code lands (1644)
3795 - fix read performance by not clobbering i_blksize on client (1598)
3796 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3797 - make LRU size a /proc tunable, clears locks when reduced (707)
3798 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3799 - server locks take references on exports now (1558)
3800 - build fixes for 2.4.20-rh trees (1663)
3801 - return an error from lov_create if all OSCs are inactive (1751)
3802 - fix import levels when a reconnect happens without a timeout (1597)
3803 - exit early from mds_open if we get a lookup error (1749)
3804 - partial page read at EOF wouldn't wait for disk before sending (1642)
3805 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3806 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3807 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3808 - write/truncate lock inversion (1639)
3809 - Don't auto-load obdclass, portals modules during cleanup (1495)
3810 - fix timestamps from jumping to "now" (1763)
3811 - extra journal assertions (1648)
3812 - add an extra multiunlink test (1771)
3813 - fix read_record/write_record API (1776)
3814 - fix leak of offset_extent, possible incorrect i_size later (1772)
3815 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3816 - Debugging for the kqswnal_get_idle_tx problems (1820)
3817 - Allow recovery to be attempted multiple times (1536)
3818 - Write out MDS last_rcvd file after it is first created (1600)
3819 - Fix tx_descriptor leak in failed transmit situations (1827)
3820 - ext3 journaling fixes for assertion failure after IO error (1871)
3821 - class_export_put() on freed export after completion AST error (1896)
3822 - Fix revalidate looping in VFS (1322)
3823 - Don't access a freed export during MDS_REINT timeout (1521)
3824 - Add open-unlink recovery support on the MDS (1673,1764)
3825 - Return an error if no MDS data was read from last_rcvd (1946)
3826 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3827 - Don't setup a disk device that doesn't match exported UUID (317)
3828 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3829 - avoid committing NULL handle in force close
3830 - local.sh is now a one-stripe LOV configuration
3831 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3832 - fix ext3 htree duplicate directory entry corruption (1516)
3833 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3834 - update to new LOV EA format (2097)
3835 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3837 2003-06-15 Phil Schwan <phil@clusterfs.com>
3840 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3841 - per-import/export recovery handling (958, 931, 959)
3842 - multiple last-rcvd slots, for serving multiple FSes (949)
3843 - connections are again shared between multiple imp/exports (963, 964)
3844 - "umount -f" would hang if any requests needed to be sent (393, 978)
3845 - avoid pinning large req buffer by copying for queued messages (989)
3846 - add "uuid" to "lctl device" command to help upcalls (991)
3847 - "open" RPCs with transnos would confuse recovery counters (1037)
3848 - do proper endian conversion of all wire messages (288, 340, 891)
3849 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3850 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3851 - fix ptlrpc_connection leak in target_handle_connect (1174)
3852 - fix import refcounting bug in OST and MDS cleanup (1134)
3853 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3854 - fix very unlikely obd_types race condition (501)
3855 - remove osc_open hack for echo_client (1187)
3856 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3857 - a failure in read_inode2 leads to deadlock (1139)
3858 - cancel ack-locks as soon as transaction is committed (1072)
3859 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3860 - make sure to commitrw after any preprw to avoid deadlock (1162)
3861 - failing to execute a file in a lustre FS would lock inode (1203)
3862 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3863 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3864 - obd_uuid2tgt would walk off the end of the list (1255)
3865 - on IA64 the osc would give portals incorrect bulk size (1258)
3866 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3867 - fix lock inversion caused by new llite matching code (1282)
3868 - limit the number of dirty pages on a client to 10MB (1286)
3869 - timed out locks were not being corrected cancelled (1289)
3870 - fix O_DIRECT above 4GB on IA-32 (1292)
3871 * major user-visible changes
3872 - fail out/fail over policy now controlled by the upcall (993)
3874 - add OBD_PING to check server availability and failure (954)
3875 - lustre messages are now sent in sending host order (288, 340, 891)
3876 - add eadatalen to MDS getattr reply (340)
3877 - OST read replies may contain second buffer, with per-page status (593)
3879 2003-03-11 Phil Schwan <phil@clusterfs.com>
3882 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3883 - failed open()s could cause deadlock; fixed (867, 869)
3884 - stop cancelling OST locks when files are closed (481)
3885 - overlapping XID spaces caused network corruption (851, 853)
3886 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3887 - setattr_raw added, to do single-RPC, server-side setattrs
3888 - lmc/lconf syntax change for OST UUIDs
3889 - fix crashy race condition between ptlrpc_free_req and osc_close
3890 - don't use request in mdc_enqueue if we hit a timeout (889)
3891 - don't set the inode i_size for regular files from the MDS (896)
3892 - handle out of order completion AST (842)
3893 - don't LBUG if a lock request times out after receiving AST (913)
3894 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3895 - if a bad lock AST arrives, send an error instead of dropping entirely
3896 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3897 - fix leak in bulk IO when only partially completed (899, 900, 926)
3898 - fix O_DIRECT for ia64 (55)
3899 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3900 - C-z after timeout could hang a process forever; fixed (977)
3902 - client-side I/O cache (678, 924, 929, 941, 970)
3904 - READPAGE and SETATTRs which don't take server-side locks get
3907 2003-02-11 Phil Schwan <phil@clusterfs.com>
3910 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3911 - Fix fsfilt_extN_readpage() to read a full page of directory
3912 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3913 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3914 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3915 - Most intent operations were removed, in favour of a new RPC mode
3916 that does a single RPC to the server and bypasses most of the VFS
3917 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3918 - Aggressively cancel local locks on DLM servers
3919 - mds_reint_unlink sends EA to the client if it's the last nlink.
3920 client uses that EA to unlink OST objects.
3921 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3922 - recursive symlinks were fixed (439)
3923 - fixed NULL deref in DEBUG_REQ
3924 - filter_update_lastobjid no longer calls sync, which annoyed extN
3925 - fixed multi-client small-writes to a single file problem (445)
3926 - fixed mtime updates during file writes (607)
3927 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3928 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3929 - fixed Linux OST size reporting problem (444, 656)
3930 - OST now updates object mtime with writes or setattr (607, 619)
3931 - client verifies file size before zeroing page past EOF (445)
3932 - OST now writes last allocated objid to disk with allocation (108)
3933 - LOV on echo now works (409)
3935 - mds_reint_unlink sends a new buffer, with the EA included. this
3936 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3937 set if a regular file was being unlinked, and it was the last link
3938 - use PtlGet from the target for bulk writes (315)
3939 - OST now updates object mtime with writes or setattr (607, 619)
3940 - LDLM now has a grant-time callback to revalidate locked items, if
3942 - Many MDS operations were reorganized to combat race conditions
3944 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3945 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3947 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3950 - Fully reactivate OST imports after reconnection (512, others)
3951 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3952 - More graceful error handling for truncating on dead OST (515)
3953 - Don't error out unless we're actually accessing dead stripes (474)
3954 - Fix garbage sizes when stripes are missing (410)
3955 - LRU counters were broken, causing constant lock purge (433, 432)
3956 - garbage on read from stripes with failed OSTs (441)
3957 - mark OSCs as active before reconnecting during recovery (438)
3958 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3959 - lfind did not preserve OST order in output (443)
3960 - symlinks cause hung clients, incorrect data (439)
3961 - stop dereferencing request after dropping refcount (457)
3962 - don't LASSERT(spin_is_locked) on non-SMP (455)
3963 - fixes for many rename() bugs
3964 - fstat didn't correctly synchronize attributes (399)
3965 - server must handle lock cancellation during blocking AST prep (487)
3966 - bulk descriptors were free()d too soon (511)
3967 - fix paths in lconf, which would load incorrect modules (451, 507)
3968 - fix confusing lconf 'host not found' error message (386)
3969 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3970 - fix race condition in mdc_blocking_ast() for inode access (526)
3971 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3972 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3973 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3974 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3975 - fix lstripe in several ways (kernel side) (527)
3976 - fix request leak in ldlm_cli_enqueue (262)
3977 - incorrect OSC was marked inactive after OST failure
3978 - call mds_fs_cleanup before unmounting filesystem (524)
3979 - fix races between taking ns_lock and ldlm_lock_change_resource
3980 - fix races updating LOV export open file list
3981 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3982 - fix recovery NULL deref in ldlm_cli_cancel_unused
3983 - fix some DLM races by using new hash table for lock handles (419)
3984 - permit the client to specify desired inodes, at replay
3985 - duplicate requests when we queue them for replay reintegration
3986 - fix last_rcvd offset calculation
3987 - sync after each recovered transaction, so we always make progress
3988 - never, not always, ERESTART requests without transnos
3989 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3991 - skip replay if the MDS says that the client is already connected
3992 - don't check for a recovery-enabled export to match lctl's UUID
3993 - don't INC_USE_COUNT for phantom exports
3994 - don't crash when cleaning up phantom exports (567)
3995 - don't double-finish or set replay data for errored mdc_open requests
3996 - abort requests when they time out, so we don't get old replies
3997 - send/receive replies for AST messages again
3998 - if the client says that it doesn't have the lock, cancel it on the
4000 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4001 mark it as destroyed, it will all work out when decref is called
4002 - fix module use counts (22, 581)
4004 - ASTs now expect a reply (server cancels lock on error reply)
4006 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4009 - fix many simultaneous client startup (392)
4010 - fix dentry->d_it clobbering
4011 - credentials weren't being shipped for readdir/getattr operations
4012 - remove invalid assertions triggered during some concurrent MD
4014 - proper Lustre versions added (336, 389)
4015 - fix memory leak for create error case (398)
4016 - fix LOV locking bug that would get cli/srv out of sync
4017 - fix echo client over LOV (409)
4018 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4019 - fix double-O_EXCL intent crash (424)
4020 - avoid sending multiple lock CANCELs (352)
4022 - MDS can do multi-client recovery (modulo bugs in new code)
4024 - many updates, edits, cleanups
4026 2002-11-18 Phil Schwan <phil@clusterfs.com>
4029 - fix null d_it dereference (346)
4030 - fix full OST/dbench hang (333)
4031 - fix permission problem with file removal (286)
4032 - fix removal of OSCs from LOV when they fail
4033 - fix NULL deref during bulk timeout (214)
4034 - fix problems related to multiple filesystems on one MDS (241)
4035 - fixed serious subtle metadata locking bugs
4036 - free locks on clients when inodes are removed due to memory
4038 - fix inode pointer in lock data (285)
4039 - partial support for multiple MDS on a single host (241)
4040 - data locks weren't cancelled at clear_inode time (290, 311)
4041 - intent locks could lead to unbounded lock growth (205)
4042 - added a maximum lock count, an LRU list, and a flusher
4043 - fix multiple rename (365)
4044 - properly abstracted the echo client
4045 - OSC locked 1 byte too many; fixed
4046 - rewrote brw callback code:
4047 - fixed recovery bugs related to LOVs (306)
4048 - fixed too-many-pages-in-one-write crash (191)
4049 - fixed (again) crash in sync_io_timeout (214)
4050 - probably fixed callback-related race (385)
4052 - Add capability to MDS protocol
4053 - LDLM cancellations and callbacks on different portals
4055 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4058 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4059 - defer open object destruction to close time (601981, 138)
4060 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4061 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4062 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4063 - don't call obd_setattr in ll_file_release for destroyed objects
4064 * protocol change to lustre_msg: move |version| and add |flags|
4065 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4066 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4067 * added replay of create, unlink, link and rename operations during
4068 MDS failover; recovery should be much more robust now
4069 * remove failed OSCs from LOVs (only lov_create uses this so far)
4070 * the lustre-HOWTO was brought (more) up to date (582544)
4072 2002-10-23 Phil Schwan <phil@clusterfs.com>
4075 - in-use dentries weren't being reused properly (617851)
4076 - prevent multiple LDLM setup (599178)
4077 - fix LOV size calculations for truncate (617853)
4078 - fix client handling of MDS intent errors (POSIX)
4079 - fix permission bug in lovstripe.c test (624321)
4080 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4081 - truncate past end of file could corrupt data
4082 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4083 - a race in recovery could return ETIMEDOUT to apps (623947)
4084 - building outside the source directory was fixed
4085 * the lustre-HOWTO was brought (more) up to date (582544)
4086 * major progress was made on recovery functionality
4088 2002-10-10 Phil Schwan <phil@clusterfs.com>
4091 - recovery deadlock fix
4092 - rm -rf causes LBUG fix (617817)
4093 - file open by multiple tasks fix (618962)
4094 - directory permissions bugs (602707 and 620007)
4095 - journal_stop fixed with locking (611313)
4096 - O_APPEND failures resolved (618273, perhaps 614459)
4097 - lconf PATH fix (619770)
4098 - IA64 build fix (621450)
4099 - RPC buffer sizes scale with amount of memory
4101 2002-10-01 Phil Schwan <phil@clusterfs.com>
4104 - locks would be cancelled without throwing away data pages,
4105 resulting in inconsistent data (605627)
4106 - inode attributes were not always being refreshed (605627, 612449)
4107 - lconf now continues to cleanup after lctl reports an error
4108 - MDS now enforces user permissions (602707)
4109 - lprocfs cleanup fixed, but not yet enabled (614157)
4110 - fixed infinite server hang, should a client not respond to an AST
4111 - avoid going into recovery if user calls readlink() with a buffer
4112 that's too small (613941)
4113 - AST RPCs no longer require replies (614867) -- this may be changed
4114 - don't crash server if client sends an IOV that's too big (611336)
4115 - fixed lock conversion deadlock (611892)
4116 - fixed the following of symlinks (614622)
4117 * recovery: the server can remove locks from a client that dies, other
4118 clients can make progress
4119 * more extN patch fixes
4120 * compile-time configurable ptlrpc buffer allocations
4122 - collaborative read cache document
4123 - Lustre Lite Performance CDR document-in-progress
4125 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4128 - fix typo in patch-2.4.18
4130 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4133 - clear ptlrpc request each time in handle_incoming_request()
4134 - unlink of files now destroys the object on the OST
4136 2002-09-19 Peter Braam <braam@clusterfs.com>
4138 * add hard link support
4139 * change obdfile creation method
4140 * kernel patch changed
4142 2002-09-19 Peter Braam <braam@clusterfs.com>
4145 - stack overflow bug in extN fixed
4147 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4149 * documentation updates
4150 - add man pages for config tools
4151 - update tests/README to describe testing with new config tools
4152 - finish metadata API descriptions
4153 * bug fixes and cleanups
4154 - statfs workaround for 16TB limit
4155 - LOV stripe allocation improved, can stripe on subset of OSTs
4156 - LOV file size/IO offset was wrong for files > 4GB in size
4157 - object EA data was being dropped, caused files to be unreadable
4158 - memory overflow with non-LOV OST caused memory corruption
4159 - fixed regression tests to work with new config tools, obdfilter
4160 - fixed bug when directory size became larger than 1 block
4161 - fixed bug (for single client case) when PWD was deleted
4162 - invalidate local directory pages when doing intent-based ops
4163 - avoid LDLM oops when lock callback contained bad data
4165 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4167 * documentation updates
4168 * bug fixes and cleanups
4169 - configuration tools
4172 - 64-bit compile warnings
4173 - 64-bit internal statfs data
4175 * test_brw on persistent OST devices
4177 * lprocfs (disabled)
4179 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4181 * documentation updates
4182 * bug fixes and cleanups
4183 * configuration tools
4185 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4188 * many small fixes to 0_5_4
4189 * io/network handling
4190 * thinkos in MDS operations
4192 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4195 * crucial basic fixes to 0.5.3
4196 * IOR, Iozone work over Elan
4199 2002-08-07 Phil Schwan <phil@clusterfs.com>
4200 * version 0_5_3, our first alpha
4201 * we use the new Portals iovs
4202 * documentation updates
4203 * bug fixes and cleanups
4204 * small changes in the DLM wire protocol
4206 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4207 * version 0_5_1 with some initial stability,
4208 * locking on MD and file I/O.
4209 * documentation updates
4210 * several bug fixes since 0.5.0
4211 * small changes in wire protocol
4213 2002-07-18 Phil Schwan <phil@clusterfs.com>
4215 * delivered as Lustre Light Alpha
4216 * fixed a crash after handling invalid MDS requests
4217 * fixed directory pages for architectures with non-4k pages sizes
4219 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4220 * release version v0_4_4
4221 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4223 * Fixes a number of interruption problems with OST operations.
4224 * Update documentation for portals header changes
4225 * Move all wire protocol structs/defines to lustre_idl.h
4226 * Fixes symlink length bug.
4227 * Add tcpdump to repository.
4229 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4230 * release version v0_4_3
4231 * Fixes statfs for inodes on extN.
4232 * Fixes bug in runtests which would delete /etc/hosts.
4233 * Use 64-bit object IDs wherever possible (not into VFS though)
4234 Remove ost_get_info, which is unused by lustre, and out of date.
4236 2002-07-03 Peter Braam <braam@clusterfs.com>
4237 * release version v0_4_2 Fixes a lookup error (type not passed)
4238 * move forward to head of Portals
4239 * move forward to latest Lustre kernel
4241 2002-06-25 Peter Braam <braam@clusterfs.com>
4242 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob