3 2.6.5-7.287.3 (SLES 9),
4 2.6.9-55.0.12.EL (RHEL 4),
5 2.6.16.53-0.16 (SLES 10),
6 2.6.18-8.1.15.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-cfs4
13 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
16 Severity : enhancement
18 Description: lost problems with lov objid file
19 Details : Fixes some scability and access to not inited memory problems
20 in work with lov objdid file.
25 Description: lfs find does not continue on file error
26 Details : Continue other files processing when a file/dir is absent.
30 Description: Inconsistent usage of lustre_pack_reply()
31 Details : Standardize the usage of lustre_pack_reply() such that it
32 always generate a CERROR on failure.
35 Frequency : occasional
37 Description: MDS hang or stay in waiting lock
38 Details : If client receive lock with CBPENDING flag ldlm need send lock
39 cancel as separate rpc, to avoid situation when cancel request
40 can't processed due all i/o threads stay in wait lock.
44 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
48 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
52 Description: Update to SLES9 kernel-2.6.5-7.287.3.
56 Description: Update to RHEL4 latest kernel.
58 Severity : enhancement
60 Description: Build SLES10 patchless client fails
61 Details : The configure was broken by run ./configure with
62 --with-linux-obj=.... argument for patchless client. When the
63 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
64 can't search header adequately. Use absolute path such as
65 -I($LINUX)/include instead.
69 Description: interrupt oig_wait produce painc on resend.
70 Details : brw_redo_request can be used for resend requests from ptlrpcd and
71 private set, and this produce situation when rq_ptlrpcd_data not
72 copyed to new allocated request and triggered LBUG on assert
73 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
74 ptlrpcd set if request is changed and can be safety changed to use
79 Description: LASSERT_{REQ,REP}SWAB macros are buggy
80 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
81 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
82 these macros and replace them with their difinitions instead.
86 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
87 Details : Update which_patch & target file for SLES10 latest kernel.
90 Description: Remove client patches from SLES 10 kernel.
91 Details : This causes SLES 10 clients to behave as patchless clients
92 even on a Lustre-patched (server) kernel.
94 Severity : enhancement
96 Description: self-adjustable client's lru lists
97 Details : use adaptive algorithm for managing client cached locks lru
98 lists according to current server load, other client's work
99 pattern, memory activities, etc. Both, server and client
100 side namespaces provide number of proc tunables for controlling
103 Severity : enhancement
105 Description: light-weight GSS support
106 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
107 and reduce performance overhead.
109 Severity : enhancement
111 Description: Linux keyring support
112 Details : Support using service of Linux keyring for Lustre GSS internal
113 context refresh/cache mechanism.
117 Description: Fix errors in lfs documentation
118 Details : Fixes man pages
122 Description: don't use GFP_* in generic Lustre code.
123 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
127 Description: obdclass is limited by single OBD_ALLOC(idarray)
128 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
132 Description: warnings with build patchless client with vanila 2.6.19 and up
133 Details : change old ctl_table style and replace ctl_table/ctl_table_header
134 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
138 Frequency : Only for SLES
139 Description: sanity_quota fail test_1
140 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
141 file, which makes TSTID[2] inunique.
145 Frequency : Only for SLES9
146 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
147 Details : Update target/ChangeLog/which_patch .
151 Frequency : Only for test_10 in sanity-quota.sh
152 Description: a bug in quota test code
153 Details : checking a test flag in a wrong place
157 Frequency : Only for test_13 in sanity-quota.sh
158 Description: a bug in quota test script
159 Details : shouldn't check overfull allocated quota
163 Description: jbd statistics
164 Details : Port older jbd statistics patch for sles10
166 Severity : enhancement
168 Description: Add printing inode info into message about error in writepage.
173 Description: reply_lock_interpret crash due to race with it and lock cancel.
174 Details : Do not replay locks that are being cancelled. Do not reference
175 locks by their address during replay, just by their handle.
179 Frequency : When flocks are used.
180 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
181 Details : Flock locks might destroy just granted lock if it could be merged
182 with another existing flock, this is done in completion handler,
183 so teach ldlm_cli_enquque_fini that this is a valid case for
189 Description: Oops in read and write path when failing to allocate lock.
190 Details : Check if lock allocation failed and return error back.
194 Description: lstripe command fails for valid OST index
195 Details : The stripe offset is compared to lov->desc.ld_tgt_count
196 instead of lov->desc.ld_active_tgt_count.
200 Description: sanity.sh failed test 103
201 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
206 Description: df doesn't work properly if diskfs blocksize != 4K
207 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
211 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
212 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
216 Description: obd_health_check_timeout too short
217 Details : set obd_health_check_timeout as 1.5x of obd_timeout
221 Description: llapi_file_create() does not allow some changes
222 Details : add llapi_file_open() that allows specifying the mode and
223 open flags, and also returns an open file handle.
227 Description: merge and cleanup kernel patches.
228 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
232 Description: Add journal checksum support.(Kernel part)
233 Details : The journal checksum feature adds two new flags i.e
234 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
235 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
236 indicates that the commit block contains the checksum for
237 the blocks described by the descriptor blocks. Now commit
238 record can be sent to disk without waiting for descriptor
239 blocks to be written to disk. This behavior is controlled
240 using JBD2_FEATURE_ASYNC_COMMIT flag.
244 Description: OSS needs mutliple precreate threads
245 Details : Add ability to start more than one create thread per OSS.
249 Description: Sanity.sh test_65k failed: setstripe should have succeeded
250 Details : execute lfs setstripe on client
254 Description: mds_obd_create error creating tmp object
255 Details : When the user sets quota on root, llog will be affected and can't
256 create files and write files.
259 Frequency : Always on ia64 patchless client, and possibly others.
261 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
262 Details : This allows the patchless client to be loaded on architectures
267 Description: RedHat Update kernel for RHEL5
268 Details : Add the kernel config file for RHEL5.
272 Description: RedHat Update kernel for RHEL5
273 Details : Modify the kernel config file more closer RHEL5.
277 Description: Build failure against Centos5 (RHEL5)
278 Details : Define PAGE_SIZE when it isn't present.
282 Description: test_7 on sanity-quota.sh can't be used on separate machines
283 Details : add facet to handle it
287 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
288 Details : with NFS, the anon dentry's parent was set to itself in
289 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
290 corresponding dentry other than use rec->ur_name.
292 Severity : enhancement
294 Description: lfs setstripe enhancement
295 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
299 Description: enable data checksumming by default
300 Details : enable checksum by default, allow --disable-checksum
301 configure option and "-o nochecksum" mount option. Checksums
302 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
306 Description: lustre support for RHEL5
307 Details : Add support for RHEL5.
311 Description: Client eviction due to failover config
312 Details : after a connection loss, the lustre client should attempt to
313 reconnect to the last active server first before trying the
314 other potential connections.
318 Description: when mds and osts use different quota unit(32bit and 64bit),
319 quota will be released repeatly.
320 Details : void sending multiple quota reqs to mds, which will keep the
321 status between the reqs.
325 Description: osts not allocated evenly to files
326 Details : change the condition to increase offset_idx
329 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
331 Description: Data corruption for OSTs that are formatted larger than 2TB
333 Details : When generating the bio request for lustre file writes the
334 sector number would overflow a temporary variable before being
335 used for the IO. The data reads correctly from Lustre (which
336 will overflow in a similar manner) but other file data or
337 filesystem metadata may be corrupted in some cases.
341 Description: Tune the kernel for good SCSI performance.
342 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
343 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
348 Description: rewrite ext2-derived code in obdclass/uuid.c
349 Details : rewrite inherited code (uuid parsing code from ext2 utils)
350 from scratch preserving functionality.
354 Description: "lfs find -obd UUID" prints directories
355 Details : "lfs find -obd UUID" will return all directory names instead
356 of just file names. It is incorrect because the directories
357 do not reside on the OSTs.
361 Description: use i_size_read and i_size_write in 2.6 port
362 Details : replace inode->i_size access with i_size_read/write()
365 Frequency : when removing large files
367 Description: scheduling issue during removal of large Lustre files
368 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
369 It causes scheduling issues when removing large files (17TB in the
373 Frequency : only with liblustre clients on XT3
375 Description: evictions taking too long
376 Details : allow llrd to evict clients directly on OSTs
379 Frequency : only on ppc
381 Description: /proc/fs/lustre/devices broken on ppc
382 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
383 We should make sure the type of 'index' is loff_t and then cast
384 explicitly as needed below. Do not assign an explicitly cast
388 Frequency : only for rhel5
390 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
391 Details : Modify the target file & which_kernel.
393 Severity : enhancement
395 Description: omit set fsid for export NFS
396 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
397 as FS_REQUIRES_DEV for avoid problems with generate fsid.
401 Frequency : Always, for kernels after 2.6.16
402 Description: Fix warning idr_remove called for id=.. which is not allocated.
403 Details : Last kernels save old s_dev before kill super and not allow
404 to restore from callback - restore it before call kill_anon_super.
408 Description: buffer overruns could theoretically occur
409 Details : llapi_semantic_traverse() modifies the "path" argument by
410 appending values to the end of the origin string, and a
411 overrun may occur. Adding buffer overrun check in liblustreapi.
415 Description: Fix error on 'ls .' at the top of the Lustre mount.
416 Details : Don't revalidate dentry if it is a root dentry.
420 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
421 Details : Modify vm-tunables-rhel4.patch.
425 Description: Kernel config for 2.6.18-vanilla.
426 Details : Modify targets/2.6-vanilla.target.in.
427 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
428 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
429 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
430 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
434 Description: improve handling recoverable errors
435 Details : if request processig with error which can be recoverable on server
436 request should be resend, otherwise page released from cache and
441 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
442 Details : Modify target file & which_patch.
443 A flaw was found in the IA32 system call emulation provided
444 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
445 value could be stored in the %RAX register, which could trigger an
446 out-of-bounds system call table access. An untrusted local user
447 could exploit this flaw to run code in the kernel
448 (ie a root privilege escalation). (CVE-2007-4573).
452 Description: O_DIRECT bypasses client statistics.
453 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
454 read_ahead_stats were updated. copied the stats section from
455 osc_send_oap_rpc() into async_internal().
459 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
463 Description: change order of libsysio includes
464 Details : '#include sysio.h' should always come before '#include xtio.h'
468 Description: handle "serious error: objid * already exists" more gracefully
469 Details : If LAST_ID value on disk is smaller than the objects existing in
470 the O/0/d* directories, it indicates disk corruption and causes an
471 LBUG(). If the object is 0-length, then we should use the existing
472 object. This will help to avoid a full fsck in most cases.
474 Severity : enhancement
476 Description: adapt the lustre_config script to support the upgrade case
477 Details : Add "-u" option for lustre_config script to support upgrading 1.4
478 server targets to 1.6 in parallel.
482 Description: To avoid grant space > avaible space when the disk is almost
483 full. Without this patch you might see the error "grant XXXX >
484 available" or some LBUG about grant, when the disk is almost
486 Details : In filter_check_grant, for non_grant cache write, we should
487 check the left space by if (*left > ungranted + bytes), instead
488 of (*left > ungranted), because only we are sure the left space
489 is enough for another "bytes", then the ungrant space should be
490 increase. In client, we should update cl_avail_grant only there
491 is OBD_MD_FLGRANT in the reply.
495 Description: Update RHEL 4 kernel to fix local root privilege escalation.
496 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
497 described in CVE-2007-4573. This problem could allow untrusted
498 local users to gain root access.
501 Frequency : when using O_DIRECT and quotas
503 Description: Incorrect file ownership on O_DIRECT output files
504 Details : block usage reported by 'lfs quota' does not take into account
505 files that have been written with O_DIRECT.
510 Description: touch file failed when fs is not full
511 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
512 otherwise we can get ENOSP while fs is not full.
516 Description: parallel lock callbacks
517 Details : Instead of sending blocking and completion callbacks as separated
518 requests, adding them to a set and sending in parallel.
521 Frequency : only for Cray XT3
522 Bugzilla : 12829/13455
523 Description: Changing primary group doesn't change the group lustre assigns to
525 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
526 overridden with the primary group provided by the group upcall,
527 whereas the supplied fsgid can be trusted if it is in the list of
528 supplementary groups returned by the group upcall.
530 --------------------------------------------------------------------------------
532 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
534 * Support for kernels:
535 2.6.5-7.283 (SLES 9),
536 2.6.9-55.EL (RHEL 4),
537 2.6.16.46-0.14 (SLES 10),
538 2.6.18.8 vanilla (kernel.org)
539 * Client support for unpatched kernels:
540 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
541 2.6.16 - 2.6.22 vanilla (kernel.org)
542 * Due to recently discovered recovery problems, we do not recommend
543 using patchless RHEL 4 clients with this or any earlier release.
544 * Recommended e2fsprogs version: 1.39.cfs8
545 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
546 * Starting with this release, the ldiskfs backing filesystem required
547 by Lustre is now in its own package, lustre-ldiskfs. This package
548 should be installed. It is versioned separately from Lustre and
549 may be released separately in future.
553 Description: block reactivating mgc import until all deactivates complete
554 Details : Fix race when failing back MDT/MGS to itself (testing)
557 Frequency : at statup only
559 Description: mds_lov_synchronize race leads to various problems
560 Details : simultaneous MDT->OST connections at startup can cause the
561 sync to abort, leaving the OSC in a bad state.
563 Severity : enhancement
565 Description: add optional extra BUILD_VERSION info
566 Details : add a new environment variable (namely LUSTRE_VERS) which allows
567 to override the lustre version.
570 Frequency : 2.6.18 servers only
572 Description: ll_kern_mount() doesn't release the module reference
573 Details : The ldiskfs module reference count never drops down to 0
574 because ll_kern_mount() doesn't release the module reference.
579 Description: server LBUG when using old ost_num_threads parameter
580 Details : Accept the old ost_num_threads parameter but warn that it
581 is deprecated, and fix an off-by-one error that caused an LBUG.
586 Description: Transient SCSI error results in persistent IO issue
587 Details : iobuf->dr_error is not reinitialized to 0 between two
591 Frequency : sometimes when underlying device returns I/O errors
593 Description: OSTs not going read-only during write failures
594 Details : OSTs are not remounted read-only when the journal commit threads
595 get I/O errors because fsfilt_ext3 calls journal_start/stop()
596 instead of the ext3 wrappers.
600 Description: poor connect scaling with increasing client count
601 Details : Don't run filter_grant_sanity_check for more than 100 exports
602 to improve scaling for large numbers of clients.
605 Frequency : SLES10 only
607 Description: sanity-quota.sh quotacheck failed: rc = -22
608 Details : Quotas cannot be enabled on SLES10.
611 Frequency : liblustre clients only
613 Description: getdirentries does not give error when run on compute nodes
614 Details : getdirentries does not fail when the size specified as an argument
615 is too small to contain at least one entry
617 Severity : enhancement
619 Description: Add LNET router traceability for debug purposes
620 Details : If a checksum failure occurs with a router as part of the
621 IO path, the NID of the last router that forwarded the bulk data
622 is printed so it can be identified.
627 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
628 Details : Due to a race condition, liblustre clients were occasionally
631 Severity : enhancement
633 Description: lfs setstripe use optional parameters instead of postional
636 Severity : enhancement
638 Description: Nanosecond timestamp support for ldiskfs
639 Details : The on-disk ldiskfs filesystem has added support for nanosecond
640 resolution timestamps. There is not yet support for this at
641 the Lustre filesystem level.
644 Frequency : during server recovery
646 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
647 Details : request with rq_no_resend flag not awake l_wait_event if they get a
651 Frequency : nfs export on patchless client
653 Description: connectathon hang when test nfs export over patchless client
654 Details : Disconnected dentry cannot be found with lookup, so we do not need
655 to unhash it or make it invalid
658 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
659 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
660 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
665 Description: open req refcounting wrong on reconnect
666 Details : If reconnect happened between getting open reply from server and
667 call to mdc_set_replay_data in ll_file_open, we will schedule
668 replay for unreferenced request that we are about to free.
669 Subsequent close will crash in variety of ways.
670 Check that request is still eligible for replay in
671 mdc_set_replay_data().
676 Description: disable writes to filesystem when reading health_check file
677 Details : the default for reading the health_check proc file has changed
678 to NOT do a journal transaction and write to disk, because this
679 can cause reads of the /proc file to hang and block HA state
680 checking on a healthy but otherwise heavily loaded system. It
681 is possible to return to the previous behaviour during configure
682 with --enable-health-write.
684 Severity : enhancement
686 Description: 64-bit inode version
687 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
688 to the inode. This will be required for version-based recovery.
693 Description: MDS fails to start if a duplicate client export is detected
694 Details : in some rare cases it was possible for a client to connect to
695 an MDS multiple times. Upon recovery the MDS would detect this
696 and fail during startup. Handle this more gracefully.
698 Severity : enhancement
700 Description: Add -o localflock option to simulate old noflock
702 Details : This will achieve local-only flock/fcntl locks
708 Description: log_commit_thread vs filter_destroy race leads to crash
709 Details : Take import reference before releasing llog record semaphore
714 Description: Wrong request locking in request set processing
715 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
716 imp_delayed_list, in this place should be used imp_lock.
719 Frequency : when reconnection
721 Description: Grant Leak when osc reconnect to OST
722 Details : When osc reconnect ost, OST(filter) should check whether it
723 should grant more space to client by comparing fed_grant and
724 cl_avail_grant, and return the granted space to client instead
725 of "new granted" space, because client will call osc_init_grant
726 to update the client grant space info.
729 Frequency : when client reconnect to OST
731 Description: Grant Leak when osc do resend and replay bulk write
732 Details : When osc reconnect to OST, OST(filter)should clear grant info of
733 bulk write request, because the grant info while be sync between
734 OSC and OST when reconnect, and we should ignore the grant info
735 these of resend/replay write req.
740 Description: Grant space more than avaiable left space sometimes.
741 Details : When then OST is about to be full, if two bulk writing from
742 different clients came to OST. Accord the avaliable space of the
743 OST, the first req should be permitted, and the second one
744 should be denied by ENOSPC. But if the seconde arrived before
745 the first one is commited. The OST might wrongly permit second
746 writing, which will cause grant space > avaiable space.
749 Frequency : when client is evicted
751 Description: Grant might be wrongly erased when osc is evicted by OST
752 Details : when the import is evicted by server, it will fork another
753 thread ptlrpc_invalidate_import_thread to invalidate the
754 import, where the grant will be set to 0. While the original
755 thread will update the grant it got when connecting. So if
756 the former happened latter, the grant will be wrongly errased
757 because of this race.
762 Description: Checking Stale with correct fid
763 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
764 to check whether it is stale, because sometimes, we want the
765 enqueue happened anyway, and op_data.fid2 will not be initialized.
767 Severity : enhancement
769 Description: update patchless client
770 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
773 Frequency : only with 2.4 kernel
775 Description: random memory corruption
776 Details : size of struct ll_inode_info is to big for union inode.u and this
777 can be cause of random memory corruption.
782 Description: Memory leak in recovery
783 Details : Lov_mds_md was not free in an error handler in mds_create_object.
784 It should also check obd_fail before fsfilt_start, otherwise if
785 fsfilt_start return -EROFS,(failover mds during mds recovery).
786 then the req will return with repmsg->transno = 0 and rc = EROFS.
787 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
788 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
789 be freed no matter whether fsfilt_commit success or not.
792 Frequency : only with huge count clients
794 Description: Prevents from taking the superblock lock in llap_from_page for
796 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
797 for llap_from_page call in ll_removepage prevents from taking the
798 superblock lock for a soon died page.
803 Description: Not check open intent error before release open handle
804 Details : in some rare cases, the open intent error is not checked before
805 release open handle, which may cause
806 ASSERTION(open_req->rq_transno != 0), because it tries to release
807 the failed open handle.
812 Description: Set cat log bitmap only after create log success.
813 Details : in some rare cases, the cat log bitmap is set too early. and it
814 should be set only after create log success.
818 Description: Accessing a block bevice can re-enable I/O when Lustre is
819 tearing down a device.
820 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
825 Description: service threads may hog cpus when there are a lot of requests
826 Details : Insert cond_resched to give other threads a chance to use some CPU
831 Description: the cat log was not initialized in recovery
832 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
833 unlink log on mds will not be initialized until mds post
834 recovery. And also in mds post recovery, the unlink log will
835 initialization will be done asynchronausly, so there will be race
836 between add unlink log and unlink log initialization.
840 Description: brw_stats were being printed incorrectly
841 Details : brw_stats were being printed as log2 but all of them were not
842 recorded as log2. Also remove some code duplication arising from
843 filter_tally_{read,write}.
847 Frequency : rare, only in recovery.
848 Description: ASSERTION(req->rq_type != LI_POISON) failed
849 Details : imp_lock should be held while iterating over imp_sending_list for
850 prevent destroy request after get timeout in ptlrpc_queue_wait.
854 Description: replay-single.sh test 52 fails
855 Details : A lock's skiplist need to be cleanup when it being unlinked
856 from its resource list.
860 Description: Short directio read returns full requested size rather than
862 Details : Direct I/O operations should return actual amount of bytes
863 transferred rather than requested size.
865 Severity : enhancement
867 Description: metadata RPC reduction (e.g. for rm performance)
868 Details : decrease the amount of synchronous RPC between clients and servers
869 by canceling conflicing lock before the operation on the client
870 and packing thier handles into the main operation RPC to server.
872 Severity : enhancement
874 Description: Async OSC create to avoid the blocking unnecessarily.
875 Details : If a OST has no remain object, system will block on the creating
876 when need to create a new object on this OST. Now, ways use
877 pre-created objects when available, instead of blocking on an
878 empty osc while others are not empty. If we must block, we block
879 for the shortest possible period of time.
881 Severity : enhancement
883 Description: refine locking for avoid write wrong info into lov_objid file
884 Details : fix possible races with add new target and write/update data in
887 --------------------------------------------------------------------------------
889 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
894 Frequency : on some architectures
896 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
897 Details : Uninitialized flags sometimes cause configuration commands to
901 Frequency : patchless clients only
903 Description: missing __iget() symbol export
904 Details : The __iget() symbol export is missing. To avoid the need for
905 this on patchless clients the deathrow inode reaper is turned
906 off, and we depend on the VM to clean up old inodes. This
907 dependency was during via the fix for bug 12181.
909 --------------------------------------------------------------------------------
911 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
913 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
914 INTEROPERATE with older versions automatically. In many cases a
915 special upgrade step is needed. Please read the
916 user documentation before upgrading any part of a live system.
917 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
918 version will not interoperate with 1.6 betas before beta5 (1.5.95).
919 * WARNING: Lustre configuration and startup changes are required with
920 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
926 Severity : enhancement
928 Description: MountConf
929 Details : Lustre configuration is now managed via mkfs and mount
930 commands instead of lmc and lconf. New obd types (MGS, MGC)
931 are added for dynamic configuration management. See
932 https://mail.clusterfs.com/wikis/lustre/MountConf for
935 Severity : enhancement
937 Description: dynamic OST addition
938 Details : OSTs can now be added to a live filesystem
940 Severity : enhancement
942 Description: startup order invariance
943 Details : MDTs and OSTs can be started in any order. Clients only
944 require the MDT to complete startup.
946 Severity : enhancement
948 Description: parallel, asynchronous orphan cleanup
949 Details : orphan cleanup is now performed in separate threads for each
950 OST, allowing parallel non-blocking operation.
952 Severity : enhancement
954 Description: optimized stripe assignment
955 Details : stripe assignments are now made based on ost space available,
956 ost previous usage, and OSS previous usage, in order to try
957 to optimize storage space and networking resources.
959 Severity : enhancement
961 Description: Permanently set tunables
962 Details : All writable /proc/fs/lustre tunables can now be permanently
963 set on a per-server basis, at mkfs time or on a live system.
965 Severity : enhancement
967 Description: Lustre message v2
968 Details : Add lustre message format v2.
970 Severity : enhancement
972 Description: client OST exclusion list
973 Details : Clients can be started with a list of OSTs that should be
974 declared "inactive" for known non-responsive OSTs.
978 Description: SPEC SFS validation failure on NFS v2 over lustre.
979 Details : Changes the blocksize for regular files to be 2x RPC size,
980 and not depend on stripe size.
982 Severity : enhancement
984 Description: Multiple MD RPCs in flight.
985 Details : Further unserialise some read-only MDS RPCs - learn about intents.
986 To avoid overly-overloading MDS, introduce a limit on number of
987 MDS RPCs in flight for a single client and add /proc controls
988 to adjust this limit.
990 Severity : enhancement
992 Description: client read/write statistics
993 Details : Add client read/write call usage stats for performance
994 analysis of user processes.
995 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
996 file access. extents_stats shows chunk size distribution.
997 extents_stats_per_process show chunk size distribution per
1000 Severity : enhancement
1002 Description: mds statistics
1003 Details : Add detailed mds operations statistics in
1004 /proc/fs/lustre/mds/*/stats.
1008 Description: Failure of copying files with lustre special EAs.
1009 Details : Client side always return success for setxattr call for lustre
1010 special xattr (currently only "trusted.lov").
1015 Description: Refcount LNET uuids
1016 Details : The global LNET uuid list grew linearly with every startup;
1017 refcount repeated list entries instead of always adding to
1020 Severity : enhancement
1022 Description: Dynamic service threads
1023 Details : Within a small range, start extra service threads
1024 automatically when the request queue builds up.
1027 Frequency : mixed-endian client/server environments
1029 Description: mixed-endian crashes
1030 Details : The new msg_v2 system had some failures in mixed-endian
1033 Severity : enhancement
1035 Description: Easy OST removal
1036 Details : OSTs can be permanently deactivated with e.g. 'lctl
1037 conf_param lustre-OST0001.osc.active=0'
1039 Severity : enhancement
1041 Description: MGS proc entries
1042 Details : Added basic proc entries for the MGS showing what filesystems
1045 Severity : enhancement
1047 Description: provide MGS failover
1048 Details : Added config lock reacquisition after MGS server failover.
1050 Severity : enhancement
1052 Description: add Linux 2.4 support
1053 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1057 Description: a large application tries to do I/O to the same resource and dies
1058 in the middle of it.
1059 Details : Check the req->rq_arrival time after the call to
1060 ost_brw_lock_get(), but before we do anything about
1061 processing it & sending the BULK transfer request. This
1062 should help move old stale pending locks off the queue as
1063 quickly as obd_timeout.
1066 Frequency : when an incorrect nid is specified during startup
1068 Description: ptlrpc connect to non-existant node causes kernel crash
1069 Details : LNET can't be re-entered from an event callback, which
1070 happened when we expire a message after the export has been
1071 cleaned up. Instead, hand the zombie cleanup off to another
1074 Severity : enhancement
1076 Description: plain/inodebits lock performance improvement
1077 Details : Grouping plain/inodebits in granted list by their request modes
1078 and bits policy, thus improving the performance of search through
1082 Frequency : only if OST filesystem is corrupted
1084 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1085 Details : for a short time RPCs with bulk IO are in the replay list,
1086 but replay of bulk IOs is unimplemented. If the OST filesystem
1087 is corrupted due to disk cache incoherency and then replay is
1088 started it is possible to trip an assertion. Avoid putting
1089 committed RPCs into the replay list at all to avoid this issue.
1092 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1094 Description: System hang on startup
1095 Details : This bug allowed the liblustre (e.g. catamount) client to
1096 return to the app before handling all startup RPCs. This
1097 could leave the node unresponsive to lustre network traffic
1098 and manifested as a server ptllnd timeout.
1100 Severity : enhancement
1102 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1103 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1104 boolean to control whether to print peer debug info when a
1105 client's RPC times out.
1108 Frequency : only for kernels with patches from Lustre below 1.4.3
1110 Description: Remove old rdonly API
1111 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1114 Frequency : only for devices with external journals
1116 Description: Set external device read-only also
1117 Details : During a commanded failover stop, we set the disk device
1118 read-only while the server shuts down. We now also set any
1119 external journal device read-only at the same time.
1122 Frequency : when upgrading from 1.4 while trying to change parameters
1124 Description: The wrong (new) MDC name was used when setting parameters for
1125 upgraded MDT's. Also allows changing of OSC (and MDC)
1126 parameters if --writeconf is specified at tunefs upgrade time.
1129 Frequency : when setting specific ost indicies
1131 Description: QOS code breaks on skipped indicies
1132 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1133 created with --index need not be sequential.
1135 Severity : enhancement
1137 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1138 Details : The uninit_groups feature works in conjunction with the kernel
1139 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1140 up the pass1 processing of e2fsck. This is a read-only feature
1141 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1142 mount filesystems that have had this feature enabled.
1144 Severity : enhancement
1146 Description: Improve multi-block allocation algorithm to avoid fragmentation
1147 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1148 allocation locality and avoid filesystem fragmentation.
1150 ------------------------------------------------------------------------------
1152 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1154 * Support for kernels:
1155 2.4.21-47.0.1.EL (RHEL 3)
1156 2.6.5-7.283 (SLES 9)
1157 2.6.9-42.0.10.EL (RHEL 4)
1158 2.6.12.6 vanilla (kernel.org)
1159 2.6.16.27-0.9 (SLES 10)
1160 * Recommended e2fsprogs version: 1.39.cfs5
1162 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1166 Frequency : occasional, depends on client load and configuration
1167 Bugzilla : 12181, 12203
1168 Description: data loss for recently-modified files
1170 Details : In some cases it is possible that recently written or created
1171 files may not be written to disk in a timely manner (this should
1172 normally be within 30s unless client IO load is very high).
1173 The problem appears as zero-length files or files that are a
1174 multiple of 1MB in size after a client crash or client eviction
1175 that are missing data at the end of the file.
1177 This problem is more likely to be hit on clients where files are
1178 repeatedly created and unlinked in the same directory, clients
1179 have a large amount of RAM, have many CPUs, the filesystem has
1180 many OSTs, the clients are rebooted frequently, and/or the files
1181 are not accessed by other nodes after being written.
1183 The presence of the problem can be detected by looking at
1184 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1185 smaller than the second (nr_unused) then dirty files will not
1186 be flushed automatically to disk. "sync; sleep 10" should be
1187 run several times on the node before unmounting it to update
1188 Lustre (this is also safe to run on nodes without this problem).
1190 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1191 that can cause this same problem, so customers using that kernel
1192 also need to update the kernel in addition to Lustre. In order
1193 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1195 It is normal that files written just before a client crash (less
1196 than 30s) may not yet have been flushed to disk, even for local
1200 Frequency : frequent on thin XT3 nodes
1202 Description: UUID collision on thin XT3 Linux nodes
1203 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1204 randomly, since we relied on an insufficiently-seeded PRNG.
1209 Description: OSS hangs after "All ost request buffers busy"
1210 Details : A deadlock between quota and journal operations caused OSS
1211 hangs after printing "All ost request buffers busy."
1214 Frequency : always on liblustre builds
1216 Description: Cleanup compiler warnings on liblustre
1219 Frequency : always on liblustre builds on XT3
1221 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1225 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1226 > 32000 subdirectories
1227 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1228 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1229 subdirectory count crosses 32000. This will aid e2fsck to
1230 correctly handle more than 32000 subdirectories.
1233 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1235 Description: System hang on startup
1236 Details : This bug allowed the liblustre (e.g. catamount) client to
1237 return to the app before handling all startup RPCs. This
1238 could leave the node unresponsive to lustre network traffic
1239 and manifested as a server ptllnd timeout.
1241 Severity : enhancement
1243 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1244 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1245 boolean to control whether to print peer debug info when a
1246 client's RPC times out.
1251 Description: make O_SYNC working on 2.6 kernels
1252 Details : 2.6 kernels use different method for mark pages for write,
1253 so need add a code to lustre for O_SYNC work.
1258 Description: Failure to close file and release space on NFS
1259 Details : Put inode details into lock acquired in ll_intent_file_open.
1260 Use mdc_intent_lock in ll_intent_open to properly
1261 detect all kind of errors unhandled by mdc_enqueue.
1266 Description: proc file read during shutdown sometimes raced obd removal,
1268 Details : Add lock to prevent obd access after proc file removal.
1271 Frequency : Only for files larger than 4GB on 32-bit clients.
1273 Description: improperly doing page alignment of locks
1274 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1275 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1280 Description: RPCs being resent when they shouldn't be
1281 Details : Some RPCs that should not be resent are being resent. This
1282 can cause inconsistencies in the RPC state machine. Do not
1283 resend such requests.
1286 Frequency : rare, only with NFS export
1288 Description: Crash on NFS re-export node
1289 Details : under very unusual load conditions an assertion is hit in
1290 ll_intent_file_open()
1293 Frequency : only if OST filesystem is corrupted
1295 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1296 Details : for a short time RPCs with bulk IO are in the replay list,
1297 but replay of bulk IOs is unimplemented. If the OST filesystem
1298 is corrupted due to disk cache incoherency and then replay is
1299 started it is possible to trip an assertion. Avoid putting
1300 committed RPCs into the replay list at all to avoid this issue.
1305 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1306 Details : Large single O_DIRECT read and write calls can fail to allocate
1307 a sufficiently large buffer to process the request. In case of
1308 allocation failure the allocation is retried with a smaller
1309 buffer and broken into smaller requests.
1311 Severity : enhancement
1313 Description: Add -o localflock option to simulate old noflock behaviour.
1314 Details : This will achieve local-only flock/fcntl locks coherentness.
1319 Description: versioning check is incomplete
1320 Details : Checking the version difference of client vs. server, report
1321 error if the gap is too big.
1326 Description: add support PG_writeback bit
1327 Details : add support for PG_writeback bit for Lustre, for more carefull
1328 work with page cache in 2.6 kernel. This also fix some deadlocks
1329 and remove hack for work O_SYNC with 2.6 kernel.
1331 Severity : enhancement
1333 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1334 Details : The uninit_groups feature works in conjunction with the kernel
1335 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1336 up the pass1 processing of e2fsck. This is a read-only feature
1337 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1338 mount filesystems that have had this feature enabled.
1340 Severity : enhancement
1342 Description: Improve multi-block allocation algorithm to avoid fragmentation
1343 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1344 allocation locality and avoid filesystem fragmentation.
1346 ------------------------------------------------------------------------------
1348 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1350 * Support for kernels:
1351 2.6.9-42.0.3.EL (RHEL 4)
1352 2.6.5-7.276 (SLES 9)
1353 2.4.21-47.0.1.EL (RHEL 3)
1354 2.6.12.6 vanilla (kernel.org)
1355 2.6.16.21-0.8 (SLES10)
1356 * Recommended e2fsprogs version: 1.39.cfs2-0
1358 * The backwards-compatible /proc/sys/portals symlink has been removed
1359 in this release. Before upgrading, please ensure that you change
1360 any configuration scripts or /etc/sysctl.conf files that access
1361 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1362 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1363 in advance of the upgrade on any system running Lustre 1.4.6 or
1364 newer, since /proc/sys/lnet was added in that version.
1365 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1369 Frequency : only when quota is used
1371 Description: avoid scanning export list for quota master
1372 Details : Change the algorithms to avoid scanning export list in order
1373 to improve the efficiency.
1376 Frequency : MDS failover only, very rarely
1378 Description: "went back in time" messages on mds failover
1379 Details : The greatest transno may be lost when the current operation
1380 finishes with an error (transno==0) and the client's last_rcvd
1381 record is over-written. Save the greatest transno in the
1382 mds_last_transno for this case.
1385 Frequency : always for specific kernels and striping counts
1387 Description: client may get "Matching packet too big" without ACL support
1388 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1389 when trying to access files in certain configurations. The
1390 clients should in fact be denied when mounting because they do
1391 not understand ACLs.
1394 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1396 Description: many clients connecting with IO in progress causes connect timeouts
1397 Details : Avoid synchronous journal commits to avoid delays caused by many
1398 clients connecting/disconnecting when bulk IO is in progress.
1399 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1400 OST_IO_PORTAL to avoid delays behind potentially many pending
1404 Frequency : occasionally with multiple writers to a single file
1406 Description: shared writes to file may result in wrong size reported by stat()
1407 Details : Allow growing of kms when extent lock is cancelled
1410 Frequency : always with random mmap IO to multi-striped file
1412 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1413 Details : Only if the hole is at the end of OST object so that kms is too
1414 small. Fix is to increase kms accordingly in ll_nopage.
1417 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1419 Description: writes to a missing object would leak memory on the OST
1420 Details : If there is an inconsistency between the MDS and OST filesystems,
1421 such that the MDS references an object that doesn't exist, writes
1422 to that object will leak memory due to incorrect cleanup in the
1423 error handling path, eventually running out of memory on the OST.
1428 Description: Creating too long symlink causes lustre errors
1429 Details : Check symlink and name lengths before sending requests to MDS.
1432 Frequency : only if flock is enabled (not on by default)
1434 Description: posix locks not released on fd closure on 2.6.9+
1435 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1436 kernels, this caused such locks not to be released on fd close and
1437 then assertions on fs unmount about still used locks.
1440 Frequency : MDS failover only, very rarely
1442 Description: clients may get ASSERTION(granted_lock != NULL)
1443 Details : When request was taking a long time, and a client was resending
1444 a getattr by name lock request. The were multiple lock requests
1445 with the same client lock handle and
1446 mds_getattr_name->fixup_handle_for_resent_request found one of the
1447 lock handles but later failed with ASSERTION(granted_lock != NULL).
1452 Description: handle->h_buffer_credits > 0, assertion failure
1453 Details : h_buffer_credits is zero after truncate, causing assertion
1454 failure. This patch extends the transaction or creates a new
1458 Frequency : NFS re-export or patchless client
1459 Bugzilla : 11179, 10796
1460 Description: Crash on NFS re-export node (__d_move)
1461 Details : We do not want to hash the dentry if we don't have a lock.
1462 But if this dentry is later used in d_move, we'd hit uninitialised
1463 list head d_hash, so we just do this to init d_hash field but
1464 leave dentry unhashed.
1467 Frequency : NFS re-export or patchless client
1469 Description: NFS exports has problem with symbolic link
1470 Details : lustre client didn't properly install dentry when re-exported
1471 to NFS or running patchless client.
1474 Frequency : NFS re-export or patchless client
1476 Description: Various nfs/patchless fixes.
1477 Details : fixes reuse disconected alias for lookup process - this fixes
1478 warning "find_exported_dentry: npd != pd",
1479 fix permission error with open files at nfs.
1480 fix apply umask when do revalidate.
1483 Frequency : occasional
1485 Description: Crash on NFS re-export node
1486 Details : calling clear_page() on the wrong pointer triggered oops in
1487 generic_mapping_read().
1490 Frequency : rarely, using O_DIRECT IO
1492 Description: unaligned directio crashes client with LASSERT
1493 Details : check for unaligned buffers before trying any requests.
1496 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1498 Description: stale data returned from RAID cache
1499 Details : If only a small amount of IO is done to the RAID device before
1500 reading it again it is possible to get stale data from the RAID
1501 cache instead of reading it from disk.
1504 Frequency : always for sles10 kernel
1506 Description: sles10 support
1507 Details : ll_follow_link: compile fixes and using of nd_set_link
1508 under newer kernels.
1511 Frequency : depends on arch, kernel and compiler version, always on sles10
1514 Description: recursive or deep enough symlinks cause stack overflow
1515 Details : getting rid of large stack-allocated variable in
1519 Frequency : depends on hardware
1521 Description: lustre write performance loss in the SLES10 kernel
1522 Details : the performance loss is caused by using of write barriers in the
1523 ext3 code. The SLES10 kernel turns barrier support on by
1524 default. The fix is to undo that change for ldiskfs.
1526 ------------------------------------------------------------------------------
1528 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1530 * Support for kernels:
1531 2.6.9-42.0.3EL (RHEL 4)
1532 2.6.5-7.276 (SLES 9)
1533 2.4.21-47.0.1.EL (RHEL 3)
1534 2.6.12.6 vanilla (kernel.org)
1538 Frequency : quota enabled and large files being deleted
1540 Description: releasing more than 4GB of quota at once hangs OST
1541 Details : If a user deletes more than 4GB of files on a single OST it
1542 will cause the OST to spin in an infinite loop. Release
1543 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1548 Description: statfs data retrieved from /proc may be stale or zero
1549 Details : When reading per-device statfs data from /proc, in the
1550 {kbytes,files}_{total,free,avail} files, it may appear
1551 as zero or be out of date.
1554 Frequency : systems with MD RAID1 external journal devices
1556 Description: lconf's call to blkid is confused by RAID1 journal devices
1557 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1558 of returning all block devices that match the journal UUID.
1561 Frequency : always, for aggregate stripe size over 4GB
1563 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1564 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1565 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1568 Frequency : always if "lfs find" used on a local file/directory
1570 Description: "lfs find" segfaults if used on a local file/directory
1571 Details : The case where a directory component was not specified wasn't
1572 handled correctly. Handle this properly.
1575 Frequency : always on ppc64
1577 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1578 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1579 perform endianness conversion.
1582 Frequency : rarely (truncate to non-zero file size after write under load)
1583 Bugzilla : 10730, 10687
1584 Description: Files padded with zeros to next 4K multiple
1585 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1586 it is possible that files that are truncated to a non-zero size
1587 immediately after being written are filled with zero bytes beyond
1588 the truncated size. No file data is lost.
1590 Severity : enhancement
1592 Description: Allow recovery/failover for liblustre clients.
1593 Details : liblustre clients were unaware of failover configurations until
1596 Severity : enhancement
1598 Description: user file locks should fail when not mounting with flock option
1599 Details : Set up an error-returning stub in ll_file_operations.lock field
1600 to prevent incorrect behaviour when client is mounted without
1601 flock option. Also, set up properly f_op->flock field for
1605 Frequency : always on ia64
1607 Description: "lfs df" loops on printing out MDS statfs information
1608 Details : The obd_ioctl_data was not initialized and in some systems
1609 this caused a failure during the ioctl that did not return
1610 an error. Initialize the struct and return an error on failure.
1613 Frequency : SLES 9 only
1615 Description: Error of copying files with lustre special EAs as root
1616 Details : Client side always return success for setxattr call for lustre
1617 special xattr (currently only "trusted.lov").
1620 Frequency : rarely on clusters with both ia64+i386 clients
1622 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1623 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1624 clients are doing concurrent writes to the same file it is
1625 possible that the ia64 clients may LASSERT because the OST
1626 extent locks are not PAGE_SIZE aligned. Ensure that grown
1627 locks are always aligned on the request boundary.
1630 Frequency : specific use, occasional
1632 Description: Overwriting in use executable truncates on-disk binary image
1633 Details : If one node attempts to overwrite an executable in use by
1634 another node, we now correctly return ETXTBSY instead of
1635 truncating the file.
1637 Severity : enhancement
1639 Description: Async OSC create to avoid the blocking unnecessarily.
1640 Details : If a OST has no remain object, system will block on the creating
1641 when need to create a new object on this OST. Now, ways use
1642 pre-created objects when available, instead of blocking on an
1643 empty osc while others are not empty. If we must block, we block
1644 for the shortest possible period of time.
1649 Description: chmod on Lustre root is propagated to other clients
1650 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1651 invalid by the follow_mount time.
1654 Frequency : liblustre clients only
1656 Description: Race in 'instant cancel' lock handling could lead to such locks
1657 never to be granted in case of SMP MDS
1658 Details : Do not destroy not yet granted but cbpending locks in
1662 Frequency : replay/resend of open
1664 Description: non null lock assetion failure in mds_intent_policy
1665 Details : Trying to replay/resend lockless open requests resulted in
1666 mds_open() returning 0 with no lock. Now it sets a flag if
1667 a lock is going to be returned.
1669 Severity : enhancement
1671 Description: Checksum enhancements
1672 Details : New checksum enhancements allow for resending RPCs that failed
1675 Severity : enhancement
1677 Description: Tunables on number of dirty pages in cacche
1678 Details : Allow to set limit on number of dirty pages cached.
1683 Description: client crash on unmount - lock still has references
1684 Details : In some error handling cases it was possible to leak a lock
1685 reference on a client while accessing a file. This was not
1686 harmful to the client during operation, but would cause the
1687 client to crash when the filesystem is unmounted.
1690 Frequency : specific case, rare
1692 Description: ETXTBSY on mds though file not in use by client
1693 Details : ETXTBSY is no longer incorrectly returned when attempting to
1694 chmod or chown a directory that the user previously tried to
1695 execute or a currently-executing binary.
1698 Frequency : extremely rare except on liblustre-based clients
1700 Description: Lustre space not freed when files are deleted
1701 Details : Clean up open-unlinked files after client eviction. Previously
1702 the unlink was skipped and the files remained as orphans.
1707 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1708 Details : In some cases it was possible to send a blocking callback to
1709 a client doing a glimpse, even though that client didn't get
1710 a lock granted. When the glimpse lock is cancelled on the OST
1711 the freed lock is left on the waiting list and corrupted the list.
1714 Frequency : all core dumps
1716 Description: Broke core dumps to lustre
1717 Details : Negative dentry may be unhashed if parent does not have UPDATE
1718 lock, but some callers, e.g. do_coredump, expect dentry to be
1719 hashed after successful create, hash it in ll_create_it.
1721 ------------------------------------------------------------------------------
1723 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1725 * Support for kernels:
1726 2.6.9-42.0.2.EL (RHEL 4)
1727 2.6.5-7.276 (SLES 9)
1728 2.4.21-40.EL (RHEL 3)
1729 2.6.12.6 vanilla (kernel.org)
1733 Frequency : always on RHEL 3
1735 Description: Number of open files grows over time
1736 Details : The number of open files grows over time, whether or not
1737 Lustre is started. This was due to a filp leak introduced
1738 by one of our kernel patches.
1740 ------------------------------------------------------------------------------
1742 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1744 * Support for kernels:
1745 2.6.9-42.EL (RHEL 4)
1746 2.6.5-7.276 (SLES 9)
1747 2.4.21-40.EL (RHEL 3)
1748 2.6.12.6 vanilla (kernel.org)
1753 Bugzilla : 5719, 9635, 9792, 9684
1754 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1755 Details : If a server is under heavy load and cannot reply to new
1756 connection requests before the client resends the (re)connect,
1757 the connection handling code can behave badly if two service
1758 threads are concurrently handing separate (re)connections from
1759 the same client. Add better locking to the connection handling
1760 code, and ensure that only a single connection will be processed
1761 for a given client UUID, even if the lock is dropped.
1763 Severity : enhancement
1765 Description: add TCP zero-copy support to kernel
1766 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1767 sends if the hardware supports scatter-gather and checksumming.
1768 This allows socklnd to do client-write and server-read more
1769 efficiently and reduce CPU utilization from skbuf copying.
1772 Frequency : only if NFS exporting from client
1774 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1775 Details : If setting attributes on a file created under NFS that had
1776 never been opened it would be possible to oops the client
1777 if the file had no objects.
1780 Frequency : always for liblustre
1782 Description: liblustre client does MDS+OSTs setattr RPC for each write
1783 Details : When doing a write from a liblustre client, the client
1784 incorrectly issued an RPC to the MDS and each OST the file was
1785 striped over in order to update the timestamps. When writing
1786 with small chunks and many clients this could overwhelm the MDS
1787 with RPCs. In all cases it would slow down the write because
1788 these RPCs are unnecessary.
1790 Severity : enhancement
1792 Description: allow number of MDS service threads to be changed at module load
1793 Details : It is now possible to change the number of MDS service threads
1794 running. Adding "options mds mds_num_threads={N}" to the MDS's
1795 /etc/modprobe.conf will set the number of threads for the next
1796 time Lustre is restarted (assuming the "mds" module is also
1797 reloaded at that time). The default number of threads will
1798 stay the same, 32 for most systems.
1803 Description: OST crash if filesystem is unformatted or corrupt
1804 Details : If an OST is started on a device that has never been formatted
1805 or if the filesystem is corrupt and cannot even mount then the
1806 error handling cleanup routines would dereference a NULL pointer.
1811 Description: NULL pointer deref in llap_from_page.
1812 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1813 incorrect) mapping if the page was truncated/reclaimed while it was
1814 searched for. Check for this condition and skip such pages when
1815 doing readahead. Introduce extra check to llap_from_page() to
1816 verify page->mapping->host is non-NULL (so page is not anonymous).
1819 Frequency : Sometimes when using sys_sendfile
1821 Description: "page not covered by a lock" warnings from ll_readpage
1822 Details : sendfile called ll_readpage without right page locks present.
1823 Now we introduced ll_file_sendfile that does necessary locking
1824 around call to generic_file_sendfile() much like we do in
1828 Frequency : with certain MDS communication failures at client mount time
1830 Description: NULL pointer deref after failed client mount
1831 Details : a client connection request may delayed by the network layer
1832 and not be sent until after the PTLRPC layer has timed out the
1833 request. If the client fails the mount immediately it will try
1834 to clean up before the network times out the request. Add a
1835 reference from the request import to the obd device and delay
1836 the cleanup until the network drops the request.
1839 Frequency : occasionally during client (re)connect
1841 Description: assertion failure during client (re)connect
1842 Details : processing a client connection request may be delayed by the
1843 client or server longer than the client connect timeout. This
1844 causes the client to resend the connection request. If the
1845 original connection request is replied in this interval, the
1846 client may trip an assertion failure in ptlrpc_connect_interpret()
1847 which thought it would be the only running connect process.
1850 Frequency : only with obd_echo servers and clients that are rebooted
1852 Description: kernel BUG accessing uninitialized data structure
1853 Details : When running an obd_echo server it did not start the ping_evictor
1854 thread, and when a client was evicted an uninitialized data
1855 structure was accessed. Start the ping_evictor in the RPC
1856 service startup instead of the OBD startup.
1858 Severity : enhancement
1859 Bugzilla : 10193 (patchless)
1860 Description: Remove dependency on various unexported kernel interfaces.
1861 Details : No longer need reparent_to_init, exit_mm, exit_files,
1862 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1865 Frequency : rare (only users of deprecated and unsupported LDAP config)
1867 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1868 Details : LDAP apparently contains 'lustreName' attributes instead of
1869 'name'. A simple remapping of the name is sufficient.
1872 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1874 Description: waiting_locks_callback trips kernel BUG if client is evicted
1875 Details : Running with the dump_on_timeout debug flag turned on makes
1876 it possible that the waiting_locks_callback() can try to dump
1877 the Lustre kernel debug logs from an interrupt handler. Defer
1878 this log dumping to the expired_lock_main() thread.
1880 Severity : enhancement
1882 Description: Support NFS exporting on 2.6 kernels.
1883 Details : Implement non-rawops metadata methods for NFS server to use without
1884 changing NFS server code.
1887 Frequency : very rare (synthetic metadata workload only)
1889 Description: two racing renames might cause an MDS thread to deadlock
1890 Details : Running the "racer" program may cause one MDS thread to rename
1891 a file from being the source of a rename to being the target of
1892 a rename at exactly the same time that another thread is doing
1893 so, and the second thread has already enqueued these locks after
1894 doing a lookup of the target and is trying to relock them in
1895 order. Ensure that we don't try to re-lock the same resource.
1898 Frequency : only very large systems with liblustre clients
1900 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1901 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1902 all OSTs in parallel. This allows the eviction of stale liblustre
1903 clients to proceed much faster than if they were done in series,
1904 and also offers similar improvements for other set_info RPCs.
1909 Description: excessive CPU usage during initial read phase on client
1910 Details : During the initial read phase on a client, it would agressively
1911 retry readahead on the file, consuming too much CPU and impacting
1912 performance (since 1.4.5.8). Improve the readahead algorithm
1913 to avoid this, and also improve some other common cases (read
1914 of small files in particular, where "small" is files smaller than
1915 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1920 Description: MDS crash when receiving packet with unknown intent.
1921 Details : Do not LBUG in unknown intent case, just return -EFAULT
1923 Severity : enhancement
1924 Bugzilla : 9293, 9385
1925 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1926 Details : Do not serialize getattr (non-intent version) and statfs.
1929 Frequency : occasional, when OST network is overloaded/intermittent
1931 Description: client evicted by OST after bulk IO timeout
1932 Details : If a client sends a bulk IO request (read or write) the OST
1933 may evict the client if it is unresposive to its data GET/PUT
1934 request. This is incorrect if the network is overloaded (takes
1935 too long to transfer the RPC data) or dropped the OST GET/PUT
1936 request. There is no need to evict the client at all, since
1937 the pinger and/or lock callbacks will handle this, and the
1938 client can restart the bulk request.
1941 Frequency : Always when mmapping file with no objects
1943 Description: client crashes when mmapping file with no objects
1944 Details : Check that we actually have objects in a file before doing any
1945 operations on objects in ll_vm_open, ll_vm_close and
1951 Description: Request leak when working with deleted CWD
1952 Details : Introduce advanced request refcount tracking for requests
1953 referenced from lustre intent.
1955 Severity : Enhancement
1957 Description: Cache open file handles on client.
1958 Details : MDS now will return special lock along with openhandle, if
1959 requested and client is allowed to hold openhandle, even if unused,
1960 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1961 closing files for every read/write openration.
1963 Severity : Enhancement
1965 Description: Cache open negative dentries on client when possible.
1966 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1967 negative dentries on lock revocation.
1972 Description: Remounting a client read-only wasn't possible with a zconf mount
1973 Details : It wasn't possible to remount a client read-only with llmount.
1975 Severity : enhancement
1976 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1977 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1978 MPICH is included to add Lustre-specific ADIO interfaces.
1979 This is based closely on the UFS ADIO layer and only differs
1980 in file creation, in order to allow the OST striping to be set.
1981 This is user-contributed code and not supported by CFS.
1986 Description: extended inode attributes (immutable, append-only) work improperly
1987 when 2.4 and 2.6 kernels are used on client/server or vice versa
1988 Details : Introduce kernel-independent values for these flags.
1990 Severity : enhancement
1993 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1994 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1995 granularity. Now, Fractional MB granularity is be supported,
1996 this is very useful for low memory system.
1998 Severity : enhancement
2000 Description: Getattr by fid
2001 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2002 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2003 have locks and avoids getattr rpc requests for it on every stat.
2006 Frequency : Always, for filesystems larger than 2TB
2008 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2009 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2010 filesystems larger than 2TB to be mounted with mballoc enabled.
2013 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2015 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2016 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2017 filesystems larger than 2TB is mounted then the kernel will
2018 silently corrupt the start of the filesystem. CONFIG_LBD is
2019 enabled for all CFS-supported kernels, but the possibility of
2020 this happening with a modified kernel config exists.
2022 Severity : enhancement
2024 Description: add client O_DIRECT support for 2.6 kernels
2025 Details : It is now possible to do O_DIRECT reads and writes to files
2026 in the Lustre client mountpoint on 2.6 kernel clients.
2028 Severity : enhancement
2030 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2031 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2032 parallel, not waiting for response from every OST before sending
2033 a rpc to the next OST.
2038 Description: setattr vs write race when updating file timestamps
2039 Details : Client processes that update a file timestamp into the past
2040 right after writing to the file (e.g. tar) it is possible that
2041 the updated file modification time can be reset to the current
2042 time due to a race between processing the setattr and write RPC.
2044 Severity : enhancement
2046 Description: Bring 'lfs find' closer in line with regular Linux find.
2047 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2048 -print0 options and obtains all the needed info through the lustre
2051 Severity : enhancement
2053 Description: support up to 1024 configured devices on one node
2054 Details : change obd_dev array from statically allocated to dynamically
2055 allocated structs as they are first used to reduce memory usage
2060 Description: Flush dirty partially truncated pages during truncate
2061 Details : Immediatelly flush partially truncated pages in filter_setattr,
2062 this way we completely avoid having any pages in page cache on OST
2063 and can retire ugly workarounds during writes to flush such pages.
2068 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2069 Details : protect inode from truncation within vfs_unlink() context
2070 just take a reference before calling vfs_unlink() and release it
2071 when parent's i_sem is free.
2076 Description: last_id value checked outside lock on OST caused LASSERT failure
2077 Details : If there were multiple MDS->OST object precreate requests in
2078 flight, it was possible that the OST's last object id was checked
2079 outside a lock and incorrectly tripped an assertion. Move checks
2080 inside locks, and discard old precreate requests.
2083 Frequency : always, if extents are used on OSTs
2085 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2086 Details : OSTs using the extents format would not zero the high 16 bits of
2087 the index physical block number. This is not a problem for any
2088 OST filesystems smaller than 16TB, and no kernels support ext3
2089 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2090 new/modified files) and can be fixed for existing filesystems
2091 with e2fsprogs-1.39-cfs1.
2096 Description: import connection selection may be incorrect if timer wraps
2097 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2098 import connections to be ignored if the 32-bit jiffies counter
2099 wraps. Use a 64-bit jiffies counter.
2102 Frequency : very large clusters immediately after boot
2104 Description: LNET request buffers exhausted under heavy short-term load
2105 Details : If a large number of client requests are generated on a service
2106 that has previously never seen so many requests it is possible
2107 that the request buffer growth cannot keep up with the spike in
2108 demand. Instead of dropping incoming requests, they are held in
2109 the LND until the RPC service can accept more requests.
2112 Frequency : Sometimes during replay
2114 Description: Assertion failure in ll_local_open after replay.
2115 Details : If replay happened on an open request reply before we were able
2116 to set replay handler, reply will become not swabbed tripping the
2117 assertion in ll_local_open. Now we set the handler right after
2118 recognising of open request
2121 Frequency : very rare
2123 Description: kernel reports "badness in vsnprintf"
2124 Details : Reading from the "recovery_status" /proc file in small chunks
2125 may cause a negative length in lprocfs_obd_rd_recovery_status()
2126 call to vsnprintf() (which is otherwise harmless). Exit early
2127 if there is no more space in the output buffer.
2129 Severity : enhancement
2131 Description: clear OBD RPC statistics by writing to them
2132 Details : It is now possible to clear the OBD RPC statistics by writing
2133 to the "stats" file.
2138 Description: Client mtime is not the same on different clients after utimes
2139 Details : In some cases, the client was using the utimes() syscall on
2140 a file cached on another node. The clients now validate the
2141 ctime from the MDS + OSTs to determine which one is right.
2146 Description: Inability to activate failout mode
2147 Details : lconf script incorrectly assumed that in pythong string's numeric
2148 value is used in comparisons.
2151 Frequency : always with multiple stripes per file
2153 Description: Inefficient object allocation for mutli-stripe files
2154 Details : When selecting which OSTs to stripe files over, for files with
2155 a stripe count that divides evenly into the number of OSTs,
2156 the MDS is always picking the same starting OST for each file.
2157 Return the OST selection heuristic to the original design.
2162 Description: mount failures may take full timeout to return an error
2163 Details : Under some heavy load conditions it is possible that a
2164 failed mount can wait for the full obd_timeout interval,
2165 possibly several minutes, before reporting an error.
2166 Instead return an error as soon as the status is known.
2168 Frequency : quota enabled and large files being deleted
2170 Description: releasing more than 4GB of quota at once hangs OST
2171 Details : If a user deletes more than 4GB of files on a single OST it
2172 will cause the OST to spin in an infinite loop. Release
2173 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2178 Description: statfs data retrieved from /proc may be stale or zero
2179 Details : When reading per-device statfs data from /proc, in the
2180 {kbytes,files}_{total,free,avail} files, it may appear
2181 as zero or be out of date.
2184 Frequency : systems with MD RAID1 external journal devices
2186 Description: lconf's call to blkid is confused by RAID1 journal devices
2187 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2188 of returning all block devices that match the journal UUID.
2191 Frequency : always, for aggregate stripe size over 4GB
2193 Description: assertion fails when trying to use 4GB stripe size
2194 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2195 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2198 Frequency : always on ppc64
2200 Description: the first write on an ext3 filesystem with mballoc got stuck
2201 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2202 perform endianness conversion.
2204 ------------------------------------------------------------------------------
2206 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2208 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2209 INTEROPERATE with older versions automatically. Please read the
2210 user documentation before upgrading any part of a live system.
2211 * WARNING: Lustre networking configuration changes are required with
2212 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2215 * Support for kernels:
2216 2.6.9-22.0.2.EL (RHEL 4)
2217 2.6.5-7.244 (SLES 9)
2218 2.6.12.6 vanilla (kernel.org)
2221 Severity : enhancement
2222 Bugzilla : 7981/8208
2223 Description: Introduced Lustre Networking (LNET)
2224 Details : LNET is new networking infrastructure for Lustre, it includes
2225 a reorganized network configuration mode (see the user
2226 documentation for full details) as well as support for routing
2227 between different network fabrics. Lustre Networking Devices
2228 (LNDs) for the supported network fabrics have also been
2229 created for this new infrastructure.
2231 Severity : enhancement
2232 Description: Introduced Access control lists
2233 Details : clients can set ACLs on files and directories in order to have
2234 more fine-grained permissions than the standard Unix UGO+RWX.
2235 The MDS must be started with the "-o acl" mount option.
2237 Severity : enhancement
2238 Description: Introduced filesystem quotas
2239 Details : Administrators may now establish per-user quotas on the
2242 Severity : enhancement
2244 Description: Configuration change for the XT3
2245 The PTLLND is now used to run Lustre over Portals on the XT3
2246 The configure option(s) --with-cray-portals are no longer used.
2247 Rather --with-portals=<path-to-portals-includes> is used to
2248 enable building on the XT3. In addition to enable XT3 specific
2249 features the option --enable-cray-xt3 must be used.
2254 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2255 Details : A race between allocating a new llcd and re-getting the llcd_lock
2256 allowed another thread to grab newly-allocated llcd.
2258 Severity : enhancement
2260 Description: 2.6 OST async journal commit and locking fix to improve performance
2261 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2262 2.6 kernels are now async as they already were in 2.4 kernels so
2263 that they can commit concurrently with the network bulk transfer.
2264 For block-allocated files the filter allocation semaphore is held
2265 to avoid filesystem fragmentation during allocation. BKL lock
2266 removed for 2.6 xattr operations where it is no longer needed.
2271 Description: lconf incorrectly determined whether two IP networks could talk
2272 Details : In some more complicated routing and multiple-network
2273 configurations, lconf will avoid trying to make a network
2274 connection to a disjoint part of the IP space. It was doing the
2275 math incorrectly for one set of cases.
2280 Description: Fix for potential infinite loop processing records in an llog.
2281 Details : If an llog record is corrupted/zeroed, it is possible to loop
2282 forever in llog_process(). Validate the llog record length
2283 and skip the remainder of the block on error.
2286 Frequency : occasional (liblustre only)
2288 Description: liblustre could not open files whose last component is a symlink
2289 Details : sysio_path_walk() would incorrectly pass the open intent to
2290 intermediate path components.
2293 Frequency : rare (liblustre only with non-standard tuning)
2294 Bugzilla : 7201 (7350)
2295 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2296 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2297 after it has been released, only do this for referenced locks
2299 Severity : enhancement
2301 Description: specifying an (invalid) directory default stripe_size of -1
2302 would reset the directory default striping
2303 Details : stripe_size -1 was used internally to signal directory stripe
2304 removal, now use "all default" to signal dir stripe removal
2305 as a directory striping of "all default" is not useful
2308 Frequency : common for large clusters running liblustre clients
2310 Description: doing an ls when liblustre clients are running is slow
2311 Details : sending a glimpse AST to a liblustre client waits for every AST
2312 to time out, as liblustre clients will not respond. Since they
2313 cannot cache data we refresh the OST lock LVB from disk instead.
2315 Severity : enhancement
2317 Description: doing an ls at the same time as file IO can be slow
2318 Details : enqueue and other "small" requests can be blocked behind many
2319 large IO requests. Create a new OST IO portal for non-IO
2320 requests so they can be processed faster.
2323 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2325 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2326 created file with mode 000
2327 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2328 allow an ftruncate() as a non-root user to a file with mode 000.
2329 Lustre now respects this flag to disable mode checks when
2330 truncating a file owned by the user
2333 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2336 Description: Revoking locks from clients that went dead or catatonic might take
2338 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2339 cancellation of such locks instant on servers without waiting for
2340 any reply from clients. Clients drops these locks when cancel
2341 notification from server is received without replying.
2344 Frequency : liblustre-only, when liblustre client dies or becomes busy
2346 Description: Doing ls on Linux clients can take a long time with active
2348 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2349 granting such locks to it in the first place if possible. Locks
2350 are taken by proxy on the OST during the read or write and
2351 dropped immediately afterward. Add connect flags handling, do
2352 not grant locks to liblustre clients for glimpse ASTs.
2354 Severity : enhancement
2356 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2357 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2358 sequential reads and behaves badly with applications doing random
2359 reads. Improve it by reading ahead at least read region, and
2360 avoiding excessive large RPC for small reads.
2362 Severity : enhancement
2364 Description: Creating more than 1000 files for a single job may cause a load
2365 imbalance on the OSTs if there are also a large number of OSTs.
2366 Details : qos_prep_create() uses an OST index reseed value that is an
2367 even multiple of the number of available OSTs so that if the
2368 reseed happens in the middle of the object allocation it will
2369 still utilize the OSTs as uniformly as possible.
2374 Description: OST or MDS may oops in ping_evictor_main()
2375 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2376 but doesn't restart at beginning of obd_exports_timed list
2379 Severity : enhancement
2381 Description: improve by-nid export eviction on the MDS and OST
2382 Details : allow multiple exports with the same NID to be evicted at one
2383 time without re-searching the exports list.
2386 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2388 Description: MDS may oops in groups_free()
2389 Details : in rare race conditions a newly allocated group_info struct is
2390 freed again, and this can be NULL. The 2.4 compatibility code
2391 for groups_free() checked for a NULL pointer, but 2.6 did not.
2394 Frequency : common for liblustre clients doing little filesystem IO
2395 Bugzilla : 9352, 7313
2396 Description: server may evict liblustre clients accessing contended locks
2397 Details : if a client is granted a lock or receives a completion AST
2398 with a blocking AST already set it would not reply to the AST
2399 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2402 Frequency : lfs setstripe, only systems with more than 160 OSTs
2404 Description: unable to set striping with a starting offset beyond OST 160
2405 Details : llapi_create_file() incorrectly limited the starting stripe
2406 index to the maximum single-file stripe count.
2409 Frequency : LDAP users only
2411 Description: lconf did not handle in-kernel recovery with LDAP properly
2412 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2414 Severity : enhancement
2416 Description: bind OST threads to NUMA nodes to improve performance
2417 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2418 node and do their allocations there to localize memory access
2420 Severity : enhancement
2422 Description: llmount can determine client NID directly from Myrinet (GM)
2423 Details : the client NID code from gmnalnid was moved directly into
2424 llmount, removing the need to use this or specifying the
2425 client NID explicitly when mounting GM clients with zeroconf
2428 Frequency : if client is started with down MDS
2430 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2431 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2432 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2433 ptlrpc_import_delayed_req() because we didn't check if the
2434 request was interrupted, and we also didn't break out of the
2435 event loop if there was no timeout
2440 Description: data loss during non-page-aligned writes to a single file from
2441 both multiple nodes and multiple threads on one node at same time
2442 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2443 inconsistency led to false short-reads, that were cached and later
2444 used by ->prepare_write() to fill in partially written page,
2445 leading to data loss.
2448 Frequency : always, if lconf --abort_recovery used
2450 Description: lconf --abort_recovery fails with 'Operation not supported'
2451 Details : lconf was attempting to abort recovery on the MDT device and not
2454 Severity : enhancement
2456 Description: remove cleanup logs
2457 Details : replace lconf-generated cleanup logs with lustre internal
2458 cleanup routines. Eliminates the need for client-cleanup and
2461 Severity : enhancement
2463 Description: add support for EAs (user and system) on lustre filesystems
2464 Details : it is now possible to store extended attributes in the Lustre
2465 client filesystem, and with the user_xattr mount option it
2466 is possible to allow users to store EAs on their files also
2468 Severity : enhancement
2470 Description: Add possibility (config option) to show minimal available OST free
2472 Details : When compiled with --enable-mindf configure option, statfs(2)
2473 (and so, df) will return least minimal free space available from
2474 all OSTs as amount of free space on FS, instead of summary of
2475 free spaces of all OSTs.
2477 Severity : enhancement
2479 Description: do not expand extent locks acquired on OST-side
2480 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2481 by server: they are not cached anyway.
2484 Frequency : when mmap is used/binaries executed from Lustre
2486 Description: Unmmap pages before throwing them away from read cache.
2487 Details : llap_shrink cache now attempts to unmap pages before discarding
2488 them (if unmapping failed - do not discard). SLES9 kernel has
2489 extra checks that trigger if this unmapping is not done first.
2494 Description: lconf didn't resolve symlinks before checking to see whether a
2495 given mountpoint was already in use
2498 Frequency : when migrating failover services
2499 Bugzilla : 6395, 9514
2500 Description: When migrating a subset of services from a node (e.g. failback
2501 from a failover service node) the remaining services would
2502 time out and evict clients.
2503 Details : lconf --force (implied by --failover) sets the global obd_timeout
2504 to 5 seconds in order to quickly disconnect, but this caused
2505 other RPCs to time out too quickly. Do not change the global
2506 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2508 Severity : enhancement
2509 Frequency : if MDS is started with down OST
2510 Bugzilla : 9439,5706
2511 Description: Allow startup/shutdown of an MDS without depending on the
2512 availability of the OSTs.
2513 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2514 Add appropriate locking and lov-osc refcounts for safe
2515 cleaning. Add osc abort_inflight calls in case the
2516 synchronize never started.
2519 Frequency : occasional (Cray XT3 only)
2521 Description: root not authorized to access files in CRAY_PORTALS environment
2522 Details : The client process capabilities were not honoured on the MDS in
2523 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2524 been accessed by an authorized user then root was able to access
2525 the file on the local client also. The root user capabilities
2526 are now allowed on the MDS, as this environment has secure UID.
2529 Frequency : occasional
2531 Description: ldiskfs "too long searching" message happens too often
2532 Details : A debugging message (otherwise harmless) prints too often on
2533 the OST console. This has been reduced to only happen when
2534 there are fragmentation problems on the filesystem.
2539 Description: Division by zero in statfs when all OSCs are inactive
2540 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2541 lov_statfs divides by value returned by lov_get_stripecnt().
2545 Bugzilla : 9489, 3273
2546 Description: First write from each client to each OST was only 4kB in size,
2547 to initialize client writeback cache, which caused sub-optimal
2548 RPCs and poor layout on disk for the first writen file.
2549 Details : Clients now request an initial cache grant at (re)connect time
2550 and so that they can start streaming writes to the cache right
2551 away and always do full-sized RPCs if there is enough data.
2552 If the OST is rebooted the client also re-establishes its grant
2553 so that client cached writes will be honoured under the grant.
2558 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2559 Details : Now I/O RPCs go to different portal number and (presumably) fast
2560 lock requests (and glimses) and other RPCs get their own service
2561 threads pool that should be able to service those RPCs
2564 Severity : enhancement
2566 Description: Ability to exchange lustre version between client and servers and
2567 issue warnings at client side if client is too old. Also for
2568 liblustre clients there is ability to refuse connection of too old
2570 Details : New 'version' field is added to connect data structure that is
2571 filled with version info. That info is later checked by server and
2575 Frequency : rare, liblustre only.
2576 Bugzilla : 9296, 9581
2577 Description: Two simultaneous writes from liblustre at offset within same page
2578 might proceed at the same time overwriting eachother with stale
2580 Details : I/O lock withing llu_file_prwv was released too early, before data
2581 actually was hitting the wire. Extended lock-holding time until
2582 server acknowledges receiving data.
2585 Frequency : extremely rare. Never observed in practice.
2587 Description: avoid generating lustre_handle cookie of 0.
2588 Details : class_handle_hash() generates handle cookies by incrementing
2589 global counter, and can hit 0 occasionaly (this is unlikely, but
2590 not impossible, because initial value of cookie counter is
2591 selected randonly). Value of 0 is used as a sentinel meaning
2592 "unassigned handle" --- avoid it. Also coalesce two critical
2593 sections in this function into one.
2595 Severity : enhancement
2597 Description: allow liblustre clients to delegate truncate locking to OST
2598 Details : To avoid overhead of locking, liblustre client instructs OST to
2599 take extent lock in ost_punch() on client's behalf. New connection
2600 flag is added to handle backward compatibility.
2602 Severity : enhancement
2603 Bugzilla : 4928, 7341, 9758
2604 Description: allow number of OST service threads to be specified
2605 Details : a module parameter allows the number of OST service threads
2606 to be specified via "options ost ost_num_threads={N}" in the
2607 OSS's /etc/modules.conf or /etc/modprobe.conf.
2611 Bugzilla : 6146, 9635, 9895
2612 Description: servers crash with bad pointer in target_handle_connect()
2613 Details : In rare cases when a client is reconnecting it was possible that
2614 the connection request was the last reference for that export.
2615 We would temporarily drop the export reference and get a new
2616 one, but this may have been the last reference and the export
2617 was just destroyed. Get new reference before dropping old one.
2619 Severity : enhancement
2620 Frequency : if client is started with failover MDS
2622 Description: Allow multiple MDS hostnames in the mount command
2623 Details : Try to read the configuration from all specified MDS
2624 hostnames during a client mount in case the "primary"
2627 Severity : enhancement
2629 Description: Stop sending data to evicted clients as soon as possible.
2630 Details : Check if the client we are about to send or are sending data to
2631 was evicted already. (Check is done every second of waiting,
2632 for which l_wait_event interface was extended to allow checking
2633 of exit condition at specified intervals).
2636 Frequency : rare, normally only when NFS exporting is done from client
2638 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2640 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2643 Frequency : always, if srand() is called before liblustre initialization
2645 Description: Liblustre uses system PRNG disturbing its usage by user application
2646 Details : Introduce internal to lustre fast and high-quality PRNG for
2647 lustre usage and make liblustre and some other places in generic
2648 lustre code to use it.
2650 Severity : enhancement
2651 Bugzilla : 9477, 9557, 9870
2652 Description: Verify that the MDS configuration logs are updated when xml is
2653 Details : Check if the .xml configuration logs are newer than the config
2654 logs stored on the MDS and report an error if this is the case.
2655 Request --write-conf, or allow starting with --old_conf.
2657 Severity : enhancement
2659 Description: Handle symlinks in the path when checking if Lustre is mounted.
2660 Details : Resolve intermediate symlinks when checking if a client has
2661 mounted a filesystem to avoid duplicate client mounts.
2666 Description: lconf can hit an error exception but still return success.
2667 Details : The lconf command catches the Command error exception at the top
2668 level script context and will exit with the associated exit
2669 status, but doesn't ensure that this exit status is non-zero.
2674 Description: failure of ptlrpc thread startup can cause oops
2675 Details : Starting a ptlrpc service thread can fail if there are a large
2676 number of threads or the server memory is very fragmented.
2677 Handle this without oopsing.
2680 Frequency : always, only if liblustre and non-default acceptor port was used
2682 Description: liblustre cannot connect to servers with non-default acceptor port
2683 Details : tcpnal_set_default_params() was not called and was therefore
2684 ignoring the environment varaible TCPNAL_PORT, as well as other
2685 TCPNAL_ environment variables
2690 Description: two objects could be created on the same OST for a single file
2691 Details : If an OST is down, in some cases it was possible to create two
2692 objects on a single OST for a single file. No problems other
2693 than potential performance impact and spurious error messages.
2697 Bugzilla : 5681, 9562
2698 Description: Client may oops in ll_unhash_aliases
2699 Details : Client dcache may become inconsistent in race condition.
2700 In some cases "getcwd" can fail if the current directory is
2706 Description: Inode refcounting problems in NFS export code
2707 Details : link_raw functions used to call d_instantiate without obtaining
2708 extra inode reference first.
2712 Bugzilla : 9942, 9903
2713 Description: Referencing freed requests leading to crash, memleaks with NFS.
2714 Details : We used to require that call to ll_revalidate_it was always
2715 followed by ll_lookup_it. Also with revalidate_special() it is
2716 possible to call ll_revalidate_it() twice for the same dentry
2717 even if first occurence returned success. This fix changes semantic
2718 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2719 reference on a request referred from the intent.
2720 ll_intent_release() then releases such a request.
2723 Frequency : rare, normally benchmark loads only
2725 Description: unlinked inodes were kept in memory on the client
2726 Details : If a client is repeatedly creating and unlinking files it
2727 can accumulate a lot of stale inodes in the inode slab cache.
2728 If there is no other client load running this can cause the
2729 client node to run out of memory. Instead flush old inodes
2730 from client cache that have the same inode number as a new inode.
2733 Frequency : SLES9 2.6.5 kernel and long filenames only
2734 Bugzilla : 9969, 10379
2735 Description: utime reports stale NFS file handle
2736 Details : SLES9 uses out-of-dentry names in some cases, which confused
2737 the lustre dentry revalidation. Change it to always use the
2741 Frequency : rare, unless heavy write-truncate concurrency is continuous
2742 Bugzilla : 4180, 6984, 7171, 9963, 9331
2743 Description: OST becomes very slow and/or deadlocked during object unlink
2744 Details : filter_destroy() was holding onto the parent directory lock
2745 while truncating+unlinking objects. For very large objects this
2746 may block other threads for a long time and slow overall OST
2747 responsiveness. It may also be possible to get a lock ordering
2748 deadlock in this case, or run out of journal credits because of
2749 the combined truncate+unlink. Solution is to do object truncate
2750 first in one transaction without parent lock, and then do the
2751 final unlink in a new transaction with the parent lock. This
2752 reduces the lock hold time dramatically.
2755 Frequency : rare, 2.4 kernels only
2757 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2758 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2759 before unmounting their filesystem in order to avoid deadlock.
2760 The kernel_locked() function in 2.4 kernels only checks whether
2761 the kernel lock is held, not whether it is this process that is
2762 holding it as 2.6 kernels do.
2767 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2768 Details : The client ptlrpc code may be trying to reconnect to a down
2769 server before a previous connection attempt has timed out.
2770 Increase the reconnect interval to be longer than the connection
2771 timeout interval to avoid sending duplicate connections to
2775 Frequency : echo_client brw_test command
2777 Description: fix echo_client to work with OST preallocated code
2778 Details : OST preallocation code (5137) didn't take echo_client IO path
2779 into account: echo_client calls filter methods outside of any
2780 OST thread and, hence, there is no per-thread preallocated
2781 pages and buffers to use. Solution: hijack pga pages for IO. As
2782 a byproduct, this avoids unnecessary data copying.
2786 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2787 Description: Client can oops in mdc_commit_close() after open replay
2788 Details : It was possible for the MDS to return an open request with no
2789 transaction number in mds_finish_transno() if the client was
2790 evicted, but without actually returning an error. Clients
2791 would later try to replay that open and may trip an assertion
2792 Simplify the client close codepath, and always return an error
2793 from the MDS in case the open is not successful.
2796 Frequency : rare, 2.6 OSTs only
2798 Description: OST may deadlock under high load on fragmented files
2799 Details : If there was a heavy load and highly-fragmented OST filesystems
2800 it was possible to have all the OST threads deadlock waiting on
2801 allocation of biovecs, because the biovecs were not released
2802 until the entire RPC IO was completed. Instead, release biovecs
2803 as soon as they are complete to ensure forward IO progress.
2805 Severity : enhancement
2807 Description: Support for specifying external journal device at mount
2808 Details : If an OST or MDS device is formatted with an external journal
2809 device, this device major/minor is stored in the ext3 superblock
2810 and may not be valid for failover. Allow detecting and
2811 specifying the external journal at mount time.
2816 Description: Mounting an MDS with pending unlinked files may cause oops
2817 Details : target_finish_recovery() calls mds_postrecov() which returned
2818 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2819 considers this an error and immediately begins cleaning up the
2820 lov, just after starting the mds_lov process
2822 Severity : enhancement
2824 Description: Implement 'lfs df' to report actual free space on per-OST basis
2825 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2826 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2827 usage of objects; '-h' to report in human readable format.
2829 ------------------------------------------------------------------------------
2831 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2838 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2839 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2840 at startup during orphan unlinking. Instead, skip buddy update
2841 before setup, as it will be regenerated after recovery is complete.
2846 Description: If an OST is inactive, its locks might reference stale inodes.
2847 Details : lov_change_cbdata() must iterate over all namespaces, even if
2848 they are inactive to clear inode references from the lock.
2850 Severity : enhancement
2851 Frequency : occasional, if non-standard max_dirty_mb used
2853 Description: Client will block write RPCs if not enough grant
2854 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2855 then the client will block writes while waiting for another RPC
2856 to complete instead of consuming its dirty limit. With change
2857 we get improved performance when max_dirty_mb is small.
2859 Severity : enhancement
2860 Bugzilla : 3389, 6253
2861 Description: Add support for supplementary groups on the MDS.
2862 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2863 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2864 lookups for user supplementary groups into a cache.
2868 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2869 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2870 for liblustre O_EXCL is enforced by MDS.
2874 Description: readdir never returns NULL in liblustre.
2875 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2876 dentry in struct dirent.
2880 Frequency : liblustre only, depends on application IO pattern
2881 Description: liblustre clients evicted if not contacting servers
2882 Details : Don't put liblustre clients into the ping_evictor list, so
2883 they will not be evicted by the pinger ever.
2885 Severity : enhancement
2887 Description: Add ability to evict clients by NID from MDS.
2888 Details : By echoing "nid:$NID" string into
2889 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2890 $NID would be instantly evicted from this MDS and from all active
2891 OSTs connected to it.
2895 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2896 Details : lookup_it_finish() used to query file size from OSTs that was not
2901 Description: service threads change working directory to that of init
2902 Details : Starting lustre service threads may pin the working directory
2903 of the parent thread, making that filesystem busy. Threads
2904 now change to the working directory of init to avoid this.
2908 Frequency : during shutdown only
2909 Description: shutdown with a failed MDS or OST can cause unmount to hang
2910 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2915 Frequency : 2.6 only
2916 Description: chmod/chown may include an extra supplementary group
2917 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2918 supplementary group and if none is specified this is used.
2921 Bugzilla : 5479 (6816)
2923 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2924 Details : If lookup is in progress on a file that is unlinked we might try
2925 to revalidate the inode and fail in revalidate after lookup is
2926 complete and ll_file_open() enqueues the open again but
2927 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2930 Frequency : always, if lconf --abort_recovery used
2932 Description: lconf --abort_recovery fails with 'Operation not supported'
2933 Details : lconf was attempting to abort recovery on the MDT device and not
2936 ------------------------------------------------------------------------------
2938 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2943 Frequency : rare (only unsupported configurations with a node running as an
2945 Bugzilla : 6514, 5137
2946 Description: Mounting a Lustre file system on a node running as an OST could
2948 Details : OSTs now preallocates memory needed to write out data at
2949 startup, instead of when needed, to avoid having to
2950 allocate memory in possibly low memory situations.
2951 Specifically, if the file system is mounted on on OST,
2952 memory pressure could force it to try to write out data,
2953 which it needed to allocate memory to do. Due to the low
2954 memory, it would be unable to do so and the node would
2955 become unresponsive.
2957 Severity : enhancement
2959 Description: Addition of lconf --service command line option
2960 Details : lconf now accepts a '--service <arg>' option, which is
2961 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2963 Severity : enhancement
2965 Description: Failover mode is now the default for OSTs.
2966 Details : By default, OSTs will now run in failover mode. To return to
2967 the old behaviour, add '--failout' to the lmc line for OSTs.
2969 Severity : enhancement
2971 Description: Health checks are now provided for MDS and OSTs
2972 Details : Additional detailed health check information on MSD and OSTs
2973 is now provided through the procfs health_check value.
2976 Frequency : occasional, depends on IO load
2978 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2979 after numerous create/delete cycles
2980 Details : The ext3 inode allocation policy would not allocate new inodes
2981 very well on the OSTs because there are no new directories
2982 being created. Instead we look for groups with free space if
2983 the parent directories are nearly full.
2988 Description: Network or server problems during mount may cause partially
2989 mounted clients instead of returning an error.
2990 Details : The config llog parsing code may overwrite the error return
2991 code during mount error handling, returning success instead
2997 Description: MDS can fail to allocate large reply buffers
2998 Details : After long uptimes the MDS can fail to allocate large reply
2999 buffers (e.g. zconf client mount config records) due to memory
3000 fragmentation or consumption by the buffer cache. Preallocate
3001 some large reply buffers so that these replies can be sent even
3002 under memory pressure.
3006 Frequency : rare (liblustre)
3007 Description: fsx running with liblustre complained that using truncate() to
3008 extend the file doesn't work. This patch corrects that issue.
3009 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3010 ATTR_SIZE and lsm use in llu_setattr_raw.
3014 Frequency : rare, only 2.6 kernels
3015 Description: Unusual file access patterns on the MDS may result in inode
3016 data being lost in very rare circumstances.
3017 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3018 code in the 2.6 kernel caused the inode and/or EA data not to
3019 be read from disk, causing single-file corruption.
3023 Frequency : rare, only 2.6 filesystems using extents
3024 Description: Heavy concurrent write and delete load may cause data corruption.
3025 Details : It was possible under high-load situations to have an extent
3026 metadata block in the block device cache from a just-unlinked
3027 file overwrite a newly-allocated data block. We now unmap any
3028 metadata buffers that alias just-allocated data blocks.
3032 Frequency : filesystems with default stripe_count larger than 77
3033 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3034 Details : lconf specifies an inode size of 4096 bytes when the default
3035 stripe_count is larger than 77. This conflicts with the default
3036 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3037 this case to avoid pinning too much memory for large EAs.
3039 ------------------------------------------------------------------------------
3041 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3046 Frequency : rare (extremely heavy IO load with hundreds of clients)
3048 Description: Client is evicted, gets IO error writing to file
3049 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3050 caused clients to be evicted for AST timeouts. The fixes in
3051 bug 5192 mean we no longer need to have such short AST timeouts
3052 so ldlm_timeout has been increased.
3055 Frequency : occasional during --force or --failover shutdown under load
3056 Bugzilla : 5949, 4834
3057 Description: Server oops/LBUG if stopped with --force or --failover under load
3058 Details : a collection of import/export refcount and cleanup ordering
3059 issues fixed for safer force cleanup
3062 Frequency : only filesystems larger than 120 OSTs
3063 Bugzilla : 5990, 6223
3064 Description: lfs getstripe would oops on a very large filesystem
3065 Details : lov_getconfig used kfree on vmalloc'd memory
3068 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3069 Bugzilla : 6242, 6243
3070 Description: reading from files that had been truncated to a non-zero size
3071 but never opened returned no data
3072 Details : ll_file_read() reads zeros from no-object files to EOF
3077 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3078 Details : There's little harm in aborting MDS/OSS recovery and letting it
3079 try again, so I removed the LASSERT and return an error instead.
3081 Severity : enhancement
3083 Description: New debugging infrastructure for tracking down data corruption
3084 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3085 (b) cover more of the client-side code path, and (c) try to narrow
3086 down where problems occurred
3090 Bugzilla : 3819, 4364, 4397, 6313
3091 Description: Racing close and eviction MDS could cause assertion in mds_close
3092 Details : It was possible to get multiple mfd references during close and
3093 client eviction, leading to one thread referencing a freed mfd.
3095 Severity: : enhancement
3096 Bugzilla : 3262, 6359
3097 Description: Attempts to reconnect to servers are now more aggressive.
3098 Details : This builds on the enhanced upcall-less recovery that was added
3099 in 1.4.2. When trying to reconnect to servers, clients will
3100 now try each server in the failover group every 10 seconds. By
3101 default, clients would previously try one server every 25 seconds.
3106 Description: After recovery, certain operations trigger a failed
3107 assertion on a client.
3108 Details : Failing over an mds, using lconf -d --failover, while a
3109 client was doing a readdir() call would cause the client to
3110 LBUG after recovery completed and the readdir() was resent.
3112 Severity : enhancement
3114 Description: Default groups are now added by lconf
3115 Details : You can now run lconf --group <servicename> without having to
3116 manually add groups with lmc.
3119 Frequency : occasional
3121 Description: Nodes with an elan id of 0 trigger a failed assertion
3124 Frequency : always when accessing e.g. tty/console device nodes
3126 Description: tty and some other devices nodes cannot be used on lustre
3127 Details : file's private_data field is used by device data and lustre
3128 values in there got lost. New field was added to struct file to
3129 store fs-specific private data.
3132 Frequency : when exporting Lustre via NFS
3134 Description: NFSD failed occasionally when looking up a path component
3135 Details : NFSD is looking up ".." which was broken in ext3 directories
3136 that had grown large enough to become hashed.
3139 Frequency : Clusters with multiple interfaces not on the same subnet
3141 Description: Nodes will repeatedly try to reconnect to an interface which it
3142 cannot reach and report an error to the log.
3143 Details : Extra peer list entries will be created by lconf with some peers
3144 unreachable. lconf now validates the peer before adding it.
3147 Frequency : Only if a default stripe is set on the filesystem root.
3149 Description: Setting a default stripe on the filesystem root prevented the
3150 filesystem from being remounted.
3151 Details : The client was sending extra request flags in the root getattr
3152 request and did not allocate a reply buffer for the dir EA.
3155 Frequency : occasional, higher if lots of files are accessed by one client
3156 Bugzilla : 6159, 6097
3157 Description: Client trips assertion regarding lsm mismatch/magic
3158 Details : While revalidating inodes the VFS looks up inodes with ifind()
3159 and in rare cases can find an inode that is being freed.
3160 The ll_test_inode() code will free the lsm during ifind()
3161 when it finds an existing inode and then the VFS later attaches
3162 this free lsm to a new inode.
3166 Bugzilla : 6422, 7030
3167 Description: MDS deadlock between mkdir and client eviction
3168 Details : Creating a new file via mkdir or mknod (starting a transaction
3169 and getting the ns lock) can deadlock with client eviction
3170 (gets ns lock and trying to finish a synchronous transaction).
3173 Frequency : occasional
3174 Description: While starting a server, the fsfilt_ext3 module could not be
3176 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3177 kernels. Previously, lconf would still use the ext3 name
3178 when trying to load modules. Now, it will correctly use
3179 ext3 on 2.4 and ldiskfs on 2.6.
3181 Severity : enhancement
3182 Description: The default stripe count has been changed to 1
3183 Details : The interpretation of the default stripe count (0, to lfs
3184 or lmc) has been changed to mean striping across a single
3185 OST, rather than all available. For general usage we have
3186 found a stripe count of 1 or 2 works best.
3188 Severity : enhancement
3189 Description: Add support for compiling against Cray portals.
3190 Details : Conditional compiling for some areas that are different
3194 Frequency : occasional
3195 Bugzilla : 6409, 6834
3196 Description: Creating files with an explicit stripe count may lead to
3197 a failed assertion on the MDS
3198 Details : If some OSTs are full or unavailable, creating files may
3199 trigger a failed assertion on the MDS. Now, Lustre will
3200 try to use other servers or return an error to the
3204 Frequency : occasional
3206 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3207 caused lock timeout during readahead (since 1.4.2).
3208 Details : Processes doing readahead might match a lock that hasn't been
3209 granted yet if there are overlapping and conflicting lock
3210 requests. The readahead process waits on ungranted lock
3211 (original lock is CBPENDING), while OST waits for that process
3212 to cancel CBPENDING read lock and eventually evicts client.
3214 Severity : enhancement
3216 Description: Initial enabling of flock support for clients
3217 Details : Implements fcntl advisory locking and file status functions.
3218 This feature is provided as an optional mount flag (default
3219 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3220 locking are implemented yet, and those that are are not guaranteed
3221 to be completely correct in production environments.
3222 mount -t lustre -o [flock|noflock] ...
3225 Frequency : occasional
3227 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3228 assertion in the ext3 JBD (journaling) layer.
3229 Details : The b_committed_data struct is protected by the big kernel lock
3230 in 2.4 kernels, serializing journal_commit_transaction() and
3231 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3232 access to this struct. In 2.6 kernels there is finer grained
3233 locking to improve SMP performance of the JBD layer.
3237 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3242 Description: merge and cleanup kernel patches.
3245 -----------------------------------------------------------------------------
3247 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3249 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3250 versions of Lustre. Please update all servers and clients to
3251 version 1.4.2 or later at the same time. You must also run
3252 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3253 to update the configuration logs.
3255 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3256 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3257 - swab llog records of type '0' so we get proper header size/idx (5861)
3258 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3259 - fix rename of one directory over another leaking an inode (5953)
3260 - avoid SetPageDirty on 2.6 (5981)
3261 - don't re-add just-being-destroyed locks to the waiting list (5653)
3262 - when creating new directories, inherit the parent's custom
3263 striping settings if present parent (3048)
3264 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3265 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3266 - don't hold client locks on temporary worklist from l_lru (5666)
3267 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3268 - automatically evict dead clients (5921)
3269 - Update file size properly in create+truncate+fstat case (6196)
3270 - Do not unhash mountpoint dentries, do not allow removal of
3272 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3273 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3274 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3275 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3276 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3277 - Correctly handle reads of files with no objects (6243)
3278 - lctl recover will also mark a device active if deactivate used (5933)
3280 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3281 - allow --write-conf on an MDS with different nettype than client (5619)
3282 - don't write config llogs to MDS for mounts not from that MDS (5617)
3283 - lconf should create multiple TCP connections from a client (5201)
3284 - init scripts are now turned off by default; run chkconfig --on
3285 lustre and chkconfig --on lustrefs to use them
3286 - upcalls are no longer needed for clients to recover to failover
3288 - add --abort-recovery option to lconf to abort recovery on device
3290 - add support for an arbitrary number of OSTs (3026)
3291 - Quota support protocol changes.
3292 - forward compatibility changes to wire structs (6007)
3293 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3294 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3295 - improved "lustre status" script
3296 - initialize blocksize for non-regular files (6062)
3297 - added --disable-server and --disable-client configure options (5782)
3298 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3299 - Vanilla 2.4.29 support
3300 - increase maximum number of obd devices to 520 (6242)
3301 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3302 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3303 SLES 9 2.6.5 kernels
3304 - limit stripes per file to 160 (the maximum EA size) (6093)
3306 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3309 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3310 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3311 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3312 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3313 - small ext3 extents cleanups and fixes (5733)
3314 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3315 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3316 - avoid needless client->OST connect, fix handle mismatch (5317)
3317 - fix DLM error path that led to out-of-sync client, long delays (5779)
3318 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3319 - fix several locking issues related to i_size (5492,5624,5654,5672)
3320 - don't move pending lock onto export if it is already evicted (5683)
3321 - fix kernel oops when creating .foo in unlinked directory (5548)
3322 - fix deadlock in obdfilter statistics vs. object create (5811)
3323 - use time_{before,after} to avoid timer jiffies wrap (5882)
3324 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3325 - Do not leak request if server was not able to process it (5154)
3326 - If mds_open unable to find parent dir, make that negative lookup(5154)
3327 - don't create new directories with extent-mapping (5909, 5936)
3329 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3330 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3331 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3332 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3333 - support for RHEL4 kernel on i686 (5773)
3334 - provide error messages when incompatible logs are encountered (5898)
3336 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3337 * version 1.4.0.10 (1.4.1 release candidate 1)
3339 - don't keep a lock reference when lock is not granted (4238)
3340 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3341 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3342 - drop import inflight refcount on signal_completed_replay error (5255)
3343 - unlock page after async write error during send (3677)
3344 - handle missing objects in filter_preprw_read properly (5265)
3345 - no transno return for symlink open, don't save no-trasno open (3440)
3346 - don't try to complete elan receive that already failed (4012)
3347 - free RPC server reply state on error (5406)
3348 - clean up thread from ptlrpc_start_thread() on error (5160)
3349 - readahead could read extra page into cache that wasn't ejected (5388)
3350 - prevent races in class_attach/setup/cleanup/detach (5260)
3351 - don't dereference de->d_inode after l_dput of de (5458)
3352 - use "int" for stripe value returned from lock_to_stripe (5544)
3353 - mballoc allocation and error-checking fixes in 2.6 (5504)
3354 - block device patches to fix I/O request sizes in 2.6 (5482)
3355 - look up hostnames for IB nals (5602)
3356 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3357 - don't start multiple acceptors for the same port (5277)
3358 - fix incorrect LASSERT in mds_getattr_name (5635)
3359 - export a proc file for general "ping" checking (5628)
3360 - fix "lfs check" to not block when the MDS is down (5628)
3362 - service request history (4965)
3363 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3364 - create an "evict_client" /proc entry on OSTs, like the MDS has
3365 - fix mount usage message, return errors per mount(8) (5168)
3366 - change grep [] to grep "[]" in tests so they work in more UMLs
3367 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3368 - remove ancient LOV_MAGIC_V0 EA support (5047)
3369 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3370 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3371 - allow readahead allocations to fail when low on memory (5383)
3372 - mmap locking landed again, after considerable improvement (2828)
3373 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3375 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3378 - send OST transaction number in read/write reply to free req (4966)
3379 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3380 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3381 - the watchdog thread now runs as interruptible (5246)
3382 - flock/lockf fixes (but it's still disabled, pending 5135)
3383 - don't use EXT3 constants in llite code (5094)
3384 - memory shortage at startup could cause assertion (5176)
3386 - reorganization of lov code
3387 - single portals codebase
3389 - add extents/mballoc support (5025)
3390 - direct I/O reads in the obdfilter (4048)
3391 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3393 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3396 - send OST transaction number in read/write reply to free req (4966)
3397 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3398 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3399 - the watchdog thread now runs as interruptible (5246)
3400 - handle missing objects in filter_preprw_read properly (5265)
3401 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3402 - drop import inflight refcount on signal_completed_replay error (5255)
3403 - unlock page after async write error during send (3677)
3404 - return original error code on reconstructed replies (3761)
3405 - no transno return for symlink open, don't save no-trasno open (3440)
3407 - add pid to ldlm debugging output (4922)
3408 - bump the watchdog timeouts -- we can't handle 30sec yet
3409 - extra debugging for orphan dentry/inode bug (5259)
3411 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3414 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3415 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3416 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3417 - fix NULL dereference in /proc/sys/portals/routes (4827)
3418 - allow failed mdc_close() operations to be interrupted (4561)
3419 - stop precreate on OST before MDS would time out on it (4778)
3420 - don't send partial-page writes before EOF from client (4410)
3421 - discard client grant for sub-page writes on large-page clients (4520)
3422 - don't free dentries not owned by NFS code, check generation (4806)
3423 - fix lsm leak if mds_create_objects() fails (4801)
3424 - limit debug_daemon file size, always print CERROR messages (4789)
3425 - use transno after validating reply (3892)
3426 - process timed out requests if import state changes (3754)
3427 - update mtime on OST during writes, return in glimpse (4829)
3428 - add mkfsoptions to LDAP (4679)
3429 - use ->max_readahead method instead of zapping global ra (5039)
3430 - don't interrupt __l_wait_event() during strace
3432 - add software watchdogs to catch hung threads quickly (4941)
3433 - make lustrefs init script start after nfs is mounted
3434 - fix CWARN/ERROR duplication (4930)
3435 - return async write errors to application if possible (2248)
3436 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3437 - print NAL number in %x format (4645)
3438 - update barely-supported suse-2.4.21-171 series (4842)
3439 - support for sles 9 %post scripts
3440 - support for building 2.6 kernel-source packages
3441 - support for sles km_* packages
3443 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3446 - ignore -ENOENT errors in osc_destroy (3639)
3447 - notify osc create thread that OSC is being cleaned up (4600)
3448 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3449 - reconstruct ost_handle() like mds_handle() (4657)
3450 - create a new thread to do import eviction to avoid deadlock (3969)
3451 - let lconf resolve symlinked-to devices (4629)
3452 - don't unlink "objects" from directory with default EA (4554)
3453 - hold socknal file ref over connect in case target is down (4394)
3454 - allow more than 32000 subdirectories in a single directory (3244)
3455 - fix blocks count for O_DIRECT writes (3751)
3456 - OST returns ENOSPC from object create when no space left (4539)
3457 - don't send truncate RPC if file size isn't changing (4410)
3458 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3459 - bind to privileged port in socknal and tcpnal (3689)
3461 - rate limit CERROR/CWARN console message to avoid overload (4519)
3462 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3463 - basic mmap support (3918)
3464 - kernel patch series update from b1_4 (4711)
3466 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3469 - avoid crash during MDS cleanup with OST shut down (2775)
3470 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3471 - don't use bad inodes on the MDS (3744)
3472 - dynamic object preallocation to improve recovery speed (4236)
3473 - don't hold spinlock over lock dumping or change debug flags (4401)
3474 - don't zero obd_dev when it is force cleaned (3651)
3475 - print grants to console if they go negative (4431)
3476 - "lctl deactivate" will stop automatic recovery attempts (3406)
3477 - look for existing locks in ldlm_handle_enqueue() (3764)
3478 - don't resolve lock handle twice in recovery avoiding race (4401)
3479 - revalidate should check working dir is a directory (4134)
3481 - don't always mark "slow" obdfilter messages as errors (4418)
3483 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3486 - don't close LustreDB during write_conf until it is done (3860)
3487 - fix typo in lconf for_each_profile (3821)
3488 - allow dumping logs from multiple threads at one time (3820)
3489 - don't allow multiple threads in OSC recovery (3812)
3490 - fix debug_size parameters (3864)
3491 - fix mds_postrecov to initialize import for llog ctxt (3121)
3492 - replace config semaphore with spinlock (3306)
3493 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3494 - don't allow enqueue to complete on a destroyed export (3822)
3495 - down write_lock before checking llog header bitmap (3825)
3496 - recover from lock replay timeout (3764)
3497 - up llog sem before sending rpc (3652)
3498 - reduce ns lock hold times when setting kms (3267)
3499 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3500 - fix NULL deref and obd_dev leak on setup error (3312)
3501 - replace some LBUG about llog ops with error handling (3841)
3502 - don't match INVALID dentries from d_lookup and spin (3784)
3503 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3504 - fix invalid assertion in ptlrpc_set_wait (3880)
3506 - add libwrap support for the TCP acceptor (3996)
3507 - add /proc/sys/portals/routes for non-root route listing (3994)
3508 - allow setting MDS UUID in .xml (2580)
3509 - print the stack of a process that LBUGs (4228)
3511 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3514 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3515 - make sure to unset replay flag from failed open requests (3440)
3516 - if default stripe count is 0, use OST count for inode size (3636)
3517 - update parent mtime/ctime on client for create/unlink (2611)
3518 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3519 - free recovery state on server during a forced cleanup (3571)
3520 - unregister_reply for resent reqs (3063)
3521 - loop back devices mounting and status check on 2.6 (3563)
3522 - fix resource-creation race that can provoke i_size == 0 (3513)
3523 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3524 - more debugging for page-accounting assertion (3746)
3525 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3526 - avoid deadlock after precreation failure (3758)
3527 - fix race and lock order deadlock in orphan handling (3450, 3750)
3528 - add validity checks when grabbing inodes from l_ast_data (3599)
3530 - add /proc/.../recovery_status to obdfilter (3428)
3531 - lightweight CDEBUG infrastructure, debug daemon (3668)
3532 - change default OSC RPC parameters to be better on small clusters
3533 - turn off OST read cache for files smaller than 32MB
3534 - install man pages and include them in rpms (3100)
3535 - add new init script for (un)mounting lustre filesystems (2593)
3536 - run chkconfig in %post for init scripts (3701)
3537 - drop scimac NAL (unmaintained)
3539 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3542 - clean kiobufs before and after use (3485)
3543 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3544 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3545 - change init script to fail more clearly if not run as root (1528)
3546 - allow clients to reconnect during replay (1742)
3547 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3548 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3549 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3550 - fix cond_resched (3554)
3551 - only evict unfinished clients after recovery (3515)
3552 - allow bulk resend, prevent data loss (3570)
3553 - dynamic ptlrpc request buffer allocation (2102)
3554 - don't allow unlinking open directory if it isn't empty (2904)
3555 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3556 - remove extraneous obd dereference causing LASSERT failure (3334)
3557 - don't use get_cycles() when creating temp. files on the mds (3156)
3558 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3559 - handle EEXIST for set-stripe, set proper directory name (3336)
3561 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3562 - fix ksocknal_fmb_callback() error messages (2918)
3564 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3567 - don't copy lvb into (possibly NULL) reply on error (2983)
3568 - don't deref dentry after dput, don't free lvb on error (2922)
3569 - use the kms to determine writeback rpc length (2947)
3570 - increment oti_logcookies when osc is inactive (2948)
3571 - update client's i_blocks count via lvb messages (2543)
3572 - handle intent open/close of special files properly (1557)
3573 - mount MDS with errors=remount-ro, like obdfilter (2009)
3574 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3575 - don't use cancelling-locks' kms values (2947)
3576 - use highest lock extent for kms, not last one (2925)
3577 - don't dereference ERR_PTR() dentry in error handling path (3107)
3578 - fix thread race in portals_debug_dumplog() (3122)
3579 - create lprocfs device entries at setup instead of at attach (1519)
3580 - common AST error handler, don't evict client on completion race (3145)
3581 - zero nameidata in detach_mnt in 2.6 (3118)
3582 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3583 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3584 - initialize RPC timeout timer earlier for 2.6 (3219)
3585 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3586 - print nal/nid for unknown nid (3258)
3587 - additional checks for oscc recovery before doing precreate (3284)
3588 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3589 - don't crash in mdc_close for bad permissions on open (3285)
3590 - zero i_rdev for non-device files (3147)
3591 - clear page->private before handing to FS, better assertion (3119)
3592 - tune the read pipeline (3236)
3593 - fix incorrect decref of invalidated dentry (2350)
3594 - provide read-ahead stats and refine rpc in flight stats (3328)
3595 - don't hold journal transaction open across create RPC (3313)
3596 - update atime on MDS at close time (3265)
3597 - close LDAP connection when recovering to avoid server load (3315)
3598 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3599 - don't leak open file on MDS after open resend (3325)
3600 - serialize filter_precreate and filter_destroy_precreated (3329)
3601 - loop device shouldn't call sync_dev() for nul device (3092)
3602 - clear page cache after eviction (2766)
3603 - resynchronize MDS->OST in background (2824)
3604 - refuse to mount the same filesystem twice on same mountpoint (3394)
3605 - allow llmount to create routes for mounting behind routers (3320)
3606 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3607 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3608 - fix rare problem with rename on htree directories (3417)
3610 - allow default OST striping configuration per directory (1414)
3611 - fix compilation for qswnal for 2.6 kernels (3125)
3612 - increase maximum number of MDS request buffers for large systems
3613 - change liblustreapi to be useful for external progs like lfsck (3098)
3614 - increase local configuration timeout for slow disks (3353)
3615 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3617 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3620 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3621 - don't overwrite extent policy data in reply if lock was blocked (2901)
3622 - drop filter export grants atomically with removal from device (2663)
3623 - del obd_self_export from work_list in class_disconnect_exports (2908)
3624 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3625 - swab reply message in mdc_close, other PPC fixes (2464)
3626 - fix destroying of named logs (2325)
3627 - overwrite old logs when running lconf --write_conf (2264)
3628 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3629 - fix race in target_handle_connect (2898)
3630 - mds_reint_create() should take same inode create lock (2926)
3631 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3632 - don't close files for self_export to avoid uninitialized obd (2936)
3633 - allow MDS with the same name as client node (2939)
3634 - hold dentry reference for closed log files for unlink (2325)
3635 - reserve space for all logs during transactions (2059)
3636 - don't evict page beyond end of stripe extent (2925)
3637 - don't oops on a deleted current working directory (2399)
3638 - handle hard links to targets without a parent properly (2517)
3639 - don't dereference NULL lock when racing during eviction (2867)
3640 - don't grow lock extents when lots of conflicting locks (2919)
3642 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3645 - account for cache space usage on clients to avoid data loss (974)
3646 - lfsck support in lustre kernel code (2349)
3647 - reduce journal credits needed for BRW writes (2370)
3648 - orphan handling to avoid losing space on client/server crashes
3649 - ptlrpcd can be blocked, stopping ALL progress (2477)
3650 - use lock value blocks to assist in proper KMS, faster stat (1021)
3651 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3652 - recovery for initial connections (2355)
3653 - fixes for mds_cleanup_orphans (1934)
3654 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3655 - block all file creations until orphan recovery completes (1901)
3656 - client remove rq_connection from request struct (2423)
3657 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3658 - recovery timer race (2670)
3659 - mdc_close recovey bug (2532)
3660 - ptlrpc cleanup bug (2710)
3661 - mds timeout on local locks (2588)
3662 - namespace lock held during RPCs (2431)
3663 - handle interrupted sync write properly (2503)
3664 - don't try to handle a message that hasn't been replied to (2699)
3665 - client assert failure during cleanup after abort recovery (2701)
3666 - leak mdc device after failed mount (2712)
3667 - ptlrpc_check_set allows timedout requests to complete (2714)
3668 - wait for inflight reqs when ptlrpcd finishes (2710)
3669 - make sure unregistered services are removed from the srv_list
3670 - reset bulk XID's when resending them (caught by 1138 test)
3671 - unregister_bulk after timeout
3672 - fix lconf error (2694)
3673 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3674 - readahead locks pages, leaves pending causing memory pressure (2673)
3675 - increase OST request buffers to 4096 on large machines (2729)
3676 - fix up permission of existing directories in simple_mkdir (2661)
3677 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3678 - don't assign transno to errored transactions (2742)
3679 - don't delete objects on OST if given a bogus objid from MDS (2751)
3680 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3681 - if rq_no_resend, then timeout request after recovery (2432)
3682 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3683 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3684 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3685 - MDS recovery shouldn't delete objects that it has given out (2730)
3686 - if enqueue arrives after completion, don't clobber LVB (2819)
3687 - don't unlock pages twice when trigger_group_io returns error (2814)
3688 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3689 - don't write pages to disk if there was an error (1450)
3690 - don't ping imports that have recovery disabled (2676)
3691 - take buffered bytes into account when balancing socknal conn (2817)
3692 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3693 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3694 - remove little-endian swabbing of llog records (1987)
3695 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3696 - retry reposting request buffers if they fail (1191)
3697 - grow extent at grant time to avoid granting a revoked lock (2809)
3698 - lock revoke doesn't evict page if covered by a second lock (2765)
3699 - disable VM readahead to avoid reading outside lock extents (2805)
3701 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3702 - updated kernel patches for hp-2.4.20 kernel (2681)
3704 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3707 - fix truncated write corruption (2366)
3708 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3710 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3711 - ensure len(uuid) < 37 in lmc (1171)
3712 - fix ia64 OOPS in llog_test (2255)
3713 - zero end of page at obdfilter for partial page writes (2648)
3714 - don't leave stale dentries around after renames (bug 2428)
3715 - fix timeouts when evicting a client with a single lock held (2642)
3716 - set deadline for the initial HELLO message to drain (2634)
3717 - print out dotted-quad IP addresses in the socknal (2302)
3719 - additional debugging for MDS client eviction problem (2443)
3720 - fix mkfsoptions support for osts (2603, 2604)
3722 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3725 - add series for the vanilla 2.6.0 kernel
3726 - add series for the vanilla 2.4.24 kernel
3727 - add series for a cray x86/64 UL kernel drop
3728 - fix xattr patches for the vanilla 2.4.19 series
3730 - generate true UUIDs in lmc (1171)
3731 - have portals stack dumping break in UML (2466)
3732 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3733 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3734 - wake callback waiting threads on client eviction (2460)
3735 - Add --ptldebug and --subsystem to lmc (1719)
3736 - update assertion to allow safe interrupt allocation
3737 - set rq_no_resend for cancel requests (2432)
3738 - recalculate ptlrpcd timeout after resend (2494)
3739 - call vfs_rmdir when removing pending directories (2368)
3740 - fix renaming a file to itself (2429)
3741 - lmc creates a default one-stripe lov (2454)
3742 - expand procfs space to handle large clusters (2326)
3743 - increase UML stack to avoid overflow
3744 - update lconf's list of debug and subsystem masks
3745 - fix lfs find --obd (2510)
3746 - /proc tunable for disabling filter read caching (2591)
3747 - stop rpm packages from altering slapd.conf (2301)
3748 - disable nagle in the socknal under 0conf (2578)
3749 - choose mds inode size based on stripe count (2572)
3750 - fix kernel-source rpm problems (2516)
3752 - add --disable-doc to avoid pdf generation (2421)
3753 - update documentation, tests, type-os, comments
3754 - avoid format warnings on ia64
3755 - remove the TOE NAL
3756 - tiny code cleanups by removing unused fields
3758 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3761 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3762 - remove the most blatant lies from BUILDING file (2371)
3763 - change default debug level to reasonable production setting
3764 - reduce client side cache size to reduce cache flush time
3765 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3766 - make TCP zerocopy and pinger support enabled by default (2476)
3767 - sync writes completed after process exits caused crashes (2319)
3768 - maintain correct mount count on the MDS (2356)
3769 - backout 1557, because 2316 wasn't really fixed
3770 - better file I/O statistics gathering in /proc
3771 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3772 - another kernel patch to fix zero-copy TCP function export
3773 - don't take duplicate lock when processing re-sent getattr (2420)
3774 - lctl uses obd_self_export instead of creating new conn (2353)
3775 - MDS/OST recovery case which requires object creation asserted (2425)
3776 - move lfs from /usr/sbin to /usr/bin in packages
3777 - fix race between mds_client_add and mds_client_free (2417)
3778 - use kmalloc instead of slabs in portals (2430)
3779 - don't create duplicate records when a failover MDS is present (2442)
3780 - remove unnecessary mount age check (2332)
3781 - don't remove directory inodes from locks prematurely (2451)
3782 - don't break if MDS service name is the same as hostname (2103)
3783 - fix races in client write RPC generation when cache full (2482)
3785 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3788 - remove now-unused request->rq_obd (278)
3789 - if an allocation fails, print out how much memory we've used (1933)
3790 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3791 - add the "configurable stack size" patch to most series files (1256)
3792 - ability to write large log records, for 100+ OST configs (2306)
3793 - fix NULL deref when filter_prep fails (2314)
3794 - fix operator precedence error in filter_sync
3795 - dynamic allocation of socknal TX descriptors (2315)
3796 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3797 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3798 - ensure that CWARN messages in Portals always get to the syslog
3799 - __init/__exit are not for prototype decls (ldlm_init/exit)
3800 - x86-64 compile warning fixes
3801 - fix gateway LMC keyword conflict (2318)
3802 - fix MDS lock inversions in getattr/reint paths (1844)
3803 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3804 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3805 - DLM race condition prevented some lock evictions (2328)
3806 - ENOMEM detection and retry on socknal sends (2230)
3807 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3808 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3809 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3810 - better error messages when a client is rejected during recovery (1505)
3811 - avoid cancelling locks which were never granted, after failure (2330)
3812 - fix i_sem/journal inversion in mds_client_add (2333)
3813 - fix truncate/getattr lock cycle deadlock (2334)
3814 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3815 - fix two rare exit paths which could leak an l_lock() ref (2321)
3816 - fencepost error in MDS/OST orphan recovery (2226)
3817 - make log record alignment 8 bytes (1988)
3818 - lstripe now fails when requested offset > ost_count (2237)
3819 - ensure that all kernel series have a complete list.h (1607)
3820 - fix crashes in special-file operations (2316)
3821 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3823 - allow configurable automake binary, for testing new versions
3824 - small update to the lfs documentation
3826 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3828 * fix negative export reference count in fsfilt_sync (2312)
3830 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3831 * release candidate 0.9.1
3833 - orphans are moved into the PENDING directory for possible recovery
3834 - replayed opens now open by fid for orphan/rename safety (1042)
3835 - last close of an orphan inode generates a transno (683)
3836 - chdir() and mount() now pin the directory entry (1020)
3837 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3838 - discard very old requests without processing them (1502)
3839 - remove some common, well-understood CERRORs (1505)
3840 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3841 - clear "grant" flags in OST replies until OST grant code lands (1644)
3842 - fix read performance by not clobbering i_blksize on client (1598)
3843 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3844 - make LRU size a /proc tunable, clears locks when reduced (707)
3845 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3846 - server locks take references on exports now (1558)
3847 - build fixes for 2.4.20-rh trees (1663)
3848 - return an error from lov_create if all OSCs are inactive (1751)
3849 - fix import levels when a reconnect happens without a timeout (1597)
3850 - exit early from mds_open if we get a lookup error (1749)
3851 - partial page read at EOF wouldn't wait for disk before sending (1642)
3852 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3853 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3854 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3855 - write/truncate lock inversion (1639)
3856 - Don't auto-load obdclass, portals modules during cleanup (1495)
3857 - fix timestamps from jumping to "now" (1763)
3858 - extra journal assertions (1648)
3859 - add an extra multiunlink test (1771)
3860 - fix read_record/write_record API (1776)
3861 - fix leak of offset_extent, possible incorrect i_size later (1772)
3862 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3863 - Debugging for the kqswnal_get_idle_tx problems (1820)
3864 - Allow recovery to be attempted multiple times (1536)
3865 - Write out MDS last_rcvd file after it is first created (1600)
3866 - Fix tx_descriptor leak in failed transmit situations (1827)
3867 - ext3 journaling fixes for assertion failure after IO error (1871)
3868 - class_export_put() on freed export after completion AST error (1896)
3869 - Fix revalidate looping in VFS (1322)
3870 - Don't access a freed export during MDS_REINT timeout (1521)
3871 - Add open-unlink recovery support on the MDS (1673,1764)
3872 - Return an error if no MDS data was read from last_rcvd (1946)
3873 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3874 - Don't setup a disk device that doesn't match exported UUID (317)
3875 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3876 - avoid committing NULL handle in force close
3877 - local.sh is now a one-stripe LOV configuration
3878 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3879 - fix ext3 htree duplicate directory entry corruption (1516)
3880 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3881 - update to new LOV EA format (2097)
3882 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3884 2003-06-15 Phil Schwan <phil@clusterfs.com>
3887 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3888 - per-import/export recovery handling (958, 931, 959)
3889 - multiple last-rcvd slots, for serving multiple FSes (949)
3890 - connections are again shared between multiple imp/exports (963, 964)
3891 - "umount -f" would hang if any requests needed to be sent (393, 978)
3892 - avoid pinning large req buffer by copying for queued messages (989)
3893 - add "uuid" to "lctl device" command to help upcalls (991)
3894 - "open" RPCs with transnos would confuse recovery counters (1037)
3895 - do proper endian conversion of all wire messages (288, 340, 891)
3896 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3897 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3898 - fix ptlrpc_connection leak in target_handle_connect (1174)
3899 - fix import refcounting bug in OST and MDS cleanup (1134)
3900 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3901 - fix very unlikely obd_types race condition (501)
3902 - remove osc_open hack for echo_client (1187)
3903 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3904 - a failure in read_inode2 leads to deadlock (1139)
3905 - cancel ack-locks as soon as transaction is committed (1072)
3906 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3907 - make sure to commitrw after any preprw to avoid deadlock (1162)
3908 - failing to execute a file in a lustre FS would lock inode (1203)
3909 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3910 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3911 - obd_uuid2tgt would walk off the end of the list (1255)
3912 - on IA64 the osc would give portals incorrect bulk size (1258)
3913 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3914 - fix lock inversion caused by new llite matching code (1282)
3915 - limit the number of dirty pages on a client to 10MB (1286)
3916 - timed out locks were not being corrected cancelled (1289)
3917 - fix O_DIRECT above 4GB on IA-32 (1292)
3918 * major user-visible changes
3919 - fail out/fail over policy now controlled by the upcall (993)
3921 - add OBD_PING to check server availability and failure (954)
3922 - lustre messages are now sent in sending host order (288, 340, 891)
3923 - add eadatalen to MDS getattr reply (340)
3924 - OST read replies may contain second buffer, with per-page status (593)
3926 2003-03-11 Phil Schwan <phil@clusterfs.com>
3929 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3930 - failed open()s could cause deadlock; fixed (867, 869)
3931 - stop cancelling OST locks when files are closed (481)
3932 - overlapping XID spaces caused network corruption (851, 853)
3933 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3934 - setattr_raw added, to do single-RPC, server-side setattrs
3935 - lmc/lconf syntax change for OST UUIDs
3936 - fix crashy race condition between ptlrpc_free_req and osc_close
3937 - don't use request in mdc_enqueue if we hit a timeout (889)
3938 - don't set the inode i_size for regular files from the MDS (896)
3939 - handle out of order completion AST (842)
3940 - don't LBUG if a lock request times out after receiving AST (913)
3941 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3942 - if a bad lock AST arrives, send an error instead of dropping entirely
3943 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3944 - fix leak in bulk IO when only partially completed (899, 900, 926)
3945 - fix O_DIRECT for ia64 (55)
3946 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3947 - C-z after timeout could hang a process forever; fixed (977)
3949 - client-side I/O cache (678, 924, 929, 941, 970)
3951 - READPAGE and SETATTRs which don't take server-side locks get
3954 2003-02-11 Phil Schwan <phil@clusterfs.com>
3957 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3958 - Fix fsfilt_extN_readpage() to read a full page of directory
3959 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3960 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3961 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3962 - Most intent operations were removed, in favour of a new RPC mode
3963 that does a single RPC to the server and bypasses most of the VFS
3964 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3965 - Aggressively cancel local locks on DLM servers
3966 - mds_reint_unlink sends EA to the client if it's the last nlink.
3967 client uses that EA to unlink OST objects.
3968 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3969 - recursive symlinks were fixed (439)
3970 - fixed NULL deref in DEBUG_REQ
3971 - filter_update_lastobjid no longer calls sync, which annoyed extN
3972 - fixed multi-client small-writes to a single file problem (445)
3973 - fixed mtime updates during file writes (607)
3974 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3975 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3976 - fixed Linux OST size reporting problem (444, 656)
3977 - OST now updates object mtime with writes or setattr (607, 619)
3978 - client verifies file size before zeroing page past EOF (445)
3979 - OST now writes last allocated objid to disk with allocation (108)
3980 - LOV on echo now works (409)
3982 - mds_reint_unlink sends a new buffer, with the EA included. this
3983 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3984 set if a regular file was being unlinked, and it was the last link
3985 - use PtlGet from the target for bulk writes (315)
3986 - OST now updates object mtime with writes or setattr (607, 619)
3987 - LDLM now has a grant-time callback to revalidate locked items, if
3989 - Many MDS operations were reorganized to combat race conditions
3991 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3992 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3994 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3997 - Fully reactivate OST imports after reconnection (512, others)
3998 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3999 - More graceful error handling for truncating on dead OST (515)
4000 - Don't error out unless we're actually accessing dead stripes (474)
4001 - Fix garbage sizes when stripes are missing (410)
4002 - LRU counters were broken, causing constant lock purge (433, 432)
4003 - garbage on read from stripes with failed OSTs (441)
4004 - mark OSCs as active before reconnecting during recovery (438)
4005 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4006 - lfind did not preserve OST order in output (443)
4007 - symlinks cause hung clients, incorrect data (439)
4008 - stop dereferencing request after dropping refcount (457)
4009 - don't LASSERT(spin_is_locked) on non-SMP (455)
4010 - fixes for many rename() bugs
4011 - fstat didn't correctly synchronize attributes (399)
4012 - server must handle lock cancellation during blocking AST prep (487)
4013 - bulk descriptors were free()d too soon (511)
4014 - fix paths in lconf, which would load incorrect modules (451, 507)
4015 - fix confusing lconf 'host not found' error message (386)
4016 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4017 - fix race condition in mdc_blocking_ast() for inode access (526)
4018 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4019 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4020 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4021 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4022 - fix lstripe in several ways (kernel side) (527)
4023 - fix request leak in ldlm_cli_enqueue (262)
4024 - incorrect OSC was marked inactive after OST failure
4025 - call mds_fs_cleanup before unmounting filesystem (524)
4026 - fix races between taking ns_lock and ldlm_lock_change_resource
4027 - fix races updating LOV export open file list
4028 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4029 - fix recovery NULL deref in ldlm_cli_cancel_unused
4030 - fix some DLM races by using new hash table for lock handles (419)
4031 - permit the client to specify desired inodes, at replay
4032 - duplicate requests when we queue them for replay reintegration
4033 - fix last_rcvd offset calculation
4034 - sync after each recovered transaction, so we always make progress
4035 - never, not always, ERESTART requests without transnos
4036 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4038 - skip replay if the MDS says that the client is already connected
4039 - don't check for a recovery-enabled export to match lctl's UUID
4040 - don't INC_USE_COUNT for phantom exports
4041 - don't crash when cleaning up phantom exports (567)
4042 - don't double-finish or set replay data for errored mdc_open requests
4043 - abort requests when they time out, so we don't get old replies
4044 - send/receive replies for AST messages again
4045 - if the client says that it doesn't have the lock, cancel it on the
4047 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4048 mark it as destroyed, it will all work out when decref is called
4049 - fix module use counts (22, 581)
4051 - ASTs now expect a reply (server cancels lock on error reply)
4053 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4056 - fix many simultaneous client startup (392)
4057 - fix dentry->d_it clobbering
4058 - credentials weren't being shipped for readdir/getattr operations
4059 - remove invalid assertions triggered during some concurrent MD
4061 - proper Lustre versions added (336, 389)
4062 - fix memory leak for create error case (398)
4063 - fix LOV locking bug that would get cli/srv out of sync
4064 - fix echo client over LOV (409)
4065 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4066 - fix double-O_EXCL intent crash (424)
4067 - avoid sending multiple lock CANCELs (352)
4069 - MDS can do multi-client recovery (modulo bugs in new code)
4071 - many updates, edits, cleanups
4073 2002-11-18 Phil Schwan <phil@clusterfs.com>
4076 - fix null d_it dereference (346)
4077 - fix full OST/dbench hang (333)
4078 - fix permission problem with file removal (286)
4079 - fix removal of OSCs from LOV when they fail
4080 - fix NULL deref during bulk timeout (214)
4081 - fix problems related to multiple filesystems on one MDS (241)
4082 - fixed serious subtle metadata locking bugs
4083 - free locks on clients when inodes are removed due to memory
4085 - fix inode pointer in lock data (285)
4086 - partial support for multiple MDS on a single host (241)
4087 - data locks weren't cancelled at clear_inode time (290, 311)
4088 - intent locks could lead to unbounded lock growth (205)
4089 - added a maximum lock count, an LRU list, and a flusher
4090 - fix multiple rename (365)
4091 - properly abstracted the echo client
4092 - OSC locked 1 byte too many; fixed
4093 - rewrote brw callback code:
4094 - fixed recovery bugs related to LOVs (306)
4095 - fixed too-many-pages-in-one-write crash (191)
4096 - fixed (again) crash in sync_io_timeout (214)
4097 - probably fixed callback-related race (385)
4099 - Add capability to MDS protocol
4100 - LDLM cancellations and callbacks on different portals
4102 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4105 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4106 - defer open object destruction to close time (601981, 138)
4107 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4108 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4109 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4110 - don't call obd_setattr in ll_file_release for destroyed objects
4111 * protocol change to lustre_msg: move |version| and add |flags|
4112 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4113 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4114 * added replay of create, unlink, link and rename operations during
4115 MDS failover; recovery should be much more robust now
4116 * remove failed OSCs from LOVs (only lov_create uses this so far)
4117 * the lustre-HOWTO was brought (more) up to date (582544)
4119 2002-10-23 Phil Schwan <phil@clusterfs.com>
4122 - in-use dentries weren't being reused properly (617851)
4123 - prevent multiple LDLM setup (599178)
4124 - fix LOV size calculations for truncate (617853)
4125 - fix client handling of MDS intent errors (POSIX)
4126 - fix permission bug in lovstripe.c test (624321)
4127 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4128 - truncate past end of file could corrupt data
4129 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4130 - a race in recovery could return ETIMEDOUT to apps (623947)
4131 - building outside the source directory was fixed
4132 * the lustre-HOWTO was brought (more) up to date (582544)
4133 * major progress was made on recovery functionality
4135 2002-10-10 Phil Schwan <phil@clusterfs.com>
4138 - recovery deadlock fix
4139 - rm -rf causes LBUG fix (617817)
4140 - file open by multiple tasks fix (618962)
4141 - directory permissions bugs (602707 and 620007)
4142 - journal_stop fixed with locking (611313)
4143 - O_APPEND failures resolved (618273, perhaps 614459)
4144 - lconf PATH fix (619770)
4145 - IA64 build fix (621450)
4146 - RPC buffer sizes scale with amount of memory
4148 2002-10-01 Phil Schwan <phil@clusterfs.com>
4151 - locks would be cancelled without throwing away data pages,
4152 resulting in inconsistent data (605627)
4153 - inode attributes were not always being refreshed (605627, 612449)
4154 - lconf now continues to cleanup after lctl reports an error
4155 - MDS now enforces user permissions (602707)
4156 - lprocfs cleanup fixed, but not yet enabled (614157)
4157 - fixed infinite server hang, should a client not respond to an AST
4158 - avoid going into recovery if user calls readlink() with a buffer
4159 that's too small (613941)
4160 - AST RPCs no longer require replies (614867) -- this may be changed
4161 - don't crash server if client sends an IOV that's too big (611336)
4162 - fixed lock conversion deadlock (611892)
4163 - fixed the following of symlinks (614622)
4164 * recovery: the server can remove locks from a client that dies, other
4165 clients can make progress
4166 * more extN patch fixes
4167 * compile-time configurable ptlrpc buffer allocations
4169 - collaborative read cache document
4170 - Lustre Lite Performance CDR document-in-progress
4172 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4175 - fix typo in patch-2.4.18
4177 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4180 - clear ptlrpc request each time in handle_incoming_request()
4181 - unlink of files now destroys the object on the OST
4183 2002-09-19 Peter Braam <braam@clusterfs.com>
4185 * add hard link support
4186 * change obdfile creation method
4187 * kernel patch changed
4189 2002-09-19 Peter Braam <braam@clusterfs.com>
4192 - stack overflow bug in extN fixed
4194 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4196 * documentation updates
4197 - add man pages for config tools
4198 - update tests/README to describe testing with new config tools
4199 - finish metadata API descriptions
4200 * bug fixes and cleanups
4201 - statfs workaround for 16TB limit
4202 - LOV stripe allocation improved, can stripe on subset of OSTs
4203 - LOV file size/IO offset was wrong for files > 4GB in size
4204 - object EA data was being dropped, caused files to be unreadable
4205 - memory overflow with non-LOV OST caused memory corruption
4206 - fixed regression tests to work with new config tools, obdfilter
4207 - fixed bug when directory size became larger than 1 block
4208 - fixed bug (for single client case) when PWD was deleted
4209 - invalidate local directory pages when doing intent-based ops
4210 - avoid LDLM oops when lock callback contained bad data
4212 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4214 * documentation updates
4215 * bug fixes and cleanups
4216 - configuration tools
4219 - 64-bit compile warnings
4220 - 64-bit internal statfs data
4222 * test_brw on persistent OST devices
4224 * lprocfs (disabled)
4226 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4228 * documentation updates
4229 * bug fixes and cleanups
4230 * configuration tools
4232 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4235 * many small fixes to 0_5_4
4236 * io/network handling
4237 * thinkos in MDS operations
4239 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4242 * crucial basic fixes to 0.5.3
4243 * IOR, Iozone work over Elan
4246 2002-08-07 Phil Schwan <phil@clusterfs.com>
4247 * version 0_5_3, our first alpha
4248 * we use the new Portals iovs
4249 * documentation updates
4250 * bug fixes and cleanups
4251 * small changes in the DLM wire protocol
4253 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4254 * version 0_5_1 with some initial stability,
4255 * locking on MD and file I/O.
4256 * documentation updates
4257 * several bug fixes since 0.5.0
4258 * small changes in wire protocol
4260 2002-07-18 Phil Schwan <phil@clusterfs.com>
4262 * delivered as Lustre Light Alpha
4263 * fixed a crash after handling invalid MDS requests
4264 * fixed directory pages for architectures with non-4k pages sizes
4266 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4267 * release version v0_4_4
4268 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4270 * Fixes a number of interruption problems with OST operations.
4271 * Update documentation for portals header changes
4272 * Move all wire protocol structs/defines to lustre_idl.h
4273 * Fixes symlink length bug.
4274 * Add tcpdump to repository.
4276 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4277 * release version v0_4_3
4278 * Fixes statfs for inodes on extN.
4279 * Fixes bug in runtests which would delete /etc/hosts.
4280 * Use 64-bit object IDs wherever possible (not into VFS though)
4281 Remove ost_get_info, which is unused by lustre, and out of date.
4283 2002-07-03 Peter Braam <braam@clusterfs.com>
4284 * release version v0_4_2 Fixes a lookup error (type not passed)
4285 * move forward to head of Portals
4286 * move forward to latest Lustre kernel
4288 2002-06-25 Peter Braam <braam@clusterfs.com>
4289 * release version v0_4_1. Hopefully stable on single node use.