3 2.6.5-7.287.3 (SLES 9),
4 2.6.9-55.0.12.EL (RHEL 4),
5 2.6.16.54-0.2.3 (SLES 10),
6 2.6.18-53.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-cfs5
13 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
17 Description: make mgs_setparam() handle fsname containing dash
18 Details : fsname containing a dash does not work with lctl conf_param
20 Severity : enhancement
22 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
24 Severity : enhancement
26 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
30 Frequency : rare, at shutdown
31 Description: access already free / zero obd_namespace.
32 Details : if client_disconnect_export was called without force flag set,
33 and exist connect request in flight, this can produce access to
34 NULL pointer (or already free pointer) when connect_interpret
35 store ocd flags in obd_namespace.
39 Frequency : only at startup
40 Description: not alloc memory with spinlock held.
41 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
44 Severity : enhancement
46 Description: make lustre randomly fail allocating memory
47 Details : Make lustre randomly failed allocating memory for testing purpose.
49 Severity : enhancement
51 Description: lost problems with lov objid file
52 Details : Fixes some scability and access to not inited memory problems
53 in work with lov objdid file.
58 Description: lfs find does not continue on file error
59 Details : Continue other files processing when a file/dir is absent.
63 Description: Inconsistent usage of lustre_pack_reply()
64 Details : Standardize the usage of lustre_pack_reply() such that it
65 always generate a CERROR on failure.
68 Frequency : occasional
70 Description: MDS hang or stay in waiting lock
71 Details : If client receive lock with CBPENDING flag ldlm need send lock
72 cancel as separate rpc, to avoid situation when cancel request
73 can't processed due all i/o threads stay in wait lock.
77 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
81 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
85 Description: Update to SLES9 kernel-2.6.5-7.287.3.
89 Description: Update to RHEL4 latest kernel.
91 Severity : enhancement
93 Description: Build SLES10 patchless client fails
94 Details : The configure was broken by run ./configure with
95 --with-linux-obj=.... argument for patchless client. When the
96 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
97 can't search header adequately. Use absolute path such as
98 -I($LINUX)/include instead.
102 Description: interrupt oig_wait produce painc on resend.
103 Details : brw_redo_request can be used for resend requests from ptlrpcd and
104 private set, and this produce situation when rq_ptlrpcd_data not
105 copyed to new allocated request and triggered LBUG on assert
106 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
107 ptlrpcd set if request is changed and can be safety changed to use
112 Description: LASSERT_{REQ,REP}SWAB macros are buggy
113 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
114 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
115 these macros and replace them with their difinitions instead.
119 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
120 Details : Update which_patch & target file for SLES10 latest kernel.
123 Description: Remove client patches from SLES 10 kernel.
124 Details : This causes SLES 10 clients to behave as patchless clients
125 even on a Lustre-patched (server) kernel.
127 Severity : enhancement
129 Description: self-adjustable client's lru lists
130 Details : use adaptive algorithm for managing client cached locks lru
131 lists according to current server load, other client's work
132 pattern, memory activities, etc. Both, server and client
133 side namespaces provide number of proc tunables for controlling
136 Severity : enhancement
138 Description: light-weight GSS support
139 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
140 and reduce performance overhead.
142 Severity : enhancement
144 Description: Linux keyring support
145 Details : Support using service of Linux keyring for Lustre GSS internal
146 context refresh/cache mechanism.
150 Description: Fix errors in lfs documentation
151 Details : Fixes man pages
155 Description: don't use GFP_* in generic Lustre code.
156 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
160 Description: obdclass is limited by single OBD_ALLOC(idarray)
161 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
165 Description: warnings with build patchless client with vanila 2.6.19 and up
166 Details : change old ctl_table style and replace ctl_table/ctl_table_header
167 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
171 Frequency : Only for SLES
172 Description: sanity_quota fail test_1
173 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
174 file, which makes TSTID[2] inunique.
178 Frequency : Only for SLES9
179 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
180 Details : Update target/ChangeLog/which_patch .
184 Frequency : Only for test_10 in sanity-quota.sh
185 Description: a bug in quota test code
186 Details : checking a test flag in a wrong place
190 Frequency : Only for test_13 in sanity-quota.sh
191 Description: a bug in quota test script
192 Details : shouldn't check overfull allocated quota
196 Description: jbd statistics
197 Details : Port older jbd statistics patch for sles10
199 Severity : enhancement
201 Description: Add printing inode info into message about error in writepage.
206 Description: reply_lock_interpret crash due to race with it and lock cancel.
207 Details : Do not replay locks that are being cancelled. Do not reference
208 locks by their address during replay, just by their handle.
212 Frequency : When flocks are used.
213 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
214 Details : Flock locks might destroy just granted lock if it could be merged
215 with another existing flock, this is done in completion handler,
216 so teach ldlm_cli_enquque_fini that this is a valid case for
222 Description: Oops in read and write path when failing to allocate lock.
223 Details : Check if lock allocation failed and return error back.
227 Description: lstripe command fails for valid OST index
228 Details : The stripe offset is compared to lov->desc.ld_tgt_count
229 instead of lov->desc.ld_active_tgt_count.
233 Description: sanity.sh failed test 103
234 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
239 Description: df doesn't work properly if diskfs blocksize != 4K
240 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
244 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
245 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
249 Description: obd_health_check_timeout too short
250 Details : set obd_health_check_timeout as 1.5x of obd_timeout
254 Description: llapi_file_create() does not allow some changes
255 Details : add llapi_file_open() that allows specifying the mode and
256 open flags, and also returns an open file handle.
260 Description: merge and cleanup kernel patches.
261 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
265 Description: Add journal checksum support.(Kernel part)
266 Details : The journal checksum feature adds two new flags i.e
267 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
268 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
269 indicates that the commit block contains the checksum for
270 the blocks described by the descriptor blocks. Now commit
271 record can be sent to disk without waiting for descriptor
272 blocks to be written to disk. This behavior is controlled
273 using JBD2_FEATURE_ASYNC_COMMIT flag.
277 Description: OSS needs mutliple precreate threads
278 Details : Add ability to start more than one create thread per OSS.
282 Description: Sanity.sh test_65k failed: setstripe should have succeeded
283 Details : execute lfs setstripe on client
287 Description: mds_obd_create error creating tmp object
288 Details : When the user sets quota on root, llog will be affected and can't
289 create files and write files.
292 Frequency : Always on ia64 patchless client, and possibly others.
294 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
295 Details : This allows the patchless client to be loaded on architectures
300 Description: RedHat Update kernel for RHEL5
301 Details : Add the kernel config file for RHEL5.
305 Description: RedHat Update kernel for RHEL5
306 Details : Modify the kernel config file more closer RHEL5.
310 Description: Build failure against Centos5 (RHEL5)
311 Details : Define PAGE_SIZE when it isn't present.
315 Description: test_7 on sanity-quota.sh can't be used on separate machines
316 Details : add facet to handle it
320 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
321 Details : with NFS, the anon dentry's parent was set to itself in
322 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
323 corresponding dentry other than use rec->ur_name.
325 Severity : enhancement
327 Description: lfs setstripe enhancement
328 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
332 Description: enable data checksumming by default
333 Details : enable checksum by default, allow --disable-checksum
334 configure option and "-o nochecksum" mount option. Checksums
335 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
339 Description: lustre support for RHEL5
340 Details : Add support for RHEL5.
344 Description: Client eviction due to failover config
345 Details : after a connection loss, the lustre client should attempt to
346 reconnect to the last active server first before trying the
347 other potential connections.
351 Description: when mds and osts use different quota unit(32bit and 64bit),
352 quota will be released repeatly.
353 Details : void sending multiple quota reqs to mds, which will keep the
354 status between the reqs.
358 Description: osts not allocated evenly to files
359 Details : change the condition to increase offset_idx
362 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
364 Description: Data corruption for OSTs that are formatted larger than 2TB
366 Details : When generating the bio request for lustre file writes the
367 sector number would overflow a temporary variable before being
368 used for the IO. The data reads correctly from Lustre (which
369 will overflow in a similar manner) but other file data or
370 filesystem metadata may be corrupted in some cases.
374 Description: Tune the kernel for good SCSI performance.
375 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
376 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
381 Description: rewrite ext2-derived code in obdclass/uuid.c
382 Details : rewrite inherited code (uuid parsing code from ext2 utils)
383 from scratch preserving functionality.
387 Description: "lfs find -obd UUID" prints directories
388 Details : "lfs find -obd UUID" will return all directory names instead
389 of just file names. It is incorrect because the directories
390 do not reside on the OSTs.
394 Description: use i_size_read and i_size_write in 2.6 port
395 Details : replace inode->i_size access with i_size_read/write()
398 Frequency : when removing large files
400 Description: scheduling issue during removal of large Lustre files
401 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
402 It causes scheduling issues when removing large files (17TB in the
406 Frequency : only with liblustre clients on XT3
408 Description: evictions taking too long
409 Details : allow llrd to evict clients directly on OSTs
412 Frequency : only on ppc
414 Description: /proc/fs/lustre/devices broken on ppc
415 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
416 We should make sure the type of 'index' is loff_t and then cast
417 explicitly as needed below. Do not assign an explicitly cast
421 Frequency : only for rhel5
423 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
424 Details : Modify the target file & which_kernel.
426 Severity : enhancement
428 Description: omit set fsid for export NFS
429 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
430 as FS_REQUIRES_DEV for avoid problems with generate fsid.
434 Frequency : Always, for kernels after 2.6.16
435 Description: Fix warning idr_remove called for id=.. which is not allocated.
436 Details : Last kernels save old s_dev before kill super and not allow
437 to restore from callback - restore it before call kill_anon_super.
441 Description: buffer overruns could theoretically occur
442 Details : llapi_semantic_traverse() modifies the "path" argument by
443 appending values to the end of the origin string, and a
444 overrun may occur. Adding buffer overrun check in liblustreapi.
448 Description: Fix error on 'ls .' at the top of the Lustre mount.
449 Details : Don't revalidate dentry if it is a root dentry.
453 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
454 Details : Modify vm-tunables-rhel4.patch.
458 Description: Kernel config for 2.6.18-vanilla.
459 Details : Modify targets/2.6-vanilla.target.in.
460 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
461 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
462 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
463 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
467 Description: improve handling recoverable errors
468 Details : if request processig with error which can be recoverable on server
469 request should be resend, otherwise page released from cache and
474 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
475 Details : Modify target file & which_patch.
476 A flaw was found in the IA32 system call emulation provided
477 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
478 value could be stored in the %RAX register, which could trigger an
479 out-of-bounds system call table access. An untrusted local user
480 could exploit this flaw to run code in the kernel
481 (ie a root privilege escalation). (CVE-2007-4573).
485 Description: O_DIRECT bypasses client statistics.
486 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
487 read_ahead_stats were updated. copied the stats section from
488 osc_send_oap_rpc() into async_internal().
492 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
496 Description: change order of libsysio includes
497 Details : '#include sysio.h' should always come before '#include xtio.h'
501 Description: handle "serious error: objid * already exists" more gracefully
502 Details : If LAST_ID value on disk is smaller than the objects existing in
503 the O/0/d* directories, it indicates disk corruption and causes an
504 LBUG(). If the object is 0-length, then we should use the existing
505 object. This will help to avoid a full fsck in most cases.
507 Severity : enhancement
509 Description: adapt the lustre_config script to support the upgrade case
510 Details : Add "-u" option for lustre_config script to support upgrading 1.4
511 server targets to 1.6 in parallel.
515 Description: To avoid grant space > avaible space when the disk is almost
516 full. Without this patch you might see the error "grant XXXX >
517 available" or some LBUG about grant, when the disk is almost
519 Details : In filter_check_grant, for non_grant cache write, we should
520 check the left space by if (*left > ungranted + bytes), instead
521 of (*left > ungranted), because only we are sure the left space
522 is enough for another "bytes", then the ungrant space should be
523 increase. In client, we should update cl_avail_grant only there
524 is OBD_MD_FLGRANT in the reply.
528 Description: Update RHEL 4 kernel to fix local root privilege escalation.
529 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
530 described in CVE-2007-4573. This problem could allow untrusted
531 local users to gain root access.
534 Frequency : when using O_DIRECT and quotas
536 Description: Incorrect file ownership on O_DIRECT output files
537 Details : block usage reported by 'lfs quota' does not take into account
538 files that have been written with O_DIRECT.
543 Description: touch file failed when fs is not full
544 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
545 otherwise we can get ENOSP while fs is not full.
549 Description: parallel lock callbacks
550 Details : Instead of sending blocking and completion callbacks as separated
551 requests, adding them to a set and sending in parallel.
554 Frequency : only for Cray XT3
555 Bugzilla : 12829/13455
556 Description: Changing primary group doesn't change the group lustre assigns to
558 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
559 overridden with the primary group provided by the group upcall,
560 whereas the supplied fsgid can be trusted if it is in the list of
561 supplementary groups returned by the group upcall.
563 Severity : enhancement
565 Description: Allow masking D_WARNING, D_ERROR messages from console
566 Details : Console messages can now be disabled via lnet.printk.
568 --------------------------------------------------------------------------------
570 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
572 * Support for kernels:
573 2.6.5-7.283 (SLES 9),
574 2.6.9-55.EL (RHEL 4),
575 2.6.16.46-0.14 (SLES 10),
576 2.6.18.8 vanilla (kernel.org)
577 * Client support for unpatched kernels:
578 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
579 2.6.16 - 2.6.22 vanilla (kernel.org)
580 * Due to recently discovered recovery problems, we do not recommend
581 using patchless RHEL 4 clients with this or any earlier release.
582 * Recommended e2fsprogs version: 1.39.cfs8
583 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
584 * Starting with this release, the ldiskfs backing filesystem required
585 by Lustre is now in its own package, lustre-ldiskfs. This package
586 should be installed. It is versioned separately from Lustre and
587 may be released separately in future.
591 Description: block reactivating mgc import until all deactivates complete
592 Details : Fix race when failing back MDT/MGS to itself (testing)
595 Frequency : at statup only
597 Description: mds_lov_synchronize race leads to various problems
598 Details : simultaneous MDT->OST connections at startup can cause the
599 sync to abort, leaving the OSC in a bad state.
601 Severity : enhancement
603 Description: add optional extra BUILD_VERSION info
604 Details : add a new environment variable (namely LUSTRE_VERS) which allows
605 to override the lustre version.
608 Frequency : 2.6.18 servers only
610 Description: ll_kern_mount() doesn't release the module reference
611 Details : The ldiskfs module reference count never drops down to 0
612 because ll_kern_mount() doesn't release the module reference.
617 Description: server LBUG when using old ost_num_threads parameter
618 Details : Accept the old ost_num_threads parameter but warn that it
619 is deprecated, and fix an off-by-one error that caused an LBUG.
624 Description: Transient SCSI error results in persistent IO issue
625 Details : iobuf->dr_error is not reinitialized to 0 between two
629 Frequency : sometimes when underlying device returns I/O errors
631 Description: OSTs not going read-only during write failures
632 Details : OSTs are not remounted read-only when the journal commit threads
633 get I/O errors because fsfilt_ext3 calls journal_start/stop()
634 instead of the ext3 wrappers.
638 Description: poor connect scaling with increasing client count
639 Details : Don't run filter_grant_sanity_check for more than 100 exports
640 to improve scaling for large numbers of clients.
643 Frequency : SLES10 only
645 Description: sanity-quota.sh quotacheck failed: rc = -22
646 Details : Quotas cannot be enabled on SLES10.
649 Frequency : liblustre clients only
651 Description: getdirentries does not give error when run on compute nodes
652 Details : getdirentries does not fail when the size specified as an argument
653 is too small to contain at least one entry
655 Severity : enhancement
657 Description: Add LNET router traceability for debug purposes
658 Details : If a checksum failure occurs with a router as part of the
659 IO path, the NID of the last router that forwarded the bulk data
660 is printed so it can be identified.
665 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
666 Details : Due to a race condition, liblustre clients were occasionally
669 Severity : enhancement
671 Description: lfs setstripe use optional parameters instead of postional
674 Severity : enhancement
676 Description: Nanosecond timestamp support for ldiskfs
677 Details : The on-disk ldiskfs filesystem has added support for nanosecond
678 resolution timestamps. There is not yet support for this at
679 the Lustre filesystem level.
682 Frequency : during server recovery
684 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
685 Details : request with rq_no_resend flag not awake l_wait_event if they get a
689 Frequency : nfs export on patchless client
691 Description: connectathon hang when test nfs export over patchless client
692 Details : Disconnected dentry cannot be found with lookup, so we do not need
693 to unhash it or make it invalid
696 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
697 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
698 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
703 Description: open req refcounting wrong on reconnect
704 Details : If reconnect happened between getting open reply from server and
705 call to mdc_set_replay_data in ll_file_open, we will schedule
706 replay for unreferenced request that we are about to free.
707 Subsequent close will crash in variety of ways.
708 Check that request is still eligible for replay in
709 mdc_set_replay_data().
714 Description: disable writes to filesystem when reading health_check file
715 Details : the default for reading the health_check proc file has changed
716 to NOT do a journal transaction and write to disk, because this
717 can cause reads of the /proc file to hang and block HA state
718 checking on a healthy but otherwise heavily loaded system. It
719 is possible to return to the previous behaviour during configure
720 with --enable-health-write.
722 Severity : enhancement
724 Description: 64-bit inode version
725 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
726 to the inode. This will be required for version-based recovery.
731 Description: MDS fails to start if a duplicate client export is detected
732 Details : in some rare cases it was possible for a client to connect to
733 an MDS multiple times. Upon recovery the MDS would detect this
734 and fail during startup. Handle this more gracefully.
736 Severity : enhancement
738 Description: Add -o localflock option to simulate old noflock
740 Details : This will achieve local-only flock/fcntl locks
746 Description: log_commit_thread vs filter_destroy race leads to crash
747 Details : Take import reference before releasing llog record semaphore
752 Description: Wrong request locking in request set processing
753 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
754 imp_delayed_list, in this place should be used imp_lock.
757 Frequency : when reconnection
759 Description: Grant Leak when osc reconnect to OST
760 Details : When osc reconnect ost, OST(filter) should check whether it
761 should grant more space to client by comparing fed_grant and
762 cl_avail_grant, and return the granted space to client instead
763 of "new granted" space, because client will call osc_init_grant
764 to update the client grant space info.
767 Frequency : when client reconnect to OST
769 Description: Grant Leak when osc do resend and replay bulk write
770 Details : When osc reconnect to OST, OST(filter)should clear grant info of
771 bulk write request, because the grant info while be sync between
772 OSC and OST when reconnect, and we should ignore the grant info
773 these of resend/replay write req.
778 Description: Grant space more than avaiable left space sometimes.
779 Details : When then OST is about to be full, if two bulk writing from
780 different clients came to OST. Accord the avaliable space of the
781 OST, the first req should be permitted, and the second one
782 should be denied by ENOSPC. But if the seconde arrived before
783 the first one is commited. The OST might wrongly permit second
784 writing, which will cause grant space > avaiable space.
787 Frequency : when client is evicted
789 Description: Grant might be wrongly erased when osc is evicted by OST
790 Details : when the import is evicted by server, it will fork another
791 thread ptlrpc_invalidate_import_thread to invalidate the
792 import, where the grant will be set to 0. While the original
793 thread will update the grant it got when connecting. So if
794 the former happened latter, the grant will be wrongly errased
795 because of this race.
800 Description: Checking Stale with correct fid
801 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
802 to check whether it is stale, because sometimes, we want the
803 enqueue happened anyway, and op_data.fid2 will not be initialized.
805 Severity : enhancement
807 Description: update patchless client
808 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
811 Frequency : only with 2.4 kernel
813 Description: random memory corruption
814 Details : size of struct ll_inode_info is to big for union inode.u and this
815 can be cause of random memory corruption.
820 Description: Memory leak in recovery
821 Details : Lov_mds_md was not free in an error handler in mds_create_object.
822 It should also check obd_fail before fsfilt_start, otherwise if
823 fsfilt_start return -EROFS,(failover mds during mds recovery).
824 then the req will return with repmsg->transno = 0 and rc = EROFS.
825 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
826 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
827 be freed no matter whether fsfilt_commit success or not.
830 Frequency : only with huge count clients
832 Description: Prevents from taking the superblock lock in llap_from_page for
834 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
835 for llap_from_page call in ll_removepage prevents from taking the
836 superblock lock for a soon died page.
841 Description: Not check open intent error before release open handle
842 Details : in some rare cases, the open intent error is not checked before
843 release open handle, which may cause
844 ASSERTION(open_req->rq_transno != 0), because it tries to release
845 the failed open handle.
850 Description: Set cat log bitmap only after create log success.
851 Details : in some rare cases, the cat log bitmap is set too early. and it
852 should be set only after create log success.
856 Description: Accessing a block bevice can re-enable I/O when Lustre is
857 tearing down a device.
858 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
863 Description: service threads may hog cpus when there are a lot of requests
864 Details : Insert cond_resched to give other threads a chance to use some CPU
869 Description: the cat log was not initialized in recovery
870 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
871 unlink log on mds will not be initialized until mds post
872 recovery. And also in mds post recovery, the unlink log will
873 initialization will be done asynchronausly, so there will be race
874 between add unlink log and unlink log initialization.
878 Description: brw_stats were being printed incorrectly
879 Details : brw_stats were being printed as log2 but all of them were not
880 recorded as log2. Also remove some code duplication arising from
881 filter_tally_{read,write}.
885 Frequency : rare, only in recovery.
886 Description: ASSERTION(req->rq_type != LI_POISON) failed
887 Details : imp_lock should be held while iterating over imp_sending_list for
888 prevent destroy request after get timeout in ptlrpc_queue_wait.
892 Description: replay-single.sh test 52 fails
893 Details : A lock's skiplist need to be cleanup when it being unlinked
894 from its resource list.
898 Description: Short directio read returns full requested size rather than
900 Details : Direct I/O operations should return actual amount of bytes
901 transferred rather than requested size.
903 Severity : enhancement
905 Description: metadata RPC reduction (e.g. for rm performance)
906 Details : decrease the amount of synchronous RPC between clients and servers
907 by canceling conflicing lock before the operation on the client
908 and packing thier handles into the main operation RPC to server.
910 Severity : enhancement
912 Description: Async OSC create to avoid the blocking unnecessarily.
913 Details : If a OST has no remain object, system will block on the creating
914 when need to create a new object on this OST. Now, ways use
915 pre-created objects when available, instead of blocking on an
916 empty osc while others are not empty. If we must block, we block
917 for the shortest possible period of time.
919 Severity : enhancement
921 Description: refine locking for avoid write wrong info into lov_objid file
922 Details : fix possible races with add new target and write/update data in
925 --------------------------------------------------------------------------------
927 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
932 Frequency : on some architectures
934 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
935 Details : Uninitialized flags sometimes cause configuration commands to
939 Frequency : patchless clients only
941 Description: missing __iget() symbol export
942 Details : The __iget() symbol export is missing. To avoid the need for
943 this on patchless clients the deathrow inode reaper is turned
944 off, and we depend on the VM to clean up old inodes. This
945 dependency was during via the fix for bug 12181.
947 --------------------------------------------------------------------------------
949 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
951 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
952 INTEROPERATE with older versions automatically. In many cases a
953 special upgrade step is needed. Please read the
954 user documentation before upgrading any part of a live system.
955 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
956 version will not interoperate with 1.6 betas before beta5 (1.5.95).
957 * WARNING: Lustre configuration and startup changes are required with
958 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
964 Severity : enhancement
966 Description: MountConf
967 Details : Lustre configuration is now managed via mkfs and mount
968 commands instead of lmc and lconf. New obd types (MGS, MGC)
969 are added for dynamic configuration management. See
970 https://mail.clusterfs.com/wikis/lustre/MountConf for
973 Severity : enhancement
975 Description: dynamic OST addition
976 Details : OSTs can now be added to a live filesystem
978 Severity : enhancement
980 Description: startup order invariance
981 Details : MDTs and OSTs can be started in any order. Clients only
982 require the MDT to complete startup.
984 Severity : enhancement
986 Description: parallel, asynchronous orphan cleanup
987 Details : orphan cleanup is now performed in separate threads for each
988 OST, allowing parallel non-blocking operation.
990 Severity : enhancement
992 Description: optimized stripe assignment
993 Details : stripe assignments are now made based on ost space available,
994 ost previous usage, and OSS previous usage, in order to try
995 to optimize storage space and networking resources.
997 Severity : enhancement
999 Description: Permanently set tunables
1000 Details : All writable /proc/fs/lustre tunables can now be permanently
1001 set on a per-server basis, at mkfs time or on a live system.
1003 Severity : enhancement
1005 Description: Lustre message v2
1006 Details : Add lustre message format v2.
1008 Severity : enhancement
1010 Description: client OST exclusion list
1011 Details : Clients can be started with a list of OSTs that should be
1012 declared "inactive" for known non-responsive OSTs.
1016 Description: SPEC SFS validation failure on NFS v2 over lustre.
1017 Details : Changes the blocksize for regular files to be 2x RPC size,
1018 and not depend on stripe size.
1020 Severity : enhancement
1022 Description: Multiple MD RPCs in flight.
1023 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1024 To avoid overly-overloading MDS, introduce a limit on number of
1025 MDS RPCs in flight for a single client and add /proc controls
1026 to adjust this limit.
1028 Severity : enhancement
1030 Description: client read/write statistics
1031 Details : Add client read/write call usage stats for performance
1032 analysis of user processes.
1033 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1034 file access. extents_stats shows chunk size distribution.
1035 extents_stats_per_process show chunk size distribution per
1038 Severity : enhancement
1040 Description: mds statistics
1041 Details : Add detailed mds operations statistics in
1042 /proc/fs/lustre/mds/*/stats.
1046 Description: Failure of copying files with lustre special EAs.
1047 Details : Client side always return success for setxattr call for lustre
1048 special xattr (currently only "trusted.lov").
1053 Description: Refcount LNET uuids
1054 Details : The global LNET uuid list grew linearly with every startup;
1055 refcount repeated list entries instead of always adding to
1058 Severity : enhancement
1060 Description: Dynamic service threads
1061 Details : Within a small range, start extra service threads
1062 automatically when the request queue builds up.
1065 Frequency : mixed-endian client/server environments
1067 Description: mixed-endian crashes
1068 Details : The new msg_v2 system had some failures in mixed-endian
1071 Severity : enhancement
1073 Description: Easy OST removal
1074 Details : OSTs can be permanently deactivated with e.g. 'lctl
1075 conf_param lustre-OST0001.osc.active=0'
1077 Severity : enhancement
1079 Description: MGS proc entries
1080 Details : Added basic proc entries for the MGS showing what filesystems
1083 Severity : enhancement
1085 Description: provide MGS failover
1086 Details : Added config lock reacquisition after MGS server failover.
1088 Severity : enhancement
1090 Description: add Linux 2.4 support
1091 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1095 Description: a large application tries to do I/O to the same resource and dies
1096 in the middle of it.
1097 Details : Check the req->rq_arrival time after the call to
1098 ost_brw_lock_get(), but before we do anything about
1099 processing it & sending the BULK transfer request. This
1100 should help move old stale pending locks off the queue as
1101 quickly as obd_timeout.
1104 Frequency : when an incorrect nid is specified during startup
1106 Description: ptlrpc connect to non-existant node causes kernel crash
1107 Details : LNET can't be re-entered from an event callback, which
1108 happened when we expire a message after the export has been
1109 cleaned up. Instead, hand the zombie cleanup off to another
1112 Severity : enhancement
1114 Description: plain/inodebits lock performance improvement
1115 Details : Grouping plain/inodebits in granted list by their request modes
1116 and bits policy, thus improving the performance of search through
1120 Frequency : only if OST filesystem is corrupted
1122 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1123 Details : for a short time RPCs with bulk IO are in the replay list,
1124 but replay of bulk IOs is unimplemented. If the OST filesystem
1125 is corrupted due to disk cache incoherency and then replay is
1126 started it is possible to trip an assertion. Avoid putting
1127 committed RPCs into the replay list at all to avoid this issue.
1130 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1132 Description: System hang on startup
1133 Details : This bug allowed the liblustre (e.g. catamount) client to
1134 return to the app before handling all startup RPCs. This
1135 could leave the node unresponsive to lustre network traffic
1136 and manifested as a server ptllnd timeout.
1138 Severity : enhancement
1140 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1141 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1142 boolean to control whether to print peer debug info when a
1143 client's RPC times out.
1146 Frequency : only for kernels with patches from Lustre below 1.4.3
1148 Description: Remove old rdonly API
1149 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1152 Frequency : only for devices with external journals
1154 Description: Set external device read-only also
1155 Details : During a commanded failover stop, we set the disk device
1156 read-only while the server shuts down. We now also set any
1157 external journal device read-only at the same time.
1160 Frequency : when upgrading from 1.4 while trying to change parameters
1162 Description: The wrong (new) MDC name was used when setting parameters for
1163 upgraded MDT's. Also allows changing of OSC (and MDC)
1164 parameters if --writeconf is specified at tunefs upgrade time.
1167 Frequency : when setting specific ost indicies
1169 Description: QOS code breaks on skipped indicies
1170 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1171 created with --index need not be sequential.
1173 Severity : enhancement
1175 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1176 Details : The uninit_groups feature works in conjunction with the kernel
1177 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1178 up the pass1 processing of e2fsck. This is a read-only feature
1179 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1180 mount filesystems that have had this feature enabled.
1182 Severity : enhancement
1184 Description: Improve multi-block allocation algorithm to avoid fragmentation
1185 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1186 allocation locality and avoid filesystem fragmentation.
1188 ------------------------------------------------------------------------------
1190 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1192 * Support for kernels:
1193 2.4.21-47.0.1.EL (RHEL 3)
1194 2.6.5-7.283 (SLES 9)
1195 2.6.9-42.0.10.EL (RHEL 4)
1196 2.6.12.6 vanilla (kernel.org)
1197 2.6.16.27-0.9 (SLES 10)
1198 * Recommended e2fsprogs version: 1.39.cfs5
1200 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1204 Frequency : occasional, depends on client load and configuration
1205 Bugzilla : 12181, 12203
1206 Description: data loss for recently-modified files
1208 Details : In some cases it is possible that recently written or created
1209 files may not be written to disk in a timely manner (this should
1210 normally be within 30s unless client IO load is very high).
1211 The problem appears as zero-length files or files that are a
1212 multiple of 1MB in size after a client crash or client eviction
1213 that are missing data at the end of the file.
1215 This problem is more likely to be hit on clients where files are
1216 repeatedly created and unlinked in the same directory, clients
1217 have a large amount of RAM, have many CPUs, the filesystem has
1218 many OSTs, the clients are rebooted frequently, and/or the files
1219 are not accessed by other nodes after being written.
1221 The presence of the problem can be detected by looking at
1222 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1223 smaller than the second (nr_unused) then dirty files will not
1224 be flushed automatically to disk. "sync; sleep 10" should be
1225 run several times on the node before unmounting it to update
1226 Lustre (this is also safe to run on nodes without this problem).
1228 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1229 that can cause this same problem, so customers using that kernel
1230 also need to update the kernel in addition to Lustre. In order
1231 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1233 It is normal that files written just before a client crash (less
1234 than 30s) may not yet have been flushed to disk, even for local
1238 Frequency : frequent on thin XT3 nodes
1240 Description: UUID collision on thin XT3 Linux nodes
1241 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1242 randomly, since we relied on an insufficiently-seeded PRNG.
1247 Description: OSS hangs after "All ost request buffers busy"
1248 Details : A deadlock between quota and journal operations caused OSS
1249 hangs after printing "All ost request buffers busy."
1252 Frequency : always on liblustre builds
1254 Description: Cleanup compiler warnings on liblustre
1257 Frequency : always on liblustre builds on XT3
1259 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1263 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1264 > 32000 subdirectories
1265 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1266 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1267 subdirectory count crosses 32000. This will aid e2fsck to
1268 correctly handle more than 32000 subdirectories.
1271 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1273 Description: System hang on startup
1274 Details : This bug allowed the liblustre (e.g. catamount) client to
1275 return to the app before handling all startup RPCs. This
1276 could leave the node unresponsive to lustre network traffic
1277 and manifested as a server ptllnd timeout.
1279 Severity : enhancement
1281 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1282 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1283 boolean to control whether to print peer debug info when a
1284 client's RPC times out.
1289 Description: make O_SYNC working on 2.6 kernels
1290 Details : 2.6 kernels use different method for mark pages for write,
1291 so need add a code to lustre for O_SYNC work.
1296 Description: Failure to close file and release space on NFS
1297 Details : Put inode details into lock acquired in ll_intent_file_open.
1298 Use mdc_intent_lock in ll_intent_open to properly
1299 detect all kind of errors unhandled by mdc_enqueue.
1304 Description: proc file read during shutdown sometimes raced obd removal,
1306 Details : Add lock to prevent obd access after proc file removal.
1309 Frequency : Only for files larger than 4GB on 32-bit clients.
1311 Description: improperly doing page alignment of locks
1312 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1313 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1318 Description: RPCs being resent when they shouldn't be
1319 Details : Some RPCs that should not be resent are being resent. This
1320 can cause inconsistencies in the RPC state machine. Do not
1321 resend such requests.
1324 Frequency : rare, only with NFS export
1326 Description: Crash on NFS re-export node
1327 Details : under very unusual load conditions an assertion is hit in
1328 ll_intent_file_open()
1331 Frequency : only if OST filesystem is corrupted
1333 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1334 Details : for a short time RPCs with bulk IO are in the replay list,
1335 but replay of bulk IOs is unimplemented. If the OST filesystem
1336 is corrupted due to disk cache incoherency and then replay is
1337 started it is possible to trip an assertion. Avoid putting
1338 committed RPCs into the replay list at all to avoid this issue.
1343 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1344 Details : Large single O_DIRECT read and write calls can fail to allocate
1345 a sufficiently large buffer to process the request. In case of
1346 allocation failure the allocation is retried with a smaller
1347 buffer and broken into smaller requests.
1349 Severity : enhancement
1351 Description: Add -o localflock option to simulate old noflock behaviour.
1352 Details : This will achieve local-only flock/fcntl locks coherentness.
1357 Description: versioning check is incomplete
1358 Details : Checking the version difference of client vs. server, report
1359 error if the gap is too big.
1364 Description: add support PG_writeback bit
1365 Details : add support for PG_writeback bit for Lustre, for more carefull
1366 work with page cache in 2.6 kernel. This also fix some deadlocks
1367 and remove hack for work O_SYNC with 2.6 kernel.
1369 Severity : enhancement
1371 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1372 Details : The uninit_groups feature works in conjunction with the kernel
1373 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1374 up the pass1 processing of e2fsck. This is a read-only feature
1375 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1376 mount filesystems that have had this feature enabled.
1378 Severity : enhancement
1380 Description: Improve multi-block allocation algorithm to avoid fragmentation
1381 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1382 allocation locality and avoid filesystem fragmentation.
1384 ------------------------------------------------------------------------------
1386 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1388 * Support for kernels:
1389 2.6.9-42.0.3.EL (RHEL 4)
1390 2.6.5-7.276 (SLES 9)
1391 2.4.21-47.0.1.EL (RHEL 3)
1392 2.6.12.6 vanilla (kernel.org)
1393 2.6.16.21-0.8 (SLES10)
1394 * Recommended e2fsprogs version: 1.39.cfs2-0
1396 * The backwards-compatible /proc/sys/portals symlink has been removed
1397 in this release. Before upgrading, please ensure that you change
1398 any configuration scripts or /etc/sysctl.conf files that access
1399 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1400 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1401 in advance of the upgrade on any system running Lustre 1.4.6 or
1402 newer, since /proc/sys/lnet was added in that version.
1403 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1407 Frequency : only when quota is used
1409 Description: avoid scanning export list for quota master
1410 Details : Change the algorithms to avoid scanning export list in order
1411 to improve the efficiency.
1414 Frequency : MDS failover only, very rarely
1416 Description: "went back in time" messages on mds failover
1417 Details : The greatest transno may be lost when the current operation
1418 finishes with an error (transno==0) and the client's last_rcvd
1419 record is over-written. Save the greatest transno in the
1420 mds_last_transno for this case.
1423 Frequency : always for specific kernels and striping counts
1425 Description: client may get "Matching packet too big" without ACL support
1426 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1427 when trying to access files in certain configurations. The
1428 clients should in fact be denied when mounting because they do
1429 not understand ACLs.
1432 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1434 Description: many clients connecting with IO in progress causes connect timeouts
1435 Details : Avoid synchronous journal commits to avoid delays caused by many
1436 clients connecting/disconnecting when bulk IO is in progress.
1437 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1438 OST_IO_PORTAL to avoid delays behind potentially many pending
1442 Frequency : occasionally with multiple writers to a single file
1444 Description: shared writes to file may result in wrong size reported by stat()
1445 Details : Allow growing of kms when extent lock is cancelled
1448 Frequency : always with random mmap IO to multi-striped file
1450 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1451 Details : Only if the hole is at the end of OST object so that kms is too
1452 small. Fix is to increase kms accordingly in ll_nopage.
1455 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1457 Description: writes to a missing object would leak memory on the OST
1458 Details : If there is an inconsistency between the MDS and OST filesystems,
1459 such that the MDS references an object that doesn't exist, writes
1460 to that object will leak memory due to incorrect cleanup in the
1461 error handling path, eventually running out of memory on the OST.
1466 Description: Creating too long symlink causes lustre errors
1467 Details : Check symlink and name lengths before sending requests to MDS.
1470 Frequency : only if flock is enabled (not on by default)
1472 Description: posix locks not released on fd closure on 2.6.9+
1473 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1474 kernels, this caused such locks not to be released on fd close and
1475 then assertions on fs unmount about still used locks.
1478 Frequency : MDS failover only, very rarely
1480 Description: clients may get ASSERTION(granted_lock != NULL)
1481 Details : When request was taking a long time, and a client was resending
1482 a getattr by name lock request. The were multiple lock requests
1483 with the same client lock handle and
1484 mds_getattr_name->fixup_handle_for_resent_request found one of the
1485 lock handles but later failed with ASSERTION(granted_lock != NULL).
1490 Description: handle->h_buffer_credits > 0, assertion failure
1491 Details : h_buffer_credits is zero after truncate, causing assertion
1492 failure. This patch extends the transaction or creates a new
1496 Frequency : NFS re-export or patchless client
1497 Bugzilla : 11179, 10796
1498 Description: Crash on NFS re-export node (__d_move)
1499 Details : We do not want to hash the dentry if we don't have a lock.
1500 But if this dentry is later used in d_move, we'd hit uninitialised
1501 list head d_hash, so we just do this to init d_hash field but
1502 leave dentry unhashed.
1505 Frequency : NFS re-export or patchless client
1507 Description: NFS exports has problem with symbolic link
1508 Details : lustre client didn't properly install dentry when re-exported
1509 to NFS or running patchless client.
1512 Frequency : NFS re-export or patchless client
1514 Description: Various nfs/patchless fixes.
1515 Details : fixes reuse disconected alias for lookup process - this fixes
1516 warning "find_exported_dentry: npd != pd",
1517 fix permission error with open files at nfs.
1518 fix apply umask when do revalidate.
1521 Frequency : occasional
1523 Description: Crash on NFS re-export node
1524 Details : calling clear_page() on the wrong pointer triggered oops in
1525 generic_mapping_read().
1528 Frequency : rarely, using O_DIRECT IO
1530 Description: unaligned directio crashes client with LASSERT
1531 Details : check for unaligned buffers before trying any requests.
1534 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1536 Description: stale data returned from RAID cache
1537 Details : If only a small amount of IO is done to the RAID device before
1538 reading it again it is possible to get stale data from the RAID
1539 cache instead of reading it from disk.
1542 Frequency : always for sles10 kernel
1544 Description: sles10 support
1545 Details : ll_follow_link: compile fixes and using of nd_set_link
1546 under newer kernels.
1549 Frequency : depends on arch, kernel and compiler version, always on sles10
1552 Description: recursive or deep enough symlinks cause stack overflow
1553 Details : getting rid of large stack-allocated variable in
1557 Frequency : depends on hardware
1559 Description: lustre write performance loss in the SLES10 kernel
1560 Details : the performance loss is caused by using of write barriers in the
1561 ext3 code. The SLES10 kernel turns barrier support on by
1562 default. The fix is to undo that change for ldiskfs.
1564 ------------------------------------------------------------------------------
1566 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1568 * Support for kernels:
1569 2.6.9-42.0.3EL (RHEL 4)
1570 2.6.5-7.276 (SLES 9)
1571 2.4.21-47.0.1.EL (RHEL 3)
1572 2.6.12.6 vanilla (kernel.org)
1576 Frequency : quota enabled and large files being deleted
1578 Description: releasing more than 4GB of quota at once hangs OST
1579 Details : If a user deletes more than 4GB of files on a single OST it
1580 will cause the OST to spin in an infinite loop. Release
1581 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1586 Description: statfs data retrieved from /proc may be stale or zero
1587 Details : When reading per-device statfs data from /proc, in the
1588 {kbytes,files}_{total,free,avail} files, it may appear
1589 as zero or be out of date.
1592 Frequency : systems with MD RAID1 external journal devices
1594 Description: lconf's call to blkid is confused by RAID1 journal devices
1595 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1596 of returning all block devices that match the journal UUID.
1599 Frequency : always, for aggregate stripe size over 4GB
1601 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1602 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1603 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1606 Frequency : always if "lfs find" used on a local file/directory
1608 Description: "lfs find" segfaults if used on a local file/directory
1609 Details : The case where a directory component was not specified wasn't
1610 handled correctly. Handle this properly.
1613 Frequency : always on ppc64
1615 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1616 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1617 perform endianness conversion.
1620 Frequency : rarely (truncate to non-zero file size after write under load)
1621 Bugzilla : 10730, 10687
1622 Description: Files padded with zeros to next 4K multiple
1623 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1624 it is possible that files that are truncated to a non-zero size
1625 immediately after being written are filled with zero bytes beyond
1626 the truncated size. No file data is lost.
1628 Severity : enhancement
1630 Description: Allow recovery/failover for liblustre clients.
1631 Details : liblustre clients were unaware of failover configurations until
1634 Severity : enhancement
1636 Description: user file locks should fail when not mounting with flock option
1637 Details : Set up an error-returning stub in ll_file_operations.lock field
1638 to prevent incorrect behaviour when client is mounted without
1639 flock option. Also, set up properly f_op->flock field for
1643 Frequency : always on ia64
1645 Description: "lfs df" loops on printing out MDS statfs information
1646 Details : The obd_ioctl_data was not initialized and in some systems
1647 this caused a failure during the ioctl that did not return
1648 an error. Initialize the struct and return an error on failure.
1651 Frequency : SLES 9 only
1653 Description: Error of copying files with lustre special EAs as root
1654 Details : Client side always return success for setxattr call for lustre
1655 special xattr (currently only "trusted.lov").
1658 Frequency : rarely on clusters with both ia64+i386 clients
1660 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1661 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1662 clients are doing concurrent writes to the same file it is
1663 possible that the ia64 clients may LASSERT because the OST
1664 extent locks are not PAGE_SIZE aligned. Ensure that grown
1665 locks are always aligned on the request boundary.
1668 Frequency : specific use, occasional
1670 Description: Overwriting in use executable truncates on-disk binary image
1671 Details : If one node attempts to overwrite an executable in use by
1672 another node, we now correctly return ETXTBSY instead of
1673 truncating the file.
1675 Severity : enhancement
1677 Description: Async OSC create to avoid the blocking unnecessarily.
1678 Details : If a OST has no remain object, system will block on the creating
1679 when need to create a new object on this OST. Now, ways use
1680 pre-created objects when available, instead of blocking on an
1681 empty osc while others are not empty. If we must block, we block
1682 for the shortest possible period of time.
1687 Description: chmod on Lustre root is propagated to other clients
1688 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1689 invalid by the follow_mount time.
1692 Frequency : liblustre clients only
1694 Description: Race in 'instant cancel' lock handling could lead to such locks
1695 never to be granted in case of SMP MDS
1696 Details : Do not destroy not yet granted but cbpending locks in
1700 Frequency : replay/resend of open
1702 Description: non null lock assetion failure in mds_intent_policy
1703 Details : Trying to replay/resend lockless open requests resulted in
1704 mds_open() returning 0 with no lock. Now it sets a flag if
1705 a lock is going to be returned.
1707 Severity : enhancement
1709 Description: Checksum enhancements
1710 Details : New checksum enhancements allow for resending RPCs that failed
1713 Severity : enhancement
1715 Description: Tunables on number of dirty pages in cacche
1716 Details : Allow to set limit on number of dirty pages cached.
1721 Description: client crash on unmount - lock still has references
1722 Details : In some error handling cases it was possible to leak a lock
1723 reference on a client while accessing a file. This was not
1724 harmful to the client during operation, but would cause the
1725 client to crash when the filesystem is unmounted.
1728 Frequency : specific case, rare
1730 Description: ETXTBSY on mds though file not in use by client
1731 Details : ETXTBSY is no longer incorrectly returned when attempting to
1732 chmod or chown a directory that the user previously tried to
1733 execute or a currently-executing binary.
1736 Frequency : extremely rare except on liblustre-based clients
1738 Description: Lustre space not freed when files are deleted
1739 Details : Clean up open-unlinked files after client eviction. Previously
1740 the unlink was skipped and the files remained as orphans.
1745 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1746 Details : In some cases it was possible to send a blocking callback to
1747 a client doing a glimpse, even though that client didn't get
1748 a lock granted. When the glimpse lock is cancelled on the OST
1749 the freed lock is left on the waiting list and corrupted the list.
1752 Frequency : all core dumps
1754 Description: Broke core dumps to lustre
1755 Details : Negative dentry may be unhashed if parent does not have UPDATE
1756 lock, but some callers, e.g. do_coredump, expect dentry to be
1757 hashed after successful create, hash it in ll_create_it.
1759 ------------------------------------------------------------------------------
1761 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1763 * Support for kernels:
1764 2.6.9-42.0.2.EL (RHEL 4)
1765 2.6.5-7.276 (SLES 9)
1766 2.4.21-40.EL (RHEL 3)
1767 2.6.12.6 vanilla (kernel.org)
1771 Frequency : always on RHEL 3
1773 Description: Number of open files grows over time
1774 Details : The number of open files grows over time, whether or not
1775 Lustre is started. This was due to a filp leak introduced
1776 by one of our kernel patches.
1778 ------------------------------------------------------------------------------
1780 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1782 * Support for kernels:
1783 2.6.9-42.EL (RHEL 4)
1784 2.6.5-7.276 (SLES 9)
1785 2.4.21-40.EL (RHEL 3)
1786 2.6.12.6 vanilla (kernel.org)
1791 Bugzilla : 5719, 9635, 9792, 9684
1792 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1793 Details : If a server is under heavy load and cannot reply to new
1794 connection requests before the client resends the (re)connect,
1795 the connection handling code can behave badly if two service
1796 threads are concurrently handing separate (re)connections from
1797 the same client. Add better locking to the connection handling
1798 code, and ensure that only a single connection will be processed
1799 for a given client UUID, even if the lock is dropped.
1801 Severity : enhancement
1803 Description: add TCP zero-copy support to kernel
1804 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1805 sends if the hardware supports scatter-gather and checksumming.
1806 This allows socklnd to do client-write and server-read more
1807 efficiently and reduce CPU utilization from skbuf copying.
1810 Frequency : only if NFS exporting from client
1812 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1813 Details : If setting attributes on a file created under NFS that had
1814 never been opened it would be possible to oops the client
1815 if the file had no objects.
1818 Frequency : always for liblustre
1820 Description: liblustre client does MDS+OSTs setattr RPC for each write
1821 Details : When doing a write from a liblustre client, the client
1822 incorrectly issued an RPC to the MDS and each OST the file was
1823 striped over in order to update the timestamps. When writing
1824 with small chunks and many clients this could overwhelm the MDS
1825 with RPCs. In all cases it would slow down the write because
1826 these RPCs are unnecessary.
1828 Severity : enhancement
1830 Description: allow number of MDS service threads to be changed at module load
1831 Details : It is now possible to change the number of MDS service threads
1832 running. Adding "options mds mds_num_threads={N}" to the MDS's
1833 /etc/modprobe.conf will set the number of threads for the next
1834 time Lustre is restarted (assuming the "mds" module is also
1835 reloaded at that time). The default number of threads will
1836 stay the same, 32 for most systems.
1841 Description: OST crash if filesystem is unformatted or corrupt
1842 Details : If an OST is started on a device that has never been formatted
1843 or if the filesystem is corrupt and cannot even mount then the
1844 error handling cleanup routines would dereference a NULL pointer.
1849 Description: NULL pointer deref in llap_from_page.
1850 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1851 incorrect) mapping if the page was truncated/reclaimed while it was
1852 searched for. Check for this condition and skip such pages when
1853 doing readahead. Introduce extra check to llap_from_page() to
1854 verify page->mapping->host is non-NULL (so page is not anonymous).
1857 Frequency : Sometimes when using sys_sendfile
1859 Description: "page not covered by a lock" warnings from ll_readpage
1860 Details : sendfile called ll_readpage without right page locks present.
1861 Now we introduced ll_file_sendfile that does necessary locking
1862 around call to generic_file_sendfile() much like we do in
1866 Frequency : with certain MDS communication failures at client mount time
1868 Description: NULL pointer deref after failed client mount
1869 Details : a client connection request may delayed by the network layer
1870 and not be sent until after the PTLRPC layer has timed out the
1871 request. If the client fails the mount immediately it will try
1872 to clean up before the network times out the request. Add a
1873 reference from the request import to the obd device and delay
1874 the cleanup until the network drops the request.
1877 Frequency : occasionally during client (re)connect
1879 Description: assertion failure during client (re)connect
1880 Details : processing a client connection request may be delayed by the
1881 client or server longer than the client connect timeout. This
1882 causes the client to resend the connection request. If the
1883 original connection request is replied in this interval, the
1884 client may trip an assertion failure in ptlrpc_connect_interpret()
1885 which thought it would be the only running connect process.
1888 Frequency : only with obd_echo servers and clients that are rebooted
1890 Description: kernel BUG accessing uninitialized data structure
1891 Details : When running an obd_echo server it did not start the ping_evictor
1892 thread, and when a client was evicted an uninitialized data
1893 structure was accessed. Start the ping_evictor in the RPC
1894 service startup instead of the OBD startup.
1896 Severity : enhancement
1897 Bugzilla : 10193 (patchless)
1898 Description: Remove dependency on various unexported kernel interfaces.
1899 Details : No longer need reparent_to_init, exit_mm, exit_files,
1900 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1903 Frequency : rare (only users of deprecated and unsupported LDAP config)
1905 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1906 Details : LDAP apparently contains 'lustreName' attributes instead of
1907 'name'. A simple remapping of the name is sufficient.
1910 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1912 Description: waiting_locks_callback trips kernel BUG if client is evicted
1913 Details : Running with the dump_on_timeout debug flag turned on makes
1914 it possible that the waiting_locks_callback() can try to dump
1915 the Lustre kernel debug logs from an interrupt handler. Defer
1916 this log dumping to the expired_lock_main() thread.
1918 Severity : enhancement
1920 Description: Support NFS exporting on 2.6 kernels.
1921 Details : Implement non-rawops metadata methods for NFS server to use without
1922 changing NFS server code.
1925 Frequency : very rare (synthetic metadata workload only)
1927 Description: two racing renames might cause an MDS thread to deadlock
1928 Details : Running the "racer" program may cause one MDS thread to rename
1929 a file from being the source of a rename to being the target of
1930 a rename at exactly the same time that another thread is doing
1931 so, and the second thread has already enqueued these locks after
1932 doing a lookup of the target and is trying to relock them in
1933 order. Ensure that we don't try to re-lock the same resource.
1936 Frequency : only very large systems with liblustre clients
1938 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1939 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1940 all OSTs in parallel. This allows the eviction of stale liblustre
1941 clients to proceed much faster than if they were done in series,
1942 and also offers similar improvements for other set_info RPCs.
1947 Description: excessive CPU usage during initial read phase on client
1948 Details : During the initial read phase on a client, it would agressively
1949 retry readahead on the file, consuming too much CPU and impacting
1950 performance (since 1.4.5.8). Improve the readahead algorithm
1951 to avoid this, and also improve some other common cases (read
1952 of small files in particular, where "small" is files smaller than
1953 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1958 Description: MDS crash when receiving packet with unknown intent.
1959 Details : Do not LBUG in unknown intent case, just return -EFAULT
1961 Severity : enhancement
1962 Bugzilla : 9293, 9385
1963 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1964 Details : Do not serialize getattr (non-intent version) and statfs.
1967 Frequency : occasional, when OST network is overloaded/intermittent
1969 Description: client evicted by OST after bulk IO timeout
1970 Details : If a client sends a bulk IO request (read or write) the OST
1971 may evict the client if it is unresposive to its data GET/PUT
1972 request. This is incorrect if the network is overloaded (takes
1973 too long to transfer the RPC data) or dropped the OST GET/PUT
1974 request. There is no need to evict the client at all, since
1975 the pinger and/or lock callbacks will handle this, and the
1976 client can restart the bulk request.
1979 Frequency : Always when mmapping file with no objects
1981 Description: client crashes when mmapping file with no objects
1982 Details : Check that we actually have objects in a file before doing any
1983 operations on objects in ll_vm_open, ll_vm_close and
1989 Description: Request leak when working with deleted CWD
1990 Details : Introduce advanced request refcount tracking for requests
1991 referenced from lustre intent.
1993 Severity : Enhancement
1995 Description: Cache open file handles on client.
1996 Details : MDS now will return special lock along with openhandle, if
1997 requested and client is allowed to hold openhandle, even if unused,
1998 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1999 closing files for every read/write openration.
2001 Severity : Enhancement
2003 Description: Cache open negative dentries on client when possible.
2004 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2005 negative dentries on lock revocation.
2010 Description: Remounting a client read-only wasn't possible with a zconf mount
2011 Details : It wasn't possible to remount a client read-only with llmount.
2013 Severity : enhancement
2014 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2015 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2016 MPICH is included to add Lustre-specific ADIO interfaces.
2017 This is based closely on the UFS ADIO layer and only differs
2018 in file creation, in order to allow the OST striping to be set.
2019 This is user-contributed code and not supported by CFS.
2024 Description: extended inode attributes (immutable, append-only) work improperly
2025 when 2.4 and 2.6 kernels are used on client/server or vice versa
2026 Details : Introduce kernel-independent values for these flags.
2028 Severity : enhancement
2031 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2032 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2033 granularity. Now, Fractional MB granularity is be supported,
2034 this is very useful for low memory system.
2036 Severity : enhancement
2038 Description: Getattr by fid
2039 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2040 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2041 have locks and avoids getattr rpc requests for it on every stat.
2044 Frequency : Always, for filesystems larger than 2TB
2046 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2047 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2048 filesystems larger than 2TB to be mounted with mballoc enabled.
2051 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2053 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2054 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2055 filesystems larger than 2TB is mounted then the kernel will
2056 silently corrupt the start of the filesystem. CONFIG_LBD is
2057 enabled for all CFS-supported kernels, but the possibility of
2058 this happening with a modified kernel config exists.
2060 Severity : enhancement
2062 Description: add client O_DIRECT support for 2.6 kernels
2063 Details : It is now possible to do O_DIRECT reads and writes to files
2064 in the Lustre client mountpoint on 2.6 kernel clients.
2066 Severity : enhancement
2068 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2069 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2070 parallel, not waiting for response from every OST before sending
2071 a rpc to the next OST.
2076 Description: setattr vs write race when updating file timestamps
2077 Details : Client processes that update a file timestamp into the past
2078 right after writing to the file (e.g. tar) it is possible that
2079 the updated file modification time can be reset to the current
2080 time due to a race between processing the setattr and write RPC.
2082 Severity : enhancement
2084 Description: Bring 'lfs find' closer in line with regular Linux find.
2085 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2086 -print0 options and obtains all the needed info through the lustre
2089 Severity : enhancement
2091 Description: support up to 1024 configured devices on one node
2092 Details : change obd_dev array from statically allocated to dynamically
2093 allocated structs as they are first used to reduce memory usage
2098 Description: Flush dirty partially truncated pages during truncate
2099 Details : Immediatelly flush partially truncated pages in filter_setattr,
2100 this way we completely avoid having any pages in page cache on OST
2101 and can retire ugly workarounds during writes to flush such pages.
2106 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2107 Details : protect inode from truncation within vfs_unlink() context
2108 just take a reference before calling vfs_unlink() and release it
2109 when parent's i_sem is free.
2114 Description: last_id value checked outside lock on OST caused LASSERT failure
2115 Details : If there were multiple MDS->OST object precreate requests in
2116 flight, it was possible that the OST's last object id was checked
2117 outside a lock and incorrectly tripped an assertion. Move checks
2118 inside locks, and discard old precreate requests.
2121 Frequency : always, if extents are used on OSTs
2123 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2124 Details : OSTs using the extents format would not zero the high 16 bits of
2125 the index physical block number. This is not a problem for any
2126 OST filesystems smaller than 16TB, and no kernels support ext3
2127 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2128 new/modified files) and can be fixed for existing filesystems
2129 with e2fsprogs-1.39-cfs1.
2134 Description: import connection selection may be incorrect if timer wraps
2135 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2136 import connections to be ignored if the 32-bit jiffies counter
2137 wraps. Use a 64-bit jiffies counter.
2140 Frequency : very large clusters immediately after boot
2142 Description: LNET request buffers exhausted under heavy short-term load
2143 Details : If a large number of client requests are generated on a service
2144 that has previously never seen so many requests it is possible
2145 that the request buffer growth cannot keep up with the spike in
2146 demand. Instead of dropping incoming requests, they are held in
2147 the LND until the RPC service can accept more requests.
2150 Frequency : Sometimes during replay
2152 Description: Assertion failure in ll_local_open after replay.
2153 Details : If replay happened on an open request reply before we were able
2154 to set replay handler, reply will become not swabbed tripping the
2155 assertion in ll_local_open. Now we set the handler right after
2156 recognising of open request
2159 Frequency : very rare
2161 Description: kernel reports "badness in vsnprintf"
2162 Details : Reading from the "recovery_status" /proc file in small chunks
2163 may cause a negative length in lprocfs_obd_rd_recovery_status()
2164 call to vsnprintf() (which is otherwise harmless). Exit early
2165 if there is no more space in the output buffer.
2167 Severity : enhancement
2169 Description: clear OBD RPC statistics by writing to them
2170 Details : It is now possible to clear the OBD RPC statistics by writing
2171 to the "stats" file.
2176 Description: Client mtime is not the same on different clients after utimes
2177 Details : In some cases, the client was using the utimes() syscall on
2178 a file cached on another node. The clients now validate the
2179 ctime from the MDS + OSTs to determine which one is right.
2184 Description: Inability to activate failout mode
2185 Details : lconf script incorrectly assumed that in pythong string's numeric
2186 value is used in comparisons.
2189 Frequency : always with multiple stripes per file
2191 Description: Inefficient object allocation for mutli-stripe files
2192 Details : When selecting which OSTs to stripe files over, for files with
2193 a stripe count that divides evenly into the number of OSTs,
2194 the MDS is always picking the same starting OST for each file.
2195 Return the OST selection heuristic to the original design.
2200 Description: mount failures may take full timeout to return an error
2201 Details : Under some heavy load conditions it is possible that a
2202 failed mount can wait for the full obd_timeout interval,
2203 possibly several minutes, before reporting an error.
2204 Instead return an error as soon as the status is known.
2206 Frequency : quota enabled and large files being deleted
2208 Description: releasing more than 4GB of quota at once hangs OST
2209 Details : If a user deletes more than 4GB of files on a single OST it
2210 will cause the OST to spin in an infinite loop. Release
2211 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2216 Description: statfs data retrieved from /proc may be stale or zero
2217 Details : When reading per-device statfs data from /proc, in the
2218 {kbytes,files}_{total,free,avail} files, it may appear
2219 as zero or be out of date.
2222 Frequency : systems with MD RAID1 external journal devices
2224 Description: lconf's call to blkid is confused by RAID1 journal devices
2225 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2226 of returning all block devices that match the journal UUID.
2229 Frequency : always, for aggregate stripe size over 4GB
2231 Description: assertion fails when trying to use 4GB stripe size
2232 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2233 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2236 Frequency : always on ppc64
2238 Description: the first write on an ext3 filesystem with mballoc got stuck
2239 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2240 perform endianness conversion.
2242 ------------------------------------------------------------------------------
2244 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2246 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2247 INTEROPERATE with older versions automatically. Please read the
2248 user documentation before upgrading any part of a live system.
2249 * WARNING: Lustre networking configuration changes are required with
2250 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2253 * Support for kernels:
2254 2.6.9-22.0.2.EL (RHEL 4)
2255 2.6.5-7.244 (SLES 9)
2256 2.6.12.6 vanilla (kernel.org)
2259 Severity : enhancement
2260 Bugzilla : 7981/8208
2261 Description: Introduced Lustre Networking (LNET)
2262 Details : LNET is new networking infrastructure for Lustre, it includes
2263 a reorganized network configuration mode (see the user
2264 documentation for full details) as well as support for routing
2265 between different network fabrics. Lustre Networking Devices
2266 (LNDs) for the supported network fabrics have also been
2267 created for this new infrastructure.
2269 Severity : enhancement
2270 Description: Introduced Access control lists
2271 Details : clients can set ACLs on files and directories in order to have
2272 more fine-grained permissions than the standard Unix UGO+RWX.
2273 The MDS must be started with the "-o acl" mount option.
2275 Severity : enhancement
2276 Description: Introduced filesystem quotas
2277 Details : Administrators may now establish per-user quotas on the
2280 Severity : enhancement
2282 Description: Configuration change for the XT3
2283 The PTLLND is now used to run Lustre over Portals on the XT3
2284 The configure option(s) --with-cray-portals are no longer used.
2285 Rather --with-portals=<path-to-portals-includes> is used to
2286 enable building on the XT3. In addition to enable XT3 specific
2287 features the option --enable-cray-xt3 must be used.
2292 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2293 Details : A race between allocating a new llcd and re-getting the llcd_lock
2294 allowed another thread to grab newly-allocated llcd.
2296 Severity : enhancement
2298 Description: 2.6 OST async journal commit and locking fix to improve performance
2299 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2300 2.6 kernels are now async as they already were in 2.4 kernels so
2301 that they can commit concurrently with the network bulk transfer.
2302 For block-allocated files the filter allocation semaphore is held
2303 to avoid filesystem fragmentation during allocation. BKL lock
2304 removed for 2.6 xattr operations where it is no longer needed.
2309 Description: lconf incorrectly determined whether two IP networks could talk
2310 Details : In some more complicated routing and multiple-network
2311 configurations, lconf will avoid trying to make a network
2312 connection to a disjoint part of the IP space. It was doing the
2313 math incorrectly for one set of cases.
2318 Description: Fix for potential infinite loop processing records in an llog.
2319 Details : If an llog record is corrupted/zeroed, it is possible to loop
2320 forever in llog_process(). Validate the llog record length
2321 and skip the remainder of the block on error.
2324 Frequency : occasional (liblustre only)
2326 Description: liblustre could not open files whose last component is a symlink
2327 Details : sysio_path_walk() would incorrectly pass the open intent to
2328 intermediate path components.
2331 Frequency : rare (liblustre only with non-standard tuning)
2332 Bugzilla : 7201 (7350)
2333 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2334 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2335 after it has been released, only do this for referenced locks
2337 Severity : enhancement
2339 Description: specifying an (invalid) directory default stripe_size of -1
2340 would reset the directory default striping
2341 Details : stripe_size -1 was used internally to signal directory stripe
2342 removal, now use "all default" to signal dir stripe removal
2343 as a directory striping of "all default" is not useful
2346 Frequency : common for large clusters running liblustre clients
2348 Description: doing an ls when liblustre clients are running is slow
2349 Details : sending a glimpse AST to a liblustre client waits for every AST
2350 to time out, as liblustre clients will not respond. Since they
2351 cannot cache data we refresh the OST lock LVB from disk instead.
2353 Severity : enhancement
2355 Description: doing an ls at the same time as file IO can be slow
2356 Details : enqueue and other "small" requests can be blocked behind many
2357 large IO requests. Create a new OST IO portal for non-IO
2358 requests so they can be processed faster.
2361 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2363 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2364 created file with mode 000
2365 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2366 allow an ftruncate() as a non-root user to a file with mode 000.
2367 Lustre now respects this flag to disable mode checks when
2368 truncating a file owned by the user
2371 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2374 Description: Revoking locks from clients that went dead or catatonic might take
2376 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2377 cancellation of such locks instant on servers without waiting for
2378 any reply from clients. Clients drops these locks when cancel
2379 notification from server is received without replying.
2382 Frequency : liblustre-only, when liblustre client dies or becomes busy
2384 Description: Doing ls on Linux clients can take a long time with active
2386 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2387 granting such locks to it in the first place if possible. Locks
2388 are taken by proxy on the OST during the read or write and
2389 dropped immediately afterward. Add connect flags handling, do
2390 not grant locks to liblustre clients for glimpse ASTs.
2392 Severity : enhancement
2394 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2395 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2396 sequential reads and behaves badly with applications doing random
2397 reads. Improve it by reading ahead at least read region, and
2398 avoiding excessive large RPC for small reads.
2400 Severity : enhancement
2402 Description: Creating more than 1000 files for a single job may cause a load
2403 imbalance on the OSTs if there are also a large number of OSTs.
2404 Details : qos_prep_create() uses an OST index reseed value that is an
2405 even multiple of the number of available OSTs so that if the
2406 reseed happens in the middle of the object allocation it will
2407 still utilize the OSTs as uniformly as possible.
2412 Description: OST or MDS may oops in ping_evictor_main()
2413 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2414 but doesn't restart at beginning of obd_exports_timed list
2417 Severity : enhancement
2419 Description: improve by-nid export eviction on the MDS and OST
2420 Details : allow multiple exports with the same NID to be evicted at one
2421 time without re-searching the exports list.
2424 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2426 Description: MDS may oops in groups_free()
2427 Details : in rare race conditions a newly allocated group_info struct is
2428 freed again, and this can be NULL. The 2.4 compatibility code
2429 for groups_free() checked for a NULL pointer, but 2.6 did not.
2432 Frequency : common for liblustre clients doing little filesystem IO
2433 Bugzilla : 9352, 7313
2434 Description: server may evict liblustre clients accessing contended locks
2435 Details : if a client is granted a lock or receives a completion AST
2436 with a blocking AST already set it would not reply to the AST
2437 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2440 Frequency : lfs setstripe, only systems with more than 160 OSTs
2442 Description: unable to set striping with a starting offset beyond OST 160
2443 Details : llapi_create_file() incorrectly limited the starting stripe
2444 index to the maximum single-file stripe count.
2447 Frequency : LDAP users only
2449 Description: lconf did not handle in-kernel recovery with LDAP properly
2450 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2452 Severity : enhancement
2454 Description: bind OST threads to NUMA nodes to improve performance
2455 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2456 node and do their allocations there to localize memory access
2458 Severity : enhancement
2460 Description: llmount can determine client NID directly from Myrinet (GM)
2461 Details : the client NID code from gmnalnid was moved directly into
2462 llmount, removing the need to use this or specifying the
2463 client NID explicitly when mounting GM clients with zeroconf
2466 Frequency : if client is started with down MDS
2468 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2469 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2470 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2471 ptlrpc_import_delayed_req() because we didn't check if the
2472 request was interrupted, and we also didn't break out of the
2473 event loop if there was no timeout
2478 Description: data loss during non-page-aligned writes to a single file from
2479 both multiple nodes and multiple threads on one node at same time
2480 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2481 inconsistency led to false short-reads, that were cached and later
2482 used by ->prepare_write() to fill in partially written page,
2483 leading to data loss.
2486 Frequency : always, if lconf --abort_recovery used
2488 Description: lconf --abort_recovery fails with 'Operation not supported'
2489 Details : lconf was attempting to abort recovery on the MDT device and not
2492 Severity : enhancement
2494 Description: remove cleanup logs
2495 Details : replace lconf-generated cleanup logs with lustre internal
2496 cleanup routines. Eliminates the need for client-cleanup and
2499 Severity : enhancement
2501 Description: add support for EAs (user and system) on lustre filesystems
2502 Details : it is now possible to store extended attributes in the Lustre
2503 client filesystem, and with the user_xattr mount option it
2504 is possible to allow users to store EAs on their files also
2506 Severity : enhancement
2508 Description: Add possibility (config option) to show minimal available OST free
2510 Details : When compiled with --enable-mindf configure option, statfs(2)
2511 (and so, df) will return least minimal free space available from
2512 all OSTs as amount of free space on FS, instead of summary of
2513 free spaces of all OSTs.
2515 Severity : enhancement
2517 Description: do not expand extent locks acquired on OST-side
2518 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2519 by server: they are not cached anyway.
2522 Frequency : when mmap is used/binaries executed from Lustre
2524 Description: Unmmap pages before throwing them away from read cache.
2525 Details : llap_shrink cache now attempts to unmap pages before discarding
2526 them (if unmapping failed - do not discard). SLES9 kernel has
2527 extra checks that trigger if this unmapping is not done first.
2532 Description: lconf didn't resolve symlinks before checking to see whether a
2533 given mountpoint was already in use
2536 Frequency : when migrating failover services
2537 Bugzilla : 6395, 9514
2538 Description: When migrating a subset of services from a node (e.g. failback
2539 from a failover service node) the remaining services would
2540 time out and evict clients.
2541 Details : lconf --force (implied by --failover) sets the global obd_timeout
2542 to 5 seconds in order to quickly disconnect, but this caused
2543 other RPCs to time out too quickly. Do not change the global
2544 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2546 Severity : enhancement
2547 Frequency : if MDS is started with down OST
2548 Bugzilla : 9439,5706
2549 Description: Allow startup/shutdown of an MDS without depending on the
2550 availability of the OSTs.
2551 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2552 Add appropriate locking and lov-osc refcounts for safe
2553 cleaning. Add osc abort_inflight calls in case the
2554 synchronize never started.
2557 Frequency : occasional (Cray XT3 only)
2559 Description: root not authorized to access files in CRAY_PORTALS environment
2560 Details : The client process capabilities were not honoured on the MDS in
2561 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2562 been accessed by an authorized user then root was able to access
2563 the file on the local client also. The root user capabilities
2564 are now allowed on the MDS, as this environment has secure UID.
2567 Frequency : occasional
2569 Description: ldiskfs "too long searching" message happens too often
2570 Details : A debugging message (otherwise harmless) prints too often on
2571 the OST console. This has been reduced to only happen when
2572 there are fragmentation problems on the filesystem.
2577 Description: Division by zero in statfs when all OSCs are inactive
2578 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2579 lov_statfs divides by value returned by lov_get_stripecnt().
2583 Bugzilla : 9489, 3273
2584 Description: First write from each client to each OST was only 4kB in size,
2585 to initialize client writeback cache, which caused sub-optimal
2586 RPCs and poor layout on disk for the first writen file.
2587 Details : Clients now request an initial cache grant at (re)connect time
2588 and so that they can start streaming writes to the cache right
2589 away and always do full-sized RPCs if there is enough data.
2590 If the OST is rebooted the client also re-establishes its grant
2591 so that client cached writes will be honoured under the grant.
2596 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2597 Details : Now I/O RPCs go to different portal number and (presumably) fast
2598 lock requests (and glimses) and other RPCs get their own service
2599 threads pool that should be able to service those RPCs
2602 Severity : enhancement
2604 Description: Ability to exchange lustre version between client and servers and
2605 issue warnings at client side if client is too old. Also for
2606 liblustre clients there is ability to refuse connection of too old
2608 Details : New 'version' field is added to connect data structure that is
2609 filled with version info. That info is later checked by server and
2613 Frequency : rare, liblustre only.
2614 Bugzilla : 9296, 9581
2615 Description: Two simultaneous writes from liblustre at offset within same page
2616 might proceed at the same time overwriting eachother with stale
2618 Details : I/O lock withing llu_file_prwv was released too early, before data
2619 actually was hitting the wire. Extended lock-holding time until
2620 server acknowledges receiving data.
2623 Frequency : extremely rare. Never observed in practice.
2625 Description: avoid generating lustre_handle cookie of 0.
2626 Details : class_handle_hash() generates handle cookies by incrementing
2627 global counter, and can hit 0 occasionaly (this is unlikely, but
2628 not impossible, because initial value of cookie counter is
2629 selected randonly). Value of 0 is used as a sentinel meaning
2630 "unassigned handle" --- avoid it. Also coalesce two critical
2631 sections in this function into one.
2633 Severity : enhancement
2635 Description: allow liblustre clients to delegate truncate locking to OST
2636 Details : To avoid overhead of locking, liblustre client instructs OST to
2637 take extent lock in ost_punch() on client's behalf. New connection
2638 flag is added to handle backward compatibility.
2640 Severity : enhancement
2641 Bugzilla : 4928, 7341, 9758
2642 Description: allow number of OST service threads to be specified
2643 Details : a module parameter allows the number of OST service threads
2644 to be specified via "options ost ost_num_threads={N}" in the
2645 OSS's /etc/modules.conf or /etc/modprobe.conf.
2649 Bugzilla : 6146, 9635, 9895
2650 Description: servers crash with bad pointer in target_handle_connect()
2651 Details : In rare cases when a client is reconnecting it was possible that
2652 the connection request was the last reference for that export.
2653 We would temporarily drop the export reference and get a new
2654 one, but this may have been the last reference and the export
2655 was just destroyed. Get new reference before dropping old one.
2657 Severity : enhancement
2658 Frequency : if client is started with failover MDS
2660 Description: Allow multiple MDS hostnames in the mount command
2661 Details : Try to read the configuration from all specified MDS
2662 hostnames during a client mount in case the "primary"
2665 Severity : enhancement
2667 Description: Stop sending data to evicted clients as soon as possible.
2668 Details : Check if the client we are about to send or are sending data to
2669 was evicted already. (Check is done every second of waiting,
2670 for which l_wait_event interface was extended to allow checking
2671 of exit condition at specified intervals).
2674 Frequency : rare, normally only when NFS exporting is done from client
2676 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2678 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2681 Frequency : always, if srand() is called before liblustre initialization
2683 Description: Liblustre uses system PRNG disturbing its usage by user application
2684 Details : Introduce internal to lustre fast and high-quality PRNG for
2685 lustre usage and make liblustre and some other places in generic
2686 lustre code to use it.
2688 Severity : enhancement
2689 Bugzilla : 9477, 9557, 9870
2690 Description: Verify that the MDS configuration logs are updated when xml is
2691 Details : Check if the .xml configuration logs are newer than the config
2692 logs stored on the MDS and report an error if this is the case.
2693 Request --write-conf, or allow starting with --old_conf.
2695 Severity : enhancement
2697 Description: Handle symlinks in the path when checking if Lustre is mounted.
2698 Details : Resolve intermediate symlinks when checking if a client has
2699 mounted a filesystem to avoid duplicate client mounts.
2704 Description: lconf can hit an error exception but still return success.
2705 Details : The lconf command catches the Command error exception at the top
2706 level script context and will exit with the associated exit
2707 status, but doesn't ensure that this exit status is non-zero.
2712 Description: failure of ptlrpc thread startup can cause oops
2713 Details : Starting a ptlrpc service thread can fail if there are a large
2714 number of threads or the server memory is very fragmented.
2715 Handle this without oopsing.
2718 Frequency : always, only if liblustre and non-default acceptor port was used
2720 Description: liblustre cannot connect to servers with non-default acceptor port
2721 Details : tcpnal_set_default_params() was not called and was therefore
2722 ignoring the environment varaible TCPNAL_PORT, as well as other
2723 TCPNAL_ environment variables
2728 Description: two objects could be created on the same OST for a single file
2729 Details : If an OST is down, in some cases it was possible to create two
2730 objects on a single OST for a single file. No problems other
2731 than potential performance impact and spurious error messages.
2735 Bugzilla : 5681, 9562
2736 Description: Client may oops in ll_unhash_aliases
2737 Details : Client dcache may become inconsistent in race condition.
2738 In some cases "getcwd" can fail if the current directory is
2744 Description: Inode refcounting problems in NFS export code
2745 Details : link_raw functions used to call d_instantiate without obtaining
2746 extra inode reference first.
2750 Bugzilla : 9942, 9903
2751 Description: Referencing freed requests leading to crash, memleaks with NFS.
2752 Details : We used to require that call to ll_revalidate_it was always
2753 followed by ll_lookup_it. Also with revalidate_special() it is
2754 possible to call ll_revalidate_it() twice for the same dentry
2755 even if first occurence returned success. This fix changes semantic
2756 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2757 reference on a request referred from the intent.
2758 ll_intent_release() then releases such a request.
2761 Frequency : rare, normally benchmark loads only
2763 Description: unlinked inodes were kept in memory on the client
2764 Details : If a client is repeatedly creating and unlinking files it
2765 can accumulate a lot of stale inodes in the inode slab cache.
2766 If there is no other client load running this can cause the
2767 client node to run out of memory. Instead flush old inodes
2768 from client cache that have the same inode number as a new inode.
2771 Frequency : SLES9 2.6.5 kernel and long filenames only
2772 Bugzilla : 9969, 10379
2773 Description: utime reports stale NFS file handle
2774 Details : SLES9 uses out-of-dentry names in some cases, which confused
2775 the lustre dentry revalidation. Change it to always use the
2779 Frequency : rare, unless heavy write-truncate concurrency is continuous
2780 Bugzilla : 4180, 6984, 7171, 9963, 9331
2781 Description: OST becomes very slow and/or deadlocked during object unlink
2782 Details : filter_destroy() was holding onto the parent directory lock
2783 while truncating+unlinking objects. For very large objects this
2784 may block other threads for a long time and slow overall OST
2785 responsiveness. It may also be possible to get a lock ordering
2786 deadlock in this case, or run out of journal credits because of
2787 the combined truncate+unlink. Solution is to do object truncate
2788 first in one transaction without parent lock, and then do the
2789 final unlink in a new transaction with the parent lock. This
2790 reduces the lock hold time dramatically.
2793 Frequency : rare, 2.4 kernels only
2795 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2796 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2797 before unmounting their filesystem in order to avoid deadlock.
2798 The kernel_locked() function in 2.4 kernels only checks whether
2799 the kernel lock is held, not whether it is this process that is
2800 holding it as 2.6 kernels do.
2805 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2806 Details : The client ptlrpc code may be trying to reconnect to a down
2807 server before a previous connection attempt has timed out.
2808 Increase the reconnect interval to be longer than the connection
2809 timeout interval to avoid sending duplicate connections to
2813 Frequency : echo_client brw_test command
2815 Description: fix echo_client to work with OST preallocated code
2816 Details : OST preallocation code (5137) didn't take echo_client IO path
2817 into account: echo_client calls filter methods outside of any
2818 OST thread and, hence, there is no per-thread preallocated
2819 pages and buffers to use. Solution: hijack pga pages for IO. As
2820 a byproduct, this avoids unnecessary data copying.
2824 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2825 Description: Client can oops in mdc_commit_close() after open replay
2826 Details : It was possible for the MDS to return an open request with no
2827 transaction number in mds_finish_transno() if the client was
2828 evicted, but without actually returning an error. Clients
2829 would later try to replay that open and may trip an assertion
2830 Simplify the client close codepath, and always return an error
2831 from the MDS in case the open is not successful.
2834 Frequency : rare, 2.6 OSTs only
2836 Description: OST may deadlock under high load on fragmented files
2837 Details : If there was a heavy load and highly-fragmented OST filesystems
2838 it was possible to have all the OST threads deadlock waiting on
2839 allocation of biovecs, because the biovecs were not released
2840 until the entire RPC IO was completed. Instead, release biovecs
2841 as soon as they are complete to ensure forward IO progress.
2843 Severity : enhancement
2845 Description: Support for specifying external journal device at mount
2846 Details : If an OST or MDS device is formatted with an external journal
2847 device, this device major/minor is stored in the ext3 superblock
2848 and may not be valid for failover. Allow detecting and
2849 specifying the external journal at mount time.
2854 Description: Mounting an MDS with pending unlinked files may cause oops
2855 Details : target_finish_recovery() calls mds_postrecov() which returned
2856 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2857 considers this an error and immediately begins cleaning up the
2858 lov, just after starting the mds_lov process
2860 Severity : enhancement
2862 Description: Implement 'lfs df' to report actual free space on per-OST basis
2863 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2864 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2865 usage of objects; '-h' to report in human readable format.
2867 ------------------------------------------------------------------------------
2869 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2876 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2877 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2878 at startup during orphan unlinking. Instead, skip buddy update
2879 before setup, as it will be regenerated after recovery is complete.
2884 Description: If an OST is inactive, its locks might reference stale inodes.
2885 Details : lov_change_cbdata() must iterate over all namespaces, even if
2886 they are inactive to clear inode references from the lock.
2888 Severity : enhancement
2889 Frequency : occasional, if non-standard max_dirty_mb used
2891 Description: Client will block write RPCs if not enough grant
2892 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2893 then the client will block writes while waiting for another RPC
2894 to complete instead of consuming its dirty limit. With change
2895 we get improved performance when max_dirty_mb is small.
2897 Severity : enhancement
2898 Bugzilla : 3389, 6253
2899 Description: Add support for supplementary groups on the MDS.
2900 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2901 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2902 lookups for user supplementary groups into a cache.
2906 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2907 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2908 for liblustre O_EXCL is enforced by MDS.
2912 Description: readdir never returns NULL in liblustre.
2913 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2914 dentry in struct dirent.
2918 Frequency : liblustre only, depends on application IO pattern
2919 Description: liblustre clients evicted if not contacting servers
2920 Details : Don't put liblustre clients into the ping_evictor list, so
2921 they will not be evicted by the pinger ever.
2923 Severity : enhancement
2925 Description: Add ability to evict clients by NID from MDS.
2926 Details : By echoing "nid:$NID" string into
2927 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2928 $NID would be instantly evicted from this MDS and from all active
2929 OSTs connected to it.
2933 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2934 Details : lookup_it_finish() used to query file size from OSTs that was not
2939 Description: service threads change working directory to that of init
2940 Details : Starting lustre service threads may pin the working directory
2941 of the parent thread, making that filesystem busy. Threads
2942 now change to the working directory of init to avoid this.
2946 Frequency : during shutdown only
2947 Description: shutdown with a failed MDS or OST can cause unmount to hang
2948 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2953 Frequency : 2.6 only
2954 Description: chmod/chown may include an extra supplementary group
2955 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2956 supplementary group and if none is specified this is used.
2959 Bugzilla : 5479 (6816)
2961 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2962 Details : If lookup is in progress on a file that is unlinked we might try
2963 to revalidate the inode and fail in revalidate after lookup is
2964 complete and ll_file_open() enqueues the open again but
2965 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2968 Frequency : always, if lconf --abort_recovery used
2970 Description: lconf --abort_recovery fails with 'Operation not supported'
2971 Details : lconf was attempting to abort recovery on the MDT device and not
2974 ------------------------------------------------------------------------------
2976 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2981 Frequency : rare (only unsupported configurations with a node running as an
2983 Bugzilla : 6514, 5137
2984 Description: Mounting a Lustre file system on a node running as an OST could
2986 Details : OSTs now preallocates memory needed to write out data at
2987 startup, instead of when needed, to avoid having to
2988 allocate memory in possibly low memory situations.
2989 Specifically, if the file system is mounted on on OST,
2990 memory pressure could force it to try to write out data,
2991 which it needed to allocate memory to do. Due to the low
2992 memory, it would be unable to do so and the node would
2993 become unresponsive.
2995 Severity : enhancement
2997 Description: Addition of lconf --service command line option
2998 Details : lconf now accepts a '--service <arg>' option, which is
2999 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3001 Severity : enhancement
3003 Description: Failover mode is now the default for OSTs.
3004 Details : By default, OSTs will now run in failover mode. To return to
3005 the old behaviour, add '--failout' to the lmc line for OSTs.
3007 Severity : enhancement
3009 Description: Health checks are now provided for MDS and OSTs
3010 Details : Additional detailed health check information on MSD and OSTs
3011 is now provided through the procfs health_check value.
3014 Frequency : occasional, depends on IO load
3016 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3017 after numerous create/delete cycles
3018 Details : The ext3 inode allocation policy would not allocate new inodes
3019 very well on the OSTs because there are no new directories
3020 being created. Instead we look for groups with free space if
3021 the parent directories are nearly full.
3026 Description: Network or server problems during mount may cause partially
3027 mounted clients instead of returning an error.
3028 Details : The config llog parsing code may overwrite the error return
3029 code during mount error handling, returning success instead
3035 Description: MDS can fail to allocate large reply buffers
3036 Details : After long uptimes the MDS can fail to allocate large reply
3037 buffers (e.g. zconf client mount config records) due to memory
3038 fragmentation or consumption by the buffer cache. Preallocate
3039 some large reply buffers so that these replies can be sent even
3040 under memory pressure.
3044 Frequency : rare (liblustre)
3045 Description: fsx running with liblustre complained that using truncate() to
3046 extend the file doesn't work. This patch corrects that issue.
3047 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3048 ATTR_SIZE and lsm use in llu_setattr_raw.
3052 Frequency : rare, only 2.6 kernels
3053 Description: Unusual file access patterns on the MDS may result in inode
3054 data being lost in very rare circumstances.
3055 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3056 code in the 2.6 kernel caused the inode and/or EA data not to
3057 be read from disk, causing single-file corruption.
3061 Frequency : rare, only 2.6 filesystems using extents
3062 Description: Heavy concurrent write and delete load may cause data corruption.
3063 Details : It was possible under high-load situations to have an extent
3064 metadata block in the block device cache from a just-unlinked
3065 file overwrite a newly-allocated data block. We now unmap any
3066 metadata buffers that alias just-allocated data blocks.
3070 Frequency : filesystems with default stripe_count larger than 77
3071 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3072 Details : lconf specifies an inode size of 4096 bytes when the default
3073 stripe_count is larger than 77. This conflicts with the default
3074 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3075 this case to avoid pinning too much memory for large EAs.
3077 ------------------------------------------------------------------------------
3079 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3084 Frequency : rare (extremely heavy IO load with hundreds of clients)
3086 Description: Client is evicted, gets IO error writing to file
3087 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3088 caused clients to be evicted for AST timeouts. The fixes in
3089 bug 5192 mean we no longer need to have such short AST timeouts
3090 so ldlm_timeout has been increased.
3093 Frequency : occasional during --force or --failover shutdown under load
3094 Bugzilla : 5949, 4834
3095 Description: Server oops/LBUG if stopped with --force or --failover under load
3096 Details : a collection of import/export refcount and cleanup ordering
3097 issues fixed for safer force cleanup
3100 Frequency : only filesystems larger than 120 OSTs
3101 Bugzilla : 5990, 6223
3102 Description: lfs getstripe would oops on a very large filesystem
3103 Details : lov_getconfig used kfree on vmalloc'd memory
3106 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3107 Bugzilla : 6242, 6243
3108 Description: reading from files that had been truncated to a non-zero size
3109 but never opened returned no data
3110 Details : ll_file_read() reads zeros from no-object files to EOF
3115 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3116 Details : There's little harm in aborting MDS/OSS recovery and letting it
3117 try again, so I removed the LASSERT and return an error instead.
3119 Severity : enhancement
3121 Description: New debugging infrastructure for tracking down data corruption
3122 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3123 (b) cover more of the client-side code path, and (c) try to narrow
3124 down where problems occurred
3128 Bugzilla : 3819, 4364, 4397, 6313
3129 Description: Racing close and eviction MDS could cause assertion in mds_close
3130 Details : It was possible to get multiple mfd references during close and
3131 client eviction, leading to one thread referencing a freed mfd.
3133 Severity: : enhancement
3134 Bugzilla : 3262, 6359
3135 Description: Attempts to reconnect to servers are now more aggressive.
3136 Details : This builds on the enhanced upcall-less recovery that was added
3137 in 1.4.2. When trying to reconnect to servers, clients will
3138 now try each server in the failover group every 10 seconds. By
3139 default, clients would previously try one server every 25 seconds.
3144 Description: After recovery, certain operations trigger a failed
3145 assertion on a client.
3146 Details : Failing over an mds, using lconf -d --failover, while a
3147 client was doing a readdir() call would cause the client to
3148 LBUG after recovery completed and the readdir() was resent.
3150 Severity : enhancement
3152 Description: Default groups are now added by lconf
3153 Details : You can now run lconf --group <servicename> without having to
3154 manually add groups with lmc.
3157 Frequency : occasional
3159 Description: Nodes with an elan id of 0 trigger a failed assertion
3162 Frequency : always when accessing e.g. tty/console device nodes
3164 Description: tty and some other devices nodes cannot be used on lustre
3165 Details : file's private_data field is used by device data and lustre
3166 values in there got lost. New field was added to struct file to
3167 store fs-specific private data.
3170 Frequency : when exporting Lustre via NFS
3172 Description: NFSD failed occasionally when looking up a path component
3173 Details : NFSD is looking up ".." which was broken in ext3 directories
3174 that had grown large enough to become hashed.
3177 Frequency : Clusters with multiple interfaces not on the same subnet
3179 Description: Nodes will repeatedly try to reconnect to an interface which it
3180 cannot reach and report an error to the log.
3181 Details : Extra peer list entries will be created by lconf with some peers
3182 unreachable. lconf now validates the peer before adding it.
3185 Frequency : Only if a default stripe is set on the filesystem root.
3187 Description: Setting a default stripe on the filesystem root prevented the
3188 filesystem from being remounted.
3189 Details : The client was sending extra request flags in the root getattr
3190 request and did not allocate a reply buffer for the dir EA.
3193 Frequency : occasional, higher if lots of files are accessed by one client
3194 Bugzilla : 6159, 6097
3195 Description: Client trips assertion regarding lsm mismatch/magic
3196 Details : While revalidating inodes the VFS looks up inodes with ifind()
3197 and in rare cases can find an inode that is being freed.
3198 The ll_test_inode() code will free the lsm during ifind()
3199 when it finds an existing inode and then the VFS later attaches
3200 this free lsm to a new inode.
3204 Bugzilla : 6422, 7030
3205 Description: MDS deadlock between mkdir and client eviction
3206 Details : Creating a new file via mkdir or mknod (starting a transaction
3207 and getting the ns lock) can deadlock with client eviction
3208 (gets ns lock and trying to finish a synchronous transaction).
3211 Frequency : occasional
3212 Description: While starting a server, the fsfilt_ext3 module could not be
3214 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3215 kernels. Previously, lconf would still use the ext3 name
3216 when trying to load modules. Now, it will correctly use
3217 ext3 on 2.4 and ldiskfs on 2.6.
3219 Severity : enhancement
3220 Description: The default stripe count has been changed to 1
3221 Details : The interpretation of the default stripe count (0, to lfs
3222 or lmc) has been changed to mean striping across a single
3223 OST, rather than all available. For general usage we have
3224 found a stripe count of 1 or 2 works best.
3226 Severity : enhancement
3227 Description: Add support for compiling against Cray portals.
3228 Details : Conditional compiling for some areas that are different
3232 Frequency : occasional
3233 Bugzilla : 6409, 6834
3234 Description: Creating files with an explicit stripe count may lead to
3235 a failed assertion on the MDS
3236 Details : If some OSTs are full or unavailable, creating files may
3237 trigger a failed assertion on the MDS. Now, Lustre will
3238 try to use other servers or return an error to the
3242 Frequency : occasional
3244 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3245 caused lock timeout during readahead (since 1.4.2).
3246 Details : Processes doing readahead might match a lock that hasn't been
3247 granted yet if there are overlapping and conflicting lock
3248 requests. The readahead process waits on ungranted lock
3249 (original lock is CBPENDING), while OST waits for that process
3250 to cancel CBPENDING read lock and eventually evicts client.
3252 Severity : enhancement
3254 Description: Initial enabling of flock support for clients
3255 Details : Implements fcntl advisory locking and file status functions.
3256 This feature is provided as an optional mount flag (default
3257 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3258 locking are implemented yet, and those that are are not guaranteed
3259 to be completely correct in production environments.
3260 mount -t lustre -o [flock|noflock] ...
3263 Frequency : occasional
3265 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3266 assertion in the ext3 JBD (journaling) layer.
3267 Details : The b_committed_data struct is protected by the big kernel lock
3268 in 2.4 kernels, serializing journal_commit_transaction() and
3269 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3270 access to this struct. In 2.6 kernels there is finer grained
3271 locking to improve SMP performance of the JBD layer.
3275 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3280 Description: merge and cleanup kernel patches.
3283 -----------------------------------------------------------------------------
3285 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3287 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3288 versions of Lustre. Please update all servers and clients to
3289 version 1.4.2 or later at the same time. You must also run
3290 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3291 to update the configuration logs.
3293 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3294 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3295 - swab llog records of type '0' so we get proper header size/idx (5861)
3296 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3297 - fix rename of one directory over another leaking an inode (5953)
3298 - avoid SetPageDirty on 2.6 (5981)
3299 - don't re-add just-being-destroyed locks to the waiting list (5653)
3300 - when creating new directories, inherit the parent's custom
3301 striping settings if present parent (3048)
3302 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3303 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3304 - don't hold client locks on temporary worklist from l_lru (5666)
3305 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3306 - automatically evict dead clients (5921)
3307 - Update file size properly in create+truncate+fstat case (6196)
3308 - Do not unhash mountpoint dentries, do not allow removal of
3310 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3311 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3312 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3313 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3314 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3315 - Correctly handle reads of files with no objects (6243)
3316 - lctl recover will also mark a device active if deactivate used (5933)
3318 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3319 - allow --write-conf on an MDS with different nettype than client (5619)
3320 - don't write config llogs to MDS for mounts not from that MDS (5617)
3321 - lconf should create multiple TCP connections from a client (5201)
3322 - init scripts are now turned off by default; run chkconfig --on
3323 lustre and chkconfig --on lustrefs to use them
3324 - upcalls are no longer needed for clients to recover to failover
3326 - add --abort-recovery option to lconf to abort recovery on device
3328 - add support for an arbitrary number of OSTs (3026)
3329 - Quota support protocol changes.
3330 - forward compatibility changes to wire structs (6007)
3331 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3332 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3333 - improved "lustre status" script
3334 - initialize blocksize for non-regular files (6062)
3335 - added --disable-server and --disable-client configure options (5782)
3336 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3337 - Vanilla 2.4.29 support
3338 - increase maximum number of obd devices to 520 (6242)
3339 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3340 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3341 SLES 9 2.6.5 kernels
3342 - limit stripes per file to 160 (the maximum EA size) (6093)
3344 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3347 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3348 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3349 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3350 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3351 - small ext3 extents cleanups and fixes (5733)
3352 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3353 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3354 - avoid needless client->OST connect, fix handle mismatch (5317)
3355 - fix DLM error path that led to out-of-sync client, long delays (5779)
3356 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3357 - fix several locking issues related to i_size (5492,5624,5654,5672)
3358 - don't move pending lock onto export if it is already evicted (5683)
3359 - fix kernel oops when creating .foo in unlinked directory (5548)
3360 - fix deadlock in obdfilter statistics vs. object create (5811)
3361 - use time_{before,after} to avoid timer jiffies wrap (5882)
3362 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3363 - Do not leak request if server was not able to process it (5154)
3364 - If mds_open unable to find parent dir, make that negative lookup(5154)
3365 - don't create new directories with extent-mapping (5909, 5936)
3367 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3368 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3369 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3370 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3371 - support for RHEL4 kernel on i686 (5773)
3372 - provide error messages when incompatible logs are encountered (5898)
3374 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3375 * version 1.4.0.10 (1.4.1 release candidate 1)
3377 - don't keep a lock reference when lock is not granted (4238)
3378 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3379 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3380 - drop import inflight refcount on signal_completed_replay error (5255)
3381 - unlock page after async write error during send (3677)
3382 - handle missing objects in filter_preprw_read properly (5265)
3383 - no transno return for symlink open, don't save no-trasno open (3440)
3384 - don't try to complete elan receive that already failed (4012)
3385 - free RPC server reply state on error (5406)
3386 - clean up thread from ptlrpc_start_thread() on error (5160)
3387 - readahead could read extra page into cache that wasn't ejected (5388)
3388 - prevent races in class_attach/setup/cleanup/detach (5260)
3389 - don't dereference de->d_inode after l_dput of de (5458)
3390 - use "int" for stripe value returned from lock_to_stripe (5544)
3391 - mballoc allocation and error-checking fixes in 2.6 (5504)
3392 - block device patches to fix I/O request sizes in 2.6 (5482)
3393 - look up hostnames for IB nals (5602)
3394 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3395 - don't start multiple acceptors for the same port (5277)
3396 - fix incorrect LASSERT in mds_getattr_name (5635)
3397 - export a proc file for general "ping" checking (5628)
3398 - fix "lfs check" to not block when the MDS is down (5628)
3400 - service request history (4965)
3401 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3402 - create an "evict_client" /proc entry on OSTs, like the MDS has
3403 - fix mount usage message, return errors per mount(8) (5168)
3404 - change grep [] to grep "[]" in tests so they work in more UMLs
3405 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3406 - remove ancient LOV_MAGIC_V0 EA support (5047)
3407 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3408 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3409 - allow readahead allocations to fail when low on memory (5383)
3410 - mmap locking landed again, after considerable improvement (2828)
3411 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3413 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3416 - send OST transaction number in read/write reply to free req (4966)
3417 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3418 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3419 - the watchdog thread now runs as interruptible (5246)
3420 - flock/lockf fixes (but it's still disabled, pending 5135)
3421 - don't use EXT3 constants in llite code (5094)
3422 - memory shortage at startup could cause assertion (5176)
3424 - reorganization of lov code
3425 - single portals codebase
3427 - add extents/mballoc support (5025)
3428 - direct I/O reads in the obdfilter (4048)
3429 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3431 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3434 - send OST transaction number in read/write reply to free req (4966)
3435 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3436 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3437 - the watchdog thread now runs as interruptible (5246)
3438 - handle missing objects in filter_preprw_read properly (5265)
3439 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3440 - drop import inflight refcount on signal_completed_replay error (5255)
3441 - unlock page after async write error during send (3677)
3442 - return original error code on reconstructed replies (3761)
3443 - no transno return for symlink open, don't save no-trasno open (3440)
3445 - add pid to ldlm debugging output (4922)
3446 - bump the watchdog timeouts -- we can't handle 30sec yet
3447 - extra debugging for orphan dentry/inode bug (5259)
3449 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3452 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3453 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3454 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3455 - fix NULL dereference in /proc/sys/portals/routes (4827)
3456 - allow failed mdc_close() operations to be interrupted (4561)
3457 - stop precreate on OST before MDS would time out on it (4778)
3458 - don't send partial-page writes before EOF from client (4410)
3459 - discard client grant for sub-page writes on large-page clients (4520)
3460 - don't free dentries not owned by NFS code, check generation (4806)
3461 - fix lsm leak if mds_create_objects() fails (4801)
3462 - limit debug_daemon file size, always print CERROR messages (4789)
3463 - use transno after validating reply (3892)
3464 - process timed out requests if import state changes (3754)
3465 - update mtime on OST during writes, return in glimpse (4829)
3466 - add mkfsoptions to LDAP (4679)
3467 - use ->max_readahead method instead of zapping global ra (5039)
3468 - don't interrupt __l_wait_event() during strace
3470 - add software watchdogs to catch hung threads quickly (4941)
3471 - make lustrefs init script start after nfs is mounted
3472 - fix CWARN/ERROR duplication (4930)
3473 - return async write errors to application if possible (2248)
3474 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3475 - print NAL number in %x format (4645)
3476 - update barely-supported suse-2.4.21-171 series (4842)
3477 - support for sles 9 %post scripts
3478 - support for building 2.6 kernel-source packages
3479 - support for sles km_* packages
3481 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3484 - ignore -ENOENT errors in osc_destroy (3639)
3485 - notify osc create thread that OSC is being cleaned up (4600)
3486 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3487 - reconstruct ost_handle() like mds_handle() (4657)
3488 - create a new thread to do import eviction to avoid deadlock (3969)
3489 - let lconf resolve symlinked-to devices (4629)
3490 - don't unlink "objects" from directory with default EA (4554)
3491 - hold socknal file ref over connect in case target is down (4394)
3492 - allow more than 32000 subdirectories in a single directory (3244)
3493 - fix blocks count for O_DIRECT writes (3751)
3494 - OST returns ENOSPC from object create when no space left (4539)
3495 - don't send truncate RPC if file size isn't changing (4410)
3496 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3497 - bind to privileged port in socknal and tcpnal (3689)
3499 - rate limit CERROR/CWARN console message to avoid overload (4519)
3500 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3501 - basic mmap support (3918)
3502 - kernel patch series update from b1_4 (4711)
3504 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3507 - avoid crash during MDS cleanup with OST shut down (2775)
3508 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3509 - don't use bad inodes on the MDS (3744)
3510 - dynamic object preallocation to improve recovery speed (4236)
3511 - don't hold spinlock over lock dumping or change debug flags (4401)
3512 - don't zero obd_dev when it is force cleaned (3651)
3513 - print grants to console if they go negative (4431)
3514 - "lctl deactivate" will stop automatic recovery attempts (3406)
3515 - look for existing locks in ldlm_handle_enqueue() (3764)
3516 - don't resolve lock handle twice in recovery avoiding race (4401)
3517 - revalidate should check working dir is a directory (4134)
3519 - don't always mark "slow" obdfilter messages as errors (4418)
3521 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3524 - don't close LustreDB during write_conf until it is done (3860)
3525 - fix typo in lconf for_each_profile (3821)
3526 - allow dumping logs from multiple threads at one time (3820)
3527 - don't allow multiple threads in OSC recovery (3812)
3528 - fix debug_size parameters (3864)
3529 - fix mds_postrecov to initialize import for llog ctxt (3121)
3530 - replace config semaphore with spinlock (3306)
3531 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3532 - don't allow enqueue to complete on a destroyed export (3822)
3533 - down write_lock before checking llog header bitmap (3825)
3534 - recover from lock replay timeout (3764)
3535 - up llog sem before sending rpc (3652)
3536 - reduce ns lock hold times when setting kms (3267)
3537 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3538 - fix NULL deref and obd_dev leak on setup error (3312)
3539 - replace some LBUG about llog ops with error handling (3841)
3540 - don't match INVALID dentries from d_lookup and spin (3784)
3541 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3542 - fix invalid assertion in ptlrpc_set_wait (3880)
3544 - add libwrap support for the TCP acceptor (3996)
3545 - add /proc/sys/portals/routes for non-root route listing (3994)
3546 - allow setting MDS UUID in .xml (2580)
3547 - print the stack of a process that LBUGs (4228)
3549 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3552 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3553 - make sure to unset replay flag from failed open requests (3440)
3554 - if default stripe count is 0, use OST count for inode size (3636)
3555 - update parent mtime/ctime on client for create/unlink (2611)
3556 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3557 - free recovery state on server during a forced cleanup (3571)
3558 - unregister_reply for resent reqs (3063)
3559 - loop back devices mounting and status check on 2.6 (3563)
3560 - fix resource-creation race that can provoke i_size == 0 (3513)
3561 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3562 - more debugging for page-accounting assertion (3746)
3563 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3564 - avoid deadlock after precreation failure (3758)
3565 - fix race and lock order deadlock in orphan handling (3450, 3750)
3566 - add validity checks when grabbing inodes from l_ast_data (3599)
3568 - add /proc/.../recovery_status to obdfilter (3428)
3569 - lightweight CDEBUG infrastructure, debug daemon (3668)
3570 - change default OSC RPC parameters to be better on small clusters
3571 - turn off OST read cache for files smaller than 32MB
3572 - install man pages and include them in rpms (3100)
3573 - add new init script for (un)mounting lustre filesystems (2593)
3574 - run chkconfig in %post for init scripts (3701)
3575 - drop scimac NAL (unmaintained)
3577 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3580 - clean kiobufs before and after use (3485)
3581 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3582 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3583 - change init script to fail more clearly if not run as root (1528)
3584 - allow clients to reconnect during replay (1742)
3585 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3586 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3587 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3588 - fix cond_resched (3554)
3589 - only evict unfinished clients after recovery (3515)
3590 - allow bulk resend, prevent data loss (3570)
3591 - dynamic ptlrpc request buffer allocation (2102)
3592 - don't allow unlinking open directory if it isn't empty (2904)
3593 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3594 - remove extraneous obd dereference causing LASSERT failure (3334)
3595 - don't use get_cycles() when creating temp. files on the mds (3156)
3596 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3597 - handle EEXIST for set-stripe, set proper directory name (3336)
3599 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3600 - fix ksocknal_fmb_callback() error messages (2918)
3602 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3605 - don't copy lvb into (possibly NULL) reply on error (2983)
3606 - don't deref dentry after dput, don't free lvb on error (2922)
3607 - use the kms to determine writeback rpc length (2947)
3608 - increment oti_logcookies when osc is inactive (2948)
3609 - update client's i_blocks count via lvb messages (2543)
3610 - handle intent open/close of special files properly (1557)
3611 - mount MDS with errors=remount-ro, like obdfilter (2009)
3612 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3613 - don't use cancelling-locks' kms values (2947)
3614 - use highest lock extent for kms, not last one (2925)
3615 - don't dereference ERR_PTR() dentry in error handling path (3107)
3616 - fix thread race in portals_debug_dumplog() (3122)
3617 - create lprocfs device entries at setup instead of at attach (1519)
3618 - common AST error handler, don't evict client on completion race (3145)
3619 - zero nameidata in detach_mnt in 2.6 (3118)
3620 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3621 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3622 - initialize RPC timeout timer earlier for 2.6 (3219)
3623 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3624 - print nal/nid for unknown nid (3258)
3625 - additional checks for oscc recovery before doing precreate (3284)
3626 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3627 - don't crash in mdc_close for bad permissions on open (3285)
3628 - zero i_rdev for non-device files (3147)
3629 - clear page->private before handing to FS, better assertion (3119)
3630 - tune the read pipeline (3236)
3631 - fix incorrect decref of invalidated dentry (2350)
3632 - provide read-ahead stats and refine rpc in flight stats (3328)
3633 - don't hold journal transaction open across create RPC (3313)
3634 - update atime on MDS at close time (3265)
3635 - close LDAP connection when recovering to avoid server load (3315)
3636 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3637 - don't leak open file on MDS after open resend (3325)
3638 - serialize filter_precreate and filter_destroy_precreated (3329)
3639 - loop device shouldn't call sync_dev() for nul device (3092)
3640 - clear page cache after eviction (2766)
3641 - resynchronize MDS->OST in background (2824)
3642 - refuse to mount the same filesystem twice on same mountpoint (3394)
3643 - allow llmount to create routes for mounting behind routers (3320)
3644 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3645 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3646 - fix rare problem with rename on htree directories (3417)
3648 - allow default OST striping configuration per directory (1414)
3649 - fix compilation for qswnal for 2.6 kernels (3125)
3650 - increase maximum number of MDS request buffers for large systems
3651 - change liblustreapi to be useful for external progs like lfsck (3098)
3652 - increase local configuration timeout for slow disks (3353)
3653 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3655 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3658 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3659 - don't overwrite extent policy data in reply if lock was blocked (2901)
3660 - drop filter export grants atomically with removal from device (2663)
3661 - del obd_self_export from work_list in class_disconnect_exports (2908)
3662 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3663 - swab reply message in mdc_close, other PPC fixes (2464)
3664 - fix destroying of named logs (2325)
3665 - overwrite old logs when running lconf --write_conf (2264)
3666 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3667 - fix race in target_handle_connect (2898)
3668 - mds_reint_create() should take same inode create lock (2926)
3669 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3670 - don't close files for self_export to avoid uninitialized obd (2936)
3671 - allow MDS with the same name as client node (2939)
3672 - hold dentry reference for closed log files for unlink (2325)
3673 - reserve space for all logs during transactions (2059)
3674 - don't evict page beyond end of stripe extent (2925)
3675 - don't oops on a deleted current working directory (2399)
3676 - handle hard links to targets without a parent properly (2517)
3677 - don't dereference NULL lock when racing during eviction (2867)
3678 - don't grow lock extents when lots of conflicting locks (2919)
3680 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3683 - account for cache space usage on clients to avoid data loss (974)
3684 - lfsck support in lustre kernel code (2349)
3685 - reduce journal credits needed for BRW writes (2370)
3686 - orphan handling to avoid losing space on client/server crashes
3687 - ptlrpcd can be blocked, stopping ALL progress (2477)
3688 - use lock value blocks to assist in proper KMS, faster stat (1021)
3689 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3690 - recovery for initial connections (2355)
3691 - fixes for mds_cleanup_orphans (1934)
3692 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3693 - block all file creations until orphan recovery completes (1901)
3694 - client remove rq_connection from request struct (2423)
3695 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3696 - recovery timer race (2670)
3697 - mdc_close recovey bug (2532)
3698 - ptlrpc cleanup bug (2710)
3699 - mds timeout on local locks (2588)
3700 - namespace lock held during RPCs (2431)
3701 - handle interrupted sync write properly (2503)
3702 - don't try to handle a message that hasn't been replied to (2699)
3703 - client assert failure during cleanup after abort recovery (2701)
3704 - leak mdc device after failed mount (2712)
3705 - ptlrpc_check_set allows timedout requests to complete (2714)
3706 - wait for inflight reqs when ptlrpcd finishes (2710)
3707 - make sure unregistered services are removed from the srv_list
3708 - reset bulk XID's when resending them (caught by 1138 test)
3709 - unregister_bulk after timeout
3710 - fix lconf error (2694)
3711 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3712 - readahead locks pages, leaves pending causing memory pressure (2673)
3713 - increase OST request buffers to 4096 on large machines (2729)
3714 - fix up permission of existing directories in simple_mkdir (2661)
3715 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3716 - don't assign transno to errored transactions (2742)
3717 - don't delete objects on OST if given a bogus objid from MDS (2751)
3718 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3719 - if rq_no_resend, then timeout request after recovery (2432)
3720 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3721 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3722 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3723 - MDS recovery shouldn't delete objects that it has given out (2730)
3724 - if enqueue arrives after completion, don't clobber LVB (2819)
3725 - don't unlock pages twice when trigger_group_io returns error (2814)
3726 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3727 - don't write pages to disk if there was an error (1450)
3728 - don't ping imports that have recovery disabled (2676)
3729 - take buffered bytes into account when balancing socknal conn (2817)
3730 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3731 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3732 - remove little-endian swabbing of llog records (1987)
3733 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3734 - retry reposting request buffers if they fail (1191)
3735 - grow extent at grant time to avoid granting a revoked lock (2809)
3736 - lock revoke doesn't evict page if covered by a second lock (2765)
3737 - disable VM readahead to avoid reading outside lock extents (2805)
3739 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3740 - updated kernel patches for hp-2.4.20 kernel (2681)
3742 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3745 - fix truncated write corruption (2366)
3746 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3748 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3749 - ensure len(uuid) < 37 in lmc (1171)
3750 - fix ia64 OOPS in llog_test (2255)
3751 - zero end of page at obdfilter for partial page writes (2648)
3752 - don't leave stale dentries around after renames (bug 2428)
3753 - fix timeouts when evicting a client with a single lock held (2642)
3754 - set deadline for the initial HELLO message to drain (2634)
3755 - print out dotted-quad IP addresses in the socknal (2302)
3757 - additional debugging for MDS client eviction problem (2443)
3758 - fix mkfsoptions support for osts (2603, 2604)
3760 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3763 - add series for the vanilla 2.6.0 kernel
3764 - add series for the vanilla 2.4.24 kernel
3765 - add series for a cray x86/64 UL kernel drop
3766 - fix xattr patches for the vanilla 2.4.19 series
3768 - generate true UUIDs in lmc (1171)
3769 - have portals stack dumping break in UML (2466)
3770 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3771 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3772 - wake callback waiting threads on client eviction (2460)
3773 - Add --ptldebug and --subsystem to lmc (1719)
3774 - update assertion to allow safe interrupt allocation
3775 - set rq_no_resend for cancel requests (2432)
3776 - recalculate ptlrpcd timeout after resend (2494)
3777 - call vfs_rmdir when removing pending directories (2368)
3778 - fix renaming a file to itself (2429)
3779 - lmc creates a default one-stripe lov (2454)
3780 - expand procfs space to handle large clusters (2326)
3781 - increase UML stack to avoid overflow
3782 - update lconf's list of debug and subsystem masks
3783 - fix lfs find --obd (2510)
3784 - /proc tunable for disabling filter read caching (2591)
3785 - stop rpm packages from altering slapd.conf (2301)
3786 - disable nagle in the socknal under 0conf (2578)
3787 - choose mds inode size based on stripe count (2572)
3788 - fix kernel-source rpm problems (2516)
3790 - add --disable-doc to avoid pdf generation (2421)
3791 - update documentation, tests, type-os, comments
3792 - avoid format warnings on ia64
3793 - remove the TOE NAL
3794 - tiny code cleanups by removing unused fields
3796 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3799 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3800 - remove the most blatant lies from BUILDING file (2371)
3801 - change default debug level to reasonable production setting
3802 - reduce client side cache size to reduce cache flush time
3803 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3804 - make TCP zerocopy and pinger support enabled by default (2476)
3805 - sync writes completed after process exits caused crashes (2319)
3806 - maintain correct mount count on the MDS (2356)
3807 - backout 1557, because 2316 wasn't really fixed
3808 - better file I/O statistics gathering in /proc
3809 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3810 - another kernel patch to fix zero-copy TCP function export
3811 - don't take duplicate lock when processing re-sent getattr (2420)
3812 - lctl uses obd_self_export instead of creating new conn (2353)
3813 - MDS/OST recovery case which requires object creation asserted (2425)
3814 - move lfs from /usr/sbin to /usr/bin in packages
3815 - fix race between mds_client_add and mds_client_free (2417)
3816 - use kmalloc instead of slabs in portals (2430)
3817 - don't create duplicate records when a failover MDS is present (2442)
3818 - remove unnecessary mount age check (2332)
3819 - don't remove directory inodes from locks prematurely (2451)
3820 - don't break if MDS service name is the same as hostname (2103)
3821 - fix races in client write RPC generation when cache full (2482)
3823 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3826 - remove now-unused request->rq_obd (278)
3827 - if an allocation fails, print out how much memory we've used (1933)
3828 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3829 - add the "configurable stack size" patch to most series files (1256)
3830 - ability to write large log records, for 100+ OST configs (2306)
3831 - fix NULL deref when filter_prep fails (2314)
3832 - fix operator precedence error in filter_sync
3833 - dynamic allocation of socknal TX descriptors (2315)
3834 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3835 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3836 - ensure that CWARN messages in Portals always get to the syslog
3837 - __init/__exit are not for prototype decls (ldlm_init/exit)
3838 - x86-64 compile warning fixes
3839 - fix gateway LMC keyword conflict (2318)
3840 - fix MDS lock inversions in getattr/reint paths (1844)
3841 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3842 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3843 - DLM race condition prevented some lock evictions (2328)
3844 - ENOMEM detection and retry on socknal sends (2230)
3845 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3846 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3847 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3848 - better error messages when a client is rejected during recovery (1505)
3849 - avoid cancelling locks which were never granted, after failure (2330)
3850 - fix i_sem/journal inversion in mds_client_add (2333)
3851 - fix truncate/getattr lock cycle deadlock (2334)
3852 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3853 - fix two rare exit paths which could leak an l_lock() ref (2321)
3854 - fencepost error in MDS/OST orphan recovery (2226)
3855 - make log record alignment 8 bytes (1988)
3856 - lstripe now fails when requested offset > ost_count (2237)
3857 - ensure that all kernel series have a complete list.h (1607)
3858 - fix crashes in special-file operations (2316)
3859 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3861 - allow configurable automake binary, for testing new versions
3862 - small update to the lfs documentation
3864 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3866 * fix negative export reference count in fsfilt_sync (2312)
3868 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3869 * release candidate 0.9.1
3871 - orphans are moved into the PENDING directory for possible recovery
3872 - replayed opens now open by fid for orphan/rename safety (1042)
3873 - last close of an orphan inode generates a transno (683)
3874 - chdir() and mount() now pin the directory entry (1020)
3875 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3876 - discard very old requests without processing them (1502)
3877 - remove some common, well-understood CERRORs (1505)
3878 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3879 - clear "grant" flags in OST replies until OST grant code lands (1644)
3880 - fix read performance by not clobbering i_blksize on client (1598)
3881 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3882 - make LRU size a /proc tunable, clears locks when reduced (707)
3883 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3884 - server locks take references on exports now (1558)
3885 - build fixes for 2.4.20-rh trees (1663)
3886 - return an error from lov_create if all OSCs are inactive (1751)
3887 - fix import levels when a reconnect happens without a timeout (1597)
3888 - exit early from mds_open if we get a lookup error (1749)
3889 - partial page read at EOF wouldn't wait for disk before sending (1642)
3890 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3891 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3892 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3893 - write/truncate lock inversion (1639)
3894 - Don't auto-load obdclass, portals modules during cleanup (1495)
3895 - fix timestamps from jumping to "now" (1763)
3896 - extra journal assertions (1648)
3897 - add an extra multiunlink test (1771)
3898 - fix read_record/write_record API (1776)
3899 - fix leak of offset_extent, possible incorrect i_size later (1772)
3900 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3901 - Debugging for the kqswnal_get_idle_tx problems (1820)
3902 - Allow recovery to be attempted multiple times (1536)
3903 - Write out MDS last_rcvd file after it is first created (1600)
3904 - Fix tx_descriptor leak in failed transmit situations (1827)
3905 - ext3 journaling fixes for assertion failure after IO error (1871)
3906 - class_export_put() on freed export after completion AST error (1896)
3907 - Fix revalidate looping in VFS (1322)
3908 - Don't access a freed export during MDS_REINT timeout (1521)
3909 - Add open-unlink recovery support on the MDS (1673,1764)
3910 - Return an error if no MDS data was read from last_rcvd (1946)
3911 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3912 - Don't setup a disk device that doesn't match exported UUID (317)
3913 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3914 - avoid committing NULL handle in force close
3915 - local.sh is now a one-stripe LOV configuration
3916 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3917 - fix ext3 htree duplicate directory entry corruption (1516)
3918 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3919 - update to new LOV EA format (2097)
3920 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3922 2003-06-15 Phil Schwan <phil@clusterfs.com>
3925 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3926 - per-import/export recovery handling (958, 931, 959)
3927 - multiple last-rcvd slots, for serving multiple FSes (949)
3928 - connections are again shared between multiple imp/exports (963, 964)
3929 - "umount -f" would hang if any requests needed to be sent (393, 978)
3930 - avoid pinning large req buffer by copying for queued messages (989)
3931 - add "uuid" to "lctl device" command to help upcalls (991)
3932 - "open" RPCs with transnos would confuse recovery counters (1037)
3933 - do proper endian conversion of all wire messages (288, 340, 891)
3934 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3935 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3936 - fix ptlrpc_connection leak in target_handle_connect (1174)
3937 - fix import refcounting bug in OST and MDS cleanup (1134)
3938 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3939 - fix very unlikely obd_types race condition (501)
3940 - remove osc_open hack for echo_client (1187)
3941 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3942 - a failure in read_inode2 leads to deadlock (1139)
3943 - cancel ack-locks as soon as transaction is committed (1072)
3944 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3945 - make sure to commitrw after any preprw to avoid deadlock (1162)
3946 - failing to execute a file in a lustre FS would lock inode (1203)
3947 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3948 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3949 - obd_uuid2tgt would walk off the end of the list (1255)
3950 - on IA64 the osc would give portals incorrect bulk size (1258)
3951 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3952 - fix lock inversion caused by new llite matching code (1282)
3953 - limit the number of dirty pages on a client to 10MB (1286)
3954 - timed out locks were not being corrected cancelled (1289)
3955 - fix O_DIRECT above 4GB on IA-32 (1292)
3956 * major user-visible changes
3957 - fail out/fail over policy now controlled by the upcall (993)
3959 - add OBD_PING to check server availability and failure (954)
3960 - lustre messages are now sent in sending host order (288, 340, 891)
3961 - add eadatalen to MDS getattr reply (340)
3962 - OST read replies may contain second buffer, with per-page status (593)
3964 2003-03-11 Phil Schwan <phil@clusterfs.com>
3967 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3968 - failed open()s could cause deadlock; fixed (867, 869)
3969 - stop cancelling OST locks when files are closed (481)
3970 - overlapping XID spaces caused network corruption (851, 853)
3971 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3972 - setattr_raw added, to do single-RPC, server-side setattrs
3973 - lmc/lconf syntax change for OST UUIDs
3974 - fix crashy race condition between ptlrpc_free_req and osc_close
3975 - don't use request in mdc_enqueue if we hit a timeout (889)
3976 - don't set the inode i_size for regular files from the MDS (896)
3977 - handle out of order completion AST (842)
3978 - don't LBUG if a lock request times out after receiving AST (913)
3979 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3980 - if a bad lock AST arrives, send an error instead of dropping entirely
3981 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3982 - fix leak in bulk IO when only partially completed (899, 900, 926)
3983 - fix O_DIRECT for ia64 (55)
3984 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3985 - C-z after timeout could hang a process forever; fixed (977)
3987 - client-side I/O cache (678, 924, 929, 941, 970)
3989 - READPAGE and SETATTRs which don't take server-side locks get
3992 2003-02-11 Phil Schwan <phil@clusterfs.com>
3995 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3996 - Fix fsfilt_extN_readpage() to read a full page of directory
3997 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3998 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3999 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4000 - Most intent operations were removed, in favour of a new RPC mode
4001 that does a single RPC to the server and bypasses most of the VFS
4002 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4003 - Aggressively cancel local locks on DLM servers
4004 - mds_reint_unlink sends EA to the client if it's the last nlink.
4005 client uses that EA to unlink OST objects.
4006 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4007 - recursive symlinks were fixed (439)
4008 - fixed NULL deref in DEBUG_REQ
4009 - filter_update_lastobjid no longer calls sync, which annoyed extN
4010 - fixed multi-client small-writes to a single file problem (445)
4011 - fixed mtime updates during file writes (607)
4012 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4013 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4014 - fixed Linux OST size reporting problem (444, 656)
4015 - OST now updates object mtime with writes or setattr (607, 619)
4016 - client verifies file size before zeroing page past EOF (445)
4017 - OST now writes last allocated objid to disk with allocation (108)
4018 - LOV on echo now works (409)
4020 - mds_reint_unlink sends a new buffer, with the EA included. this
4021 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4022 set if a regular file was being unlinked, and it was the last link
4023 - use PtlGet from the target for bulk writes (315)
4024 - OST now updates object mtime with writes or setattr (607, 619)
4025 - LDLM now has a grant-time callback to revalidate locked items, if
4027 - Many MDS operations were reorganized to combat race conditions
4029 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4030 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4032 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4035 - Fully reactivate OST imports after reconnection (512, others)
4036 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4037 - More graceful error handling for truncating on dead OST (515)
4038 - Don't error out unless we're actually accessing dead stripes (474)
4039 - Fix garbage sizes when stripes are missing (410)
4040 - LRU counters were broken, causing constant lock purge (433, 432)
4041 - garbage on read from stripes with failed OSTs (441)
4042 - mark OSCs as active before reconnecting during recovery (438)
4043 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4044 - lfind did not preserve OST order in output (443)
4045 - symlinks cause hung clients, incorrect data (439)
4046 - stop dereferencing request after dropping refcount (457)
4047 - don't LASSERT(spin_is_locked) on non-SMP (455)
4048 - fixes for many rename() bugs
4049 - fstat didn't correctly synchronize attributes (399)
4050 - server must handle lock cancellation during blocking AST prep (487)
4051 - bulk descriptors were free()d too soon (511)
4052 - fix paths in lconf, which would load incorrect modules (451, 507)
4053 - fix confusing lconf 'host not found' error message (386)
4054 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4055 - fix race condition in mdc_blocking_ast() for inode access (526)
4056 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4057 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4058 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4059 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4060 - fix lstripe in several ways (kernel side) (527)
4061 - fix request leak in ldlm_cli_enqueue (262)
4062 - incorrect OSC was marked inactive after OST failure
4063 - call mds_fs_cleanup before unmounting filesystem (524)
4064 - fix races between taking ns_lock and ldlm_lock_change_resource
4065 - fix races updating LOV export open file list
4066 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4067 - fix recovery NULL deref in ldlm_cli_cancel_unused
4068 - fix some DLM races by using new hash table for lock handles (419)
4069 - permit the client to specify desired inodes, at replay
4070 - duplicate requests when we queue them for replay reintegration
4071 - fix last_rcvd offset calculation
4072 - sync after each recovered transaction, so we always make progress
4073 - never, not always, ERESTART requests without transnos
4074 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4076 - skip replay if the MDS says that the client is already connected
4077 - don't check for a recovery-enabled export to match lctl's UUID
4078 - don't INC_USE_COUNT for phantom exports
4079 - don't crash when cleaning up phantom exports (567)
4080 - don't double-finish or set replay data for errored mdc_open requests
4081 - abort requests when they time out, so we don't get old replies
4082 - send/receive replies for AST messages again
4083 - if the client says that it doesn't have the lock, cancel it on the
4085 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4086 mark it as destroyed, it will all work out when decref is called
4087 - fix module use counts (22, 581)
4089 - ASTs now expect a reply (server cancels lock on error reply)
4091 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4094 - fix many simultaneous client startup (392)
4095 - fix dentry->d_it clobbering
4096 - credentials weren't being shipped for readdir/getattr operations
4097 - remove invalid assertions triggered during some concurrent MD
4099 - proper Lustre versions added (336, 389)
4100 - fix memory leak for create error case (398)
4101 - fix LOV locking bug that would get cli/srv out of sync
4102 - fix echo client over LOV (409)
4103 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4104 - fix double-O_EXCL intent crash (424)
4105 - avoid sending multiple lock CANCELs (352)
4107 - MDS can do multi-client recovery (modulo bugs in new code)
4109 - many updates, edits, cleanups
4111 2002-11-18 Phil Schwan <phil@clusterfs.com>
4114 - fix null d_it dereference (346)
4115 - fix full OST/dbench hang (333)
4116 - fix permission problem with file removal (286)
4117 - fix removal of OSCs from LOV when they fail
4118 - fix NULL deref during bulk timeout (214)
4119 - fix problems related to multiple filesystems on one MDS (241)
4120 - fixed serious subtle metadata locking bugs
4121 - free locks on clients when inodes are removed due to memory
4123 - fix inode pointer in lock data (285)
4124 - partial support for multiple MDS on a single host (241)
4125 - data locks weren't cancelled at clear_inode time (290, 311)
4126 - intent locks could lead to unbounded lock growth (205)
4127 - added a maximum lock count, an LRU list, and a flusher
4128 - fix multiple rename (365)
4129 - properly abstracted the echo client
4130 - OSC locked 1 byte too many; fixed
4131 - rewrote brw callback code:
4132 - fixed recovery bugs related to LOVs (306)
4133 - fixed too-many-pages-in-one-write crash (191)
4134 - fixed (again) crash in sync_io_timeout (214)
4135 - probably fixed callback-related race (385)
4137 - Add capability to MDS protocol
4138 - LDLM cancellations and callbacks on different portals
4140 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4143 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4144 - defer open object destruction to close time (601981, 138)
4145 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4146 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4147 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4148 - don't call obd_setattr in ll_file_release for destroyed objects
4149 * protocol change to lustre_msg: move |version| and add |flags|
4150 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4151 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4152 * added replay of create, unlink, link and rename operations during
4153 MDS failover; recovery should be much more robust now
4154 * remove failed OSCs from LOVs (only lov_create uses this so far)
4155 * the lustre-HOWTO was brought (more) up to date (582544)
4157 2002-10-23 Phil Schwan <phil@clusterfs.com>
4160 - in-use dentries weren't being reused properly (617851)
4161 - prevent multiple LDLM setup (599178)
4162 - fix LOV size calculations for truncate (617853)
4163 - fix client handling of MDS intent errors (POSIX)
4164 - fix permission bug in lovstripe.c test (624321)
4165 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4166 - truncate past end of file could corrupt data
4167 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4168 - a race in recovery could return ETIMEDOUT to apps (623947)
4169 - building outside the source directory was fixed
4170 * the lustre-HOWTO was brought (more) up to date (582544)
4171 * major progress was made on recovery functionality
4173 2002-10-10 Phil Schwan <phil@clusterfs.com>
4176 - recovery deadlock fix
4177 - rm -rf causes LBUG fix (617817)
4178 - file open by multiple tasks fix (618962)
4179 - directory permissions bugs (602707 and 620007)
4180 - journal_stop fixed with locking (611313)
4181 - O_APPEND failures resolved (618273, perhaps 614459)
4182 - lconf PATH fix (619770)
4183 - IA64 build fix (621450)
4184 - RPC buffer sizes scale with amount of memory
4186 2002-10-01 Phil Schwan <phil@clusterfs.com>
4189 - locks would be cancelled without throwing away data pages,
4190 resulting in inconsistent data (605627)
4191 - inode attributes were not always being refreshed (605627, 612449)
4192 - lconf now continues to cleanup after lctl reports an error
4193 - MDS now enforces user permissions (602707)
4194 - lprocfs cleanup fixed, but not yet enabled (614157)
4195 - fixed infinite server hang, should a client not respond to an AST
4196 - avoid going into recovery if user calls readlink() with a buffer
4197 that's too small (613941)
4198 - AST RPCs no longer require replies (614867) -- this may be changed
4199 - don't crash server if client sends an IOV that's too big (611336)
4200 - fixed lock conversion deadlock (611892)
4201 - fixed the following of symlinks (614622)
4202 * recovery: the server can remove locks from a client that dies, other
4203 clients can make progress
4204 * more extN patch fixes
4205 * compile-time configurable ptlrpc buffer allocations
4207 - collaborative read cache document
4208 - Lustre Lite Performance CDR document-in-progress
4210 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4213 - fix typo in patch-2.4.18
4215 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4218 - clear ptlrpc request each time in handle_incoming_request()
4219 - unlink of files now destroys the object on the OST
4221 2002-09-19 Peter Braam <braam@clusterfs.com>
4223 * add hard link support
4224 * change obdfile creation method
4225 * kernel patch changed
4227 2002-09-19 Peter Braam <braam@clusterfs.com>
4230 - stack overflow bug in extN fixed
4232 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4234 * documentation updates
4235 - add man pages for config tools
4236 - update tests/README to describe testing with new config tools
4237 - finish metadata API descriptions
4238 * bug fixes and cleanups
4239 - statfs workaround for 16TB limit
4240 - LOV stripe allocation improved, can stripe on subset of OSTs
4241 - LOV file size/IO offset was wrong for files > 4GB in size
4242 - object EA data was being dropped, caused files to be unreadable
4243 - memory overflow with non-LOV OST caused memory corruption
4244 - fixed regression tests to work with new config tools, obdfilter
4245 - fixed bug when directory size became larger than 1 block
4246 - fixed bug (for single client case) when PWD was deleted
4247 - invalidate local directory pages when doing intent-based ops
4248 - avoid LDLM oops when lock callback contained bad data
4250 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4252 * documentation updates
4253 * bug fixes and cleanups
4254 - configuration tools
4257 - 64-bit compile warnings
4258 - 64-bit internal statfs data
4260 * test_brw on persistent OST devices
4262 * lprocfs (disabled)
4264 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4266 * documentation updates
4267 * bug fixes and cleanups
4268 * configuration tools
4270 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4273 * many small fixes to 0_5_4
4274 * io/network handling
4275 * thinkos in MDS operations
4277 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4280 * crucial basic fixes to 0.5.3
4281 * IOR, Iozone work over Elan
4284 2002-08-07 Phil Schwan <phil@clusterfs.com>
4285 * version 0_5_3, our first alpha
4286 * we use the new Portals iovs
4287 * documentation updates
4288 * bug fixes and cleanups
4289 * small changes in the DLM wire protocol
4291 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4292 * version 0_5_1 with some initial stability,
4293 * locking on MD and file I/O.
4294 * documentation updates
4295 * several bug fixes since 0.5.0
4296 * small changes in wire protocol
4298 2002-07-18 Phil Schwan <phil@clusterfs.com>
4300 * delivered as Lustre Light Alpha
4301 * fixed a crash after handling invalid MDS requests
4302 * fixed directory pages for architectures with non-4k pages sizes
4304 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4305 * release version v0_4_4
4306 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4308 * Fixes a number of interruption problems with OST operations.
4309 * Update documentation for portals header changes
4310 * Move all wire protocol structs/defines to lustre_idl.h
4311 * Fixes symlink length bug.
4312 * Add tcpdump to repository.
4314 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4315 * release version v0_4_3
4316 * Fixes statfs for inodes on extN.
4317 * Fixes bug in runtests which would delete /etc/hosts.
4318 * Use 64-bit object IDs wherever possible (not into VFS though)
4319 Remove ost_get_info, which is unused by lustre, and out of date.
4321 2002-07-03 Peter Braam <braam@clusterfs.com>
4322 * release version v0_4_2 Fixes a lookup error (type not passed)
4323 * move forward to head of Portals
4324 * move forward to latest Lustre kernel
4326 2002-06-25 Peter Braam <braam@clusterfs.com>
4327 * release version v0_4_1. Hopefully stable on single node use.