4 2.6.9-55.0.9.EL (RHEL 4),
5 2.6.16.53-0.8 (SLES 10),
6 2.6.18-8.1.14.EL5 (RHEL 5).
7 2.6.18.8 vanilla (kernel.org)
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.9-42.0.10.EL (RHEL 4),
11 2.6.16 - 2.6.21 vanilla (kernel.org)
12 * Recommended e2fsprogs version: 1.40.2-cfs1
13 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
17 Description: Remove 2.4 kernel support from lustre 1.8.0
18 Details : Remove 2.5.0 check from quota_check.c, quota_ctl.c,
19 quota_interface.c in quota subsystem.
20 Remove 2.4 kernel check from obd_mount.c, linux/linux-sysctl.c in
22 Remove lock_24kernel from fsfilt_ext3.c fsfilt_reiserfs.c
23 lvfs_linux.c upcall_cache.c in lvfs subsystem.
24 Remove 2.5.0 check from dcache.c dir.c file.c llite_internal.h
25 llite_lib.c llite_mmap.c lproc_llite.c namei.c rw.c symlink.c
27 Remove 2.5.0 check from handler.c lproc_mds.c mds_fs.c
28 mds_internal.h mds_lib.c mds_open.c mds_reint.c mds_xattr.c
30 Remove 2.4 kernel check from lproc_osc.c, osc_request.c in osc
32 Remove 2.4 kernel check from filter.c, lproc_obdfilter.c in
34 Remove 2.4 kernel code from lustre/obdecho/echo_client.c.
35 Remove 2.4 kernel check from lustre/mgs/lproc_mgs.c,
37 Remove 2.4 kernel code from lustre/mgc/mgc_request.c.
38 Remove 2.4 kernel check from lustre/mdt/mdt_lproc.c.
39 Remove 2.4 kernel code from lustre/mdc/mdc_request.c.
40 Remove 2.4 kernel check from lustre/lov/lproc_lov.c.
41 Remove 2.4 kernel check from lustre/lmv/lproc_lmv.c.
43 Severity : enhancement
45 Description: Build SLES10 patchless client fails
46 Details : The configure was broken by run ./configure with
47 --with-linux-obj=.... argument for patchless client. When the
48 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
49 can't search header adequately. Use absolute path such as
50 -I($LINUX)/include instead.
54 Description: interrupt oig_wait produce painc on resend.
55 Details : brw_redo_request can be used for resend requests from ptlrpcd and
56 private set, and this produce situation when rq_ptlrpcd_data not
57 copyed to new allocated request and triggered LBUG on assert
58 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
59 ptlrpcd set if request is changed and can be safety changed to use
64 Description: LASSERT_{REQ,REP}SWAB macros are buggy
65 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
66 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
67 these macros and replace them with their difinitions instead.
71 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
72 Details : Update which_patch & target file for SLES10 latest kernel.
75 Description: Remove client patches from SLES 10 kernel.
76 Details : This causes SLES 10 clients to behave as patchless clients
77 even on a Lustre-patched (server) kernel.
79 Severity : enhancement
81 Description: self-adjustable client's lru lists
82 Details : use adaptive algorithm for managing client cached locks lru
83 lists according to current server load, other client's work
84 pattern, memory activities, etc. Both, server and client
85 side namespaces provide number of proc tunables for controlling
88 Severity : enhancement
90 Description: light-weight GSS support
91 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
92 and reduce performance overhead.
94 Severity : enhancement
96 Description: Linux keyring support
97 Details : Support using service of Linux keyring for Lustre GSS internal
98 context refresh/cache mechanism.
102 Description: Fix errors in lfs documentation
103 Details : Fixes man pages
107 Description: don't use GFP_* in generic Lustre code.
108 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
112 Description: obdclass is limited by single OBD_ALLOC(idarray)
113 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
117 Description: warnings with build patchless client with vanila 2.6.19 and up
118 Details : change old ctl_table style and replace ctl_table/ctl_table_header
119 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
123 Frequency : Only for SLES
124 Description: sanity_quota fail test_1
125 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
126 file, which makes TSTID[2] inunique.
130 Frequency : Only for SLES9
131 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
132 Details : Update target/ChangeLog/which_patch .
136 Frequency : Only for test_10 in sanity-quota.sh
137 Description: a bug in quota test code
138 Details : checking a test flag in a wrong place
142 Frequency : Only for test_13 in sanity-quota.sh
143 Description: a bug in quota test script
144 Details : shouldn't check overfull allocated quota
148 Description: jbd statistics
149 Details : Port older jbd statistics patch for sles10
151 Severity : enhancement
153 Description: Add printing inode info into message about error in writepage.
158 Description: reply_lock_interpret crash due to race with it and lock cancel.
159 Details : Do not replay locks that are being cancelled. Do not reference
160 locks by their address during replay, just by their handle.
164 Frequency : When flocks are used.
165 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
166 Details : Flock locks might destroy just granted lock if it could be merged
167 with another existing flock, this is done in completion handler,
168 so teach ldlm_cli_enquque_fini that this is a valid case for
174 Description: Oops in read and write path when failing to allocate lock.
175 Details : Check if lock allocation failed and return error back.
179 Description: lstripe command fails for valid OST index
180 Details : The stripe offset is compared to lov->desc.ld_tgt_count
181 instead of lov->desc.ld_active_tgt_count.
185 Description: sanity.sh failed test 103
186 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
191 Description: df doesn't work properly if diskfs blocksize != 4K
192 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
196 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
197 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
201 Description: obd_health_check_timeout too short
202 Details : set obd_health_check_timeout as 1.5x of obd_timeout
206 Description: llapi_file_create() does not allow some changes
207 Details : add llapi_file_open() that allows specifying the mode and
208 open flags, and also returns an open file handle.
212 Description: merge and cleanup kernel patches.
213 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
217 Description: Add journal checksum support.(Kernel part)
218 Details : The journal checksum feature adds two new flags i.e
219 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
220 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
221 indicates that the commit block contains the checksum for
222 the blocks described by the descriptor blocks. Now commit
223 record can be sent to disk without waiting for descriptor
224 blocks to be written to disk. This behavior is controlled
225 using JBD2_FEATURE_ASYNC_COMMIT flag.
229 Description: OSS needs mutliple precreate threads
230 Details : Add ability to start more than one create thread per OSS.
234 Description: Sanity.sh test_65k failed: setstripe should have succeeded
235 Details : execute lfs setstripe on client
239 Description: mds_obd_create error creating tmp object
240 Details : When the user sets quota on root, llog will be affected and can't
241 create files and write files.
244 Frequency : Always on ia64 patchless client, and possibly others.
246 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
247 Details : This allows the patchless client to be loaded on architectures
252 Description: RedHat Update kernel for RHEL5
253 Details : Add the kernel config file for RHEL5.
257 Description: RedHat Update kernel for RHEL5
258 Details : Modify the kernel config file more closer RHEL5.
262 Description: Build failure against Centos5 (RHEL5)
263 Details : Define PAGE_SIZE when it isn't present.
267 Description: test_7 on sanity-quota.sh can't be used on separate machines
268 Details : add facet to handle it
272 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
273 Details : with NFS, the anon dentry's parent was set to itself in
274 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
275 corresponding dentry other than use rec->ur_name.
277 Severity : enhancement
279 Description: lfs setstripe enhancement
280 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
284 Description: testing performance impact of enabling checksumming
285 Details : enable checksum by default, allow --disable-checksum
286 configure option and "-o nochecksum" mount option
290 Description: lustre support for RHEL5
291 Details : Add support for RHEL5.
295 Description: Client eviction due to failover config
296 Details : after a connection loss, the lustre client should attempt to
297 reconnect to the last active server first before trying the
298 other potential connections.
302 Description: when mds and osts use different quota unit(32bit and 64bit),
303 quota will be released repeatly.
304 Details : void sending multiple quota reqs to mds, which will keep the
305 status between the reqs.
309 Description: osts not allocated evenly to files
310 Details : change the condition to increase offset_idx
313 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
315 Description: Data corruption for OSTs that are formatted larger than 2TB
317 Details : When generating the bio request for lustre file writes the
318 sector number would overflow a temporary variable before being
319 used for the IO. The data reads correctly from Lustre (which
320 will overflow in a similar manner) but other file data or
321 filesystem metadata may be corrupted in some cases.
325 Description: Tune the kernel for good SCSI performance.
326 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
327 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
332 Description: rewrite ext2-derived code in obdclass/uuid.c
333 Details : rewrite inherited code (uuid parsing code from ext2 utils)
334 from scratch preserving functionality.
338 Description: "lfs find -obd UUID" prints directories
339 Details : "lfs find -obd UUID" will return all directory names instead
340 of just file names. It is incorrect because the directories
341 do not reside on the OSTs.
345 Description: use i_size_read and i_size_write in 2.6 port
346 Details : replace inode->i_size access with i_size_read/write()
349 Frequency : when removing large files
351 Description: scheduling issue during removal of large Lustre files
352 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
353 It causes scheduling issues when removing large files (17TB in the
357 Frequency : only with liblustre clients on XT3
359 Description: evictions taking too long
360 Details : allow llrd to evict clients directly on OSTs
363 Frequency : only on ppc
365 Description: /proc/fs/lustre/devices broken on ppc
366 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
367 We should make sure the type of 'index' is loff_t and then cast
368 explicitly as needed below. Do not assign an explicitly cast
372 Frequency : only for rhel5
374 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
375 Details : Modify the target file & which_kernel.
377 Severity : enhancement
379 Description: omit set fsid for export NFS
380 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
381 as FS_REQUIRES_DEV for avoid problems with generate fsid.
385 Frequency : Always, for kernels after 2.6.16
386 Description: Fix warning idr_remove called for id=.. which is not
388 Details : Last kernels save old s_dev before kill super and not allow
389 to restore from callback - restore it before call kill_anon_super.
393 Description: buffer overruns could theoretically occur
394 Details : llapi_semantic_traverse() modifies the "path" argument by
395 appending values to the end of the origin string, and a
396 overrun may occur. Adding buffer overrun check in liblustreapi.
400 Description: Fix error on 'ls .' at the top of the Lustre mount.
401 Details : Don't revalidate dentry if it is a root dentry.
405 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
406 Details : Modify vm-tunables-rhel4.patch.
410 Description: Kernel config for 2.6.18-vanilla.
411 Details : Modify targets/2.6-vanilla.target.in.
412 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
413 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
414 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
415 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
419 Description: improve handling recoverable errors
420 Details : if request processig with error which can be recoverable on server
421 request should be resend, otherwise page released from cache and
426 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
427 Details : Modify target file & which_patch.
428 A flaw was found in the IA32 system call emulation provided
429 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
430 value could be stored in the %RAX register, which could trigger an
431 out-of-bounds system call table access. An untrusted local user
432 could exploit this flaw to run code in the kernel
433 (ie a root privilege escalation). (CVE-2007-4573).
437 Description: O_DIRECT bypasses client statistics.
438 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
439 read_ahead_stats were updated. copied the stats section from
440 osc_send_oap_rpc() into async_internal().
444 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
448 Description: change order of libsysio includes
449 Details : '#include sysio.h' should always come before '#include xtio.h'
453 Description: handle "serious error: objid * already exists" more gracefully
454 Details : If LAST_ID value on disk is smaller than the objects existing in
455 the O/0/d* directories, it indicates disk corruption and causes an
456 LBUG(). If the object is 0-length, then we should use the existing
457 object. This will help to avoid a full fsck in most cases.
459 Severity : enhancement
461 Description: adapt the lustre_config script to support the upgrade case
462 Details : Add "-u" option for lustre_config script to support upgrading 1.4
463 server targets to 1.6 in parallel.
467 Description: To avoid grant space > avaible space when the disk is almost
468 full. Without this patch you might see the error "grant XXXX >
469 available" or some LBUG about grant, when the disk is almost
471 Details : In filter_check_grant, for non_grant cache write, we should
472 check the left space by if (*left > ungranted + bytes), instead
473 of (*left > ungranted), because only we are sure the left space
474 is enough for another "bytes", then the ungrant space should be
475 increase. In client, we should update cl_avail_grant only there
476 is OBD_MD_FLGRANT in the reply.
480 Description: Update RHEL 4 kernel to fix local root privilege escalation.
481 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
482 described in CVE-2007-4573. This problem could allow untrusted
483 local users to gain root access.
487 Frequency : when using O_DIRECT and quotas
489 Description: Incorrect file ownership on O_DIRECT output files
490 Details : block usage reported by 'lfs quota' does not take into account
491 files that have been written with O_DIRECT.
493 --------------------------------------------------------------------------------
495 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
497 * Support for kernels:
498 2.6.5-7.283 (SLES 9),
499 2.6.9-55.EL (RHEL 4),
500 2.6.16.46-0.14 (SLES 10),
501 2.6.18.8 vanilla (kernel.org)
502 * Client support for unpatched kernels:
503 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
504 2.6.16 - 2.6.22 vanilla (kernel.org)
505 * Due to recently discovered recovery problems, we do not recommend
506 using patchless RHEL 4 clients with this or any earlier release.
507 * Recommended e2fsprogs version: 1.39.cfs8
508 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
509 * Starting with this release, the ldiskfs backing filesystem required
510 by Lustre is now in its own package, lustre-ldiskfs. This package
511 should be installed. It is versioned separately from Lustre and
512 may be released separately in future.
516 Description: block reactivating mgc import until all deactivates complete
517 Details : Fix race when failing back MDT/MGS to itself (testing)
520 Frequency : at statup only
522 Description: mds_lov_synchronize race leads to various problems
523 Details : simultaneous MDT->OST connections at startup can cause the
524 sync to abort, leaving the OSC in a bad state.
526 Severity : enhancement
528 Description: add optional extra BUILD_VERSION info
529 Details : add a new environment variable (namely LUSTRE_VERS) which allows
530 to override the lustre version.
533 Frequency : 2.6.18 servers only
535 Description: ll_kern_mount() doesn't release the module reference
536 Details : The ldiskfs module reference count never drops down to 0
537 because ll_kern_mount() doesn't release the module reference.
542 Description: server LBUG when using old ost_num_threads parameter
543 Details : Accept the old ost_num_threads parameter but warn that it
544 is deprecated, and fix an off-by-one error that caused an LBUG.
549 Description: Transient SCSI error results in persistent IO issue
550 Details : iobuf->dr_error is not reinitialized to 0 between two
554 Frequency : sometimes when underlying device returns I/O errors
556 Description: OSTs not going read-only during write failures
557 Details : OSTs are not remounted read-only when the journal commit threads
558 get I/O errors because fsfilt_ext3 calls journal_start/stop()
559 instead of the ext3 wrappers.
563 Description: poor connect scaling with increasing client count
564 Details : Don't run filter_grant_sanity_check for more than 100 exports
565 to improve scaling for large numbers of clients.
568 Frequency : SLES10 only
570 Description: sanity-quota.sh quotacheck failed: rc = -22
571 Details : Quotas cannot be enabled on SLES10.
574 Frequency : liblustre clients only
576 Description: getdirentries does not give error when run on compute nodes
577 Details : getdirentries does not fail when the size specified as an argument
578 is too small to contain at least one entry
580 Severity : enhancement
582 Description: Add LNET router traceability for debug purposes
583 Details : If a checksum failure occurs with a router as part of the
584 IO path, the NID of the last router that forwarded the bulk data
585 is printed so it can be identified.
590 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
591 Details : Due to a race condition, liblustre clients were occasionally
594 Severity : enhancement
596 Description: lfs setstripe use optional parameters instead of postional
599 Severity : enhancement
601 Description: Nanosecond timestamp support for ldiskfs
602 Details : The on-disk ldiskfs filesystem has added support for nanosecond
603 resolution timestamps. There is not yet support for this at
604 the Lustre filesystem level.
607 Frequency : during server recovery
609 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
610 Details : request with rq_no_resend flag not awake l_wait_event if they get a
614 Frequency : nfs export on patchless client
616 Description: connectathon hang when test nfs export over patchless client
617 Details : Disconnected dentry cannot be found with lookup, so we do not need
618 to unhash it or make it invalid
621 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
622 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
623 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
628 Description: open req refcounting wrong on reconnect
629 Details : If reconnect happened between getting open reply from server and
630 call to mdc_set_replay_data in ll_file_open, we will schedule
631 replay for unreferenced request that we are about to free.
632 Subsequent close will crash in variety of ways.
633 Check that request is still eligible for replay in
634 mdc_set_replay_data().
639 Description: disable writes to filesystem when reading health_check file
640 Details : the default for reading the health_check proc file has changed
641 to NOT do a journal transaction and write to disk, because this
642 can cause reads of the /proc file to hang and block HA state
643 checking on a healthy but otherwise heavily loaded system. It
644 is possible to return to the previous behaviour during configure
645 with --enable-health-write.
647 Severity : enhancement
649 Description: 64-bit inode version
650 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
651 to the inode. This will be required for version-based recovery.
656 Description: MDS fails to start if a duplicate client export is detected
657 Details : in some rare cases it was possible for a client to connect to
658 an MDS multiple times. Upon recovery the MDS would detect this
659 and fail during startup. Handle this more gracefully.
661 Severity : enhancement
663 Description: Add -o localflock option to simulate old noflock
665 Details : This will achieve local-only flock/fcntl locks
671 Description: log_commit_thread vs filter_destroy race leads to crash
672 Details : Take import reference before releasing llog record semaphore
677 Description: Wrong request locking in request set processing
678 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
679 imp_delayed_list, in this place should be used imp_lock.
682 Frequency : when reconnection
684 Description: Grant Leak when osc reconnect to OST
685 Details : When osc reconnect ost, OST(filter) should check whether it
686 should grant more space to client by comparing fed_grant and
687 cl_avail_grant, and return the granted space to client instead
688 of "new granted" space, because client will call osc_init_grant
689 to update the client grant space info.
692 Frequency : when client reconnect to OST
694 Description: Grant Leak when osc do resend and replay bulk write
695 Details : When osc reconnect to OST, OST(filter)should clear grant info of
696 bulk write request, because the grant info while be sync between
697 OSC and OST when reconnect, and we should ignore the grant info
698 these of resend/replay write req.
703 Description: Grant space more than avaiable left space sometimes.
704 Details : When then OST is about to be full, if two bulk writing from
705 different clients came to OST. Accord the avaliable space of the
706 OST, the first req should be permitted, and the second one
707 should be denied by ENOSPC. But if the seconde arrived before
708 the first one is commited. The OST might wrongly permit second
709 writing, which will cause grant space > avaiable space.
712 Frequency : when client is evicted
714 Description: Grant might be wrongly erased when osc is evicted by OST
715 Details : when the import is evicted by server, it will fork another
716 thread ptlrpc_invalidate_import_thread to invalidate the
717 import, where the grant will be set to 0. While the original
718 thread will update the grant it got when connecting. So if
719 the former happened latter, the grant will be wrongly errased
720 because of this race.
725 Description: Checking Stale with correct fid
726 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
727 to check whether it is stale, because sometimes, we want the
728 enqueue happened anyway, and op_data.fid2 will not be initialized.
730 Severity : enhancement
732 Description: update patchless client
733 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
736 Frequency : only with 2.4 kernel
738 Description: random memory corruption
739 Details : size of struct ll_inode_info is to big for union inode.u and this
740 can be cause of random memory corruption.
745 Description: Memory leak in recovery
746 Details : Lov_mds_md was not free in an error handler in mds_create_object.
747 It should also check obd_fail before fsfilt_start, otherwise if
748 fsfilt_start return -EROFS,(failover mds during mds recovery).
749 then the req will return with repmsg->transno = 0 and rc = EROFS.
750 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
751 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
752 be freed no matter whether fsfilt_commit success or not.
755 Frequency : only with huge count clients
757 Description: Prevents from taking the superblock lock in llap_from_page for
759 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
760 for llap_from_page call in ll_removepage prevents from taking the
761 superblock lock for a soon died page.
766 Description: Not check open intent error before release open handle
767 Details : in some rare cases, the open intent error is not checked before
768 release open handle, which may cause
769 ASSERTION(open_req->rq_transno != 0), because it tries to release
770 the failed open handle.
775 Description: Set cat log bitmap only after create log success.
776 Details : in some rare cases, the cat log bitmap is set too early. and it
777 should be set only after create log success.
781 Description: Accessing a block bevice can re-enable I/O when Lustre is
782 tearing down a device.
783 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
788 Description: service threads may hog cpus when there are a lot of requests
790 Details : Insert cond_resched to give other threads a chance to use some of
796 Description: the cat log was not initialized in recovery
797 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
798 unlink log on mds will not be initialized until mds post
799 recovery. And also in mds post recovery, the unlink log will
800 initialization will be done asynchronausly, so there will be race
801 between add unlink log and unlink log initialization.
805 Description: brw_stats were being printed incorrectly
806 Details : brw_stats were being printed as log2 but all of them were not
807 recorded as log2. Also remove some code duplication arising from
808 filter_tally_{read,write}.
812 Frequency : rare, only in recovery.
813 Description: ASSERTION(req->rq_type != LI_POISON) failed
814 Details : imp_lock should be held while iterating over imp_sending_list for
815 prevent destroy request after get timeout in ptlrpc_queue_wait.
819 Description: replay-single.sh test 52 fails
820 Details : A lock's skiplist need to be cleanup when it being unlinked
821 from its resource list.
825 Description: Short directio read returns full requested size rather than
827 Details : Direct I/O operations should return actual amount of bytes
828 transferred rather than requested size.
830 Severity : enhancement
832 Description: metadata RPC reduction (e.g. for rm performance)
833 Details : decrease the amount of synchronous RPC between clients and servers
834 by canceling conflicing lock before the operation on the client side
835 and packing thier handles into the main operation RPC to server.
837 Severity : enhancement
839 Description: Async OSC create to avoid the blocking unnecessarily.
840 Details : If a OST has no remain object, system will block on the creating
841 when need to create a new object on this OST. Now, ways use
842 pre-created objects when available, instead of blocking on an
843 empty osc while others are not empty. If we must block, we block
844 for the shortest possible period of time.
846 --------------------------------------------------------------------------------
848 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
853 Frequency : on some architectures
855 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
856 Details : Uninitialized flags sometimes cause configuration commands to
860 Frequency : patchless clients only
862 Description: missing __iget() symbol export
863 Details : The __iget() symbol export is missing. To avoid the need for
864 this on patchless clients the deathrow inode reaper is turned
865 off, and we depend on the VM to clean up old inodes. This
866 dependency was during via the fix for bug 12181.
868 --------------------------------------------------------------------------------
870 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
872 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
873 INTEROPERATE with older versions automatically. In many cases a
874 special upgrade step is needed. Please read the
875 user documentation before upgrading any part of a live system.
876 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
877 version will not interoperate with 1.6 betas before beta5 (1.5.95).
878 * WARNING: Lustre configuration and startup changes are required with
879 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
885 Severity : enhancement
887 Description: MountConf
888 Details : Lustre configuration is now managed via mkfs and mount
889 commands instead of lmc and lconf. New obd types (MGS, MGC)
890 are added for dynamic configuration management. See
891 https://mail.clusterfs.com/wikis/lustre/MountConf for
894 Severity : enhancement
896 Description: dynamic OST addition
897 Details : OSTs can now be added to a live filesystem
899 Severity : enhancement
901 Description: startup order invariance
902 Details : MDTs and OSTs can be started in any order. Clients only
903 require the MDT to complete startup.
905 Severity : enhancement
907 Description: parallel, asynchronous orphan cleanup
908 Details : orphan cleanup is now performed in separate threads for each
909 OST, allowing parallel non-blocking operation.
911 Severity : enhancement
913 Description: optimized stripe assignment
914 Details : stripe assignments are now made based on ost space available,
915 ost previous usage, and OSS previous usage, in order to try
916 to optimize storage space and networking resources.
918 Severity : enhancement
920 Description: Permanently set tunables
921 Details : All writable /proc/fs/lustre tunables can now be permanently
922 set on a per-server basis, at mkfs time or on a live
925 Severity : enhancement
927 Description: Lustre message v2
928 Details : Add lustre message format v2.
930 Severity : enhancement
932 Description: client OST exclusion list
933 Details : Clients can be started with a list of OSTs that should be
934 declared "inactive" for known non-responsive OSTs.
938 Description: SPEC SFS validation failure on NFS v2 over lustre.
939 Details : Changes the blocksize for regular files to be 2x RPC size,
940 and not depend on stripe size.
942 Severity : enhancement
944 Description: Multiple MD RPCs in flight.
945 Details : Further unserialise some read-only MDS RPCs - learn about intents.
946 To avoid overly-overloading MDS, introduce a limit on number of
947 MDS RPCs in flight for a single client and add /proc controls
948 to adjust this limit.
950 Severity : enhancement
952 Description: client read/write statistics
953 Details : Add client read/write call usage stats for performance
954 analysis of user processes.
955 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
956 file access. extents_stats shows chunk size distribution.
957 extents_stats_per_process show chunk size distribution per
960 Severity : enhancement
962 Description: mds statistics
963 Details : Add detailed mds operations statistics in
964 /proc/fs/lustre/mds/*/stats.
968 Description: Failure of copying files with lustre special EAs.
969 Details : Client side always return success for setxattr call for lustre
970 special xattr (currently only "trusted.lov").
975 Description: Refcount LNET uuids
976 Details : The global LNET uuid list grew linearly with every startup;
977 refcount repeated list entries instead of always adding to
980 Severity : enhancement
982 Description: Dynamic service threads
983 Details : Within a small range, start extra service threads
984 automatically when the request queue builds up.
987 Frequency : mixed-endian client/server environments
989 Description: mixed-endian crashes
990 Details : The new msg_v2 system had some failures in mixed-endian
993 Severity : enhancement
995 Description: Easy OST removal
996 Details : OSTs can be permanently deactivated with e.g. 'lctl
997 conf_param lustre-OST0001.osc.active=0'
999 Severity : enhancement
1001 Description: MGS proc entries
1002 Details : Added basic proc entries for the MGS showing what filesystems
1005 Severity : enhancement
1007 Description: provide MGS failover
1008 Details : Added config lock reacquisition after MGS server failover.
1010 Severity : enhancement
1012 Description: add Linux 2.4 support
1013 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1017 Description: a large application tries to do I/O to the same resource and dies
1018 in the middle of it.
1019 Details : Check the req->rq_arrival time after the call to
1020 ost_brw_lock_get(), but before we do anything about
1021 processing it & sending the BULK transfer request. This
1022 should help move old stale pending locks off the queue as
1023 quickly as obd_timeout.
1026 Frequency : when an incorrect nid is specified during startup
1028 Description: ptlrpc connect to non-existant node causes kernel crash
1029 Details : LNET can't be re-entered from an event callback, which
1030 happened when we expire a message after the export has been
1031 cleaned up. Instead, hand the zombie cleanup off to another
1034 Severity : enhancement
1036 Description: plain/inodebits lock performance improvement
1037 Details : Grouping plain/inodebits in granted list by their request modes
1038 and bits policy, thus improving the performance of search through
1042 Frequency : only if OST filesystem is corrupted
1044 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1045 Details : for a short time RPCs with bulk IO are in the replay list,
1046 but replay of bulk IOs is unimplemented. If the OST filesystem
1047 is corrupted due to disk cache incoherency and then replay is
1048 started it is possible to trip an assertion. Avoid putting
1049 committed RPCs into the replay list at all to avoid this issue.
1052 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1054 Description: System hang on startup
1055 Details : This bug allowed the liblustre (e.g. catamount) client to
1056 return to the app before handling all startup RPCs. This
1057 could leave the node unresponsive to lustre network traffic
1058 and manifested as a server ptllnd timeout.
1060 Severity : enhancement
1062 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1063 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1064 boolean to control whether to print peer debug info when a
1065 client's RPC times out.
1068 Frequency : only for kernels with patches from Lustre below 1.4.3
1070 Description: Remove old rdonly API
1071 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1074 Frequency : only for devices with external journals
1076 Description: Set external device read-only also
1077 Details : During a commanded failover stop, we set the disk device
1078 read-only while the server shuts down. We now also set any
1079 external journal device read-only at the same time.
1082 Frequency : when upgrading from 1.4 while trying to change parameters
1084 Description: The wrong (new) MDC name was used when setting parameters for
1085 upgraded MDT's. Also allows changing of OSC (and MDC)
1086 parameters if --writeconf is specified at tunefs upgrade time.
1089 Frequency : when setting specific ost indicies
1091 Description: QOS code breaks on skipped indicies
1092 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1093 created with --index need not be sequential.
1095 Severity : enhancement
1097 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1098 Details : The uninit_groups feature works in conjunction with the kernel
1099 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1100 up the pass1 processing of e2fsck. This is a read-only feature
1101 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1102 mount filesystems that have had this feature enabled.
1104 Severity : enhancement
1106 Description: Improve multi-block allocation algorithm to avoid fragmentation
1107 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1108 allocation locality and avoid filesystem fragmentation.
1110 ------------------------------------------------------------------------------
1112 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1114 * Support for kernels:
1115 2.4.21-47.0.1.EL (RHEL 3)
1116 2.6.5-7.283 (SLES 9)
1117 2.6.9-42.0.10.EL (RHEL 4)
1118 2.6.12.6 vanilla (kernel.org)
1119 2.6.16.27-0.9 (SLES 10)
1120 * Recommended e2fsprogs version: 1.39.cfs5
1122 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1126 Frequency : occasional, depends on client load and configuration
1127 Bugzilla : 12181, 12203
1128 Description: data loss for recently-modified files
1130 Details : In some cases it is possible that recently written or created
1131 files may not be written to disk in a timely manner (this should
1132 normally be within 30s unless client IO load is very high).
1133 The problem appears as zero-length files or files that are a
1134 multiple of 1MB in size after a client crash or client eviction
1135 that are missing data at the end of the file.
1137 This problem is more likely to be hit on clients where files are
1138 repeatedly created and unlinked in the same directory, clients
1139 have a large amount of RAM, have many CPUs, the filesystem has
1140 many OSTs, the clients are rebooted frequently, and/or the files
1141 are not accessed by other nodes after being written.
1143 The presence of the problem can be detected by looking at
1144 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1145 smaller than the second (nr_unused) then dirty files will not
1146 be flushed automatically to disk. "sync; sleep 10" should be
1147 run several times on the node before unmounting it to update
1148 Lustre (this is also safe to run on nodes without this problem).
1150 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1151 that can cause this same problem, so customers using that kernel
1152 also need to update the kernel in addition to Lustre. In order
1153 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1155 It is normal that files written just before a client crash (less
1156 than 30s) may not yet have been flushed to disk, even for local
1160 Frequency : frequent on thin XT3 nodes
1162 Description: UUID collision on thin XT3 Linux nodes
1163 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1164 randomly, since we relied on an insufficiently-seeded PRNG.
1169 Description: OSS hangs after "All ost request buffers busy"
1170 Details : A deadlock between quota and journal operations caused OSS
1171 hangs after printing "All ost request buffers busy."
1174 Frequency : always on liblustre builds
1176 Description: Cleanup compiler warnings on liblustre
1179 Frequency : always on liblustre builds on XT3
1181 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1185 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1186 > 32000 subdirectories
1187 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1188 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1189 subdirectory count crosses 32000. This will aid e2fsck to
1190 correctly handle more than 32000 subdirectories.
1193 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1195 Description: System hang on startup
1196 Details : This bug allowed the liblustre (e.g. catamount) client to
1197 return to the app before handling all startup RPCs. This
1198 could leave the node unresponsive to lustre network traffic
1199 and manifested as a server ptllnd timeout.
1201 Severity : enhancement
1203 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1204 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1205 boolean to control whether to print peer debug info when a
1206 client's RPC times out.
1211 Description: make O_SYNC working on 2.6 kernels
1212 Details : 2.6 kernels use different method for mark pages for write,
1213 so need add a code to lustre for O_SYNC work.
1218 Description: Failure to close file and release space on NFS
1219 Details : Put inode details into lock acquired in ll_intent_file_open.
1220 Use mdc_intent_lock in ll_intent_open to properly
1221 detect all kind of errors unhandled by mdc_enqueue.
1226 Description: proc file read during shutdown sometimes raced obd removal,
1228 Details : Add lock to prevent obd access after proc file removal.
1231 Frequency : Only for files larger than 4GB on 32-bit clients.
1233 Description: improperly doing page alignment of locks
1234 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1235 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1240 Description: RPCs being resent when they shouldn't be
1241 Details : Some RPCs that should not be resent are being resent. This
1242 can cause inconsistencies in the RPC state machine. Do not
1243 resend such requests.
1246 Frequency : rare, only with NFS export
1248 Description: Crash on NFS re-export node
1249 Details : under very unusual load conditions an assertion is hit in
1250 ll_intent_file_open()
1253 Frequency : only if OST filesystem is corrupted
1255 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1256 Details : for a short time RPCs with bulk IO are in the replay list,
1257 but replay of bulk IOs is unimplemented. If the OST filesystem
1258 is corrupted due to disk cache incoherency and then replay is
1259 started it is possible to trip an assertion. Avoid putting
1260 committed RPCs into the replay list at all to avoid this issue.
1265 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1266 Details : Large single O_DIRECT read and write calls can fail to allocate
1267 a sufficiently large buffer to process the request. In case of
1268 allocation failure the allocation is retried with a smaller
1269 buffer and broken into smaller requests.
1271 Severity : enhancement
1273 Description: Add -o localflock option to simulate old noflock behaviour.
1274 Details : This will achieve local-only flock/fcntl locks coherentness.
1279 Description: versioning check is incomplete
1280 Details : Checking the version difference of client vs. server, report
1281 error if the gap is too big.
1286 Description: add support PG_writeback bit
1287 Details : add support for PG_writeback bit for Lustre, for more carefull
1288 work with page cache in 2.6 kernel. This also fix some deadlocks
1289 and remove hack for work O_SYNC with 2.6 kernel.
1291 Severity : enhancement
1293 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1294 Details : The uninit_groups feature works in conjunction with the kernel
1295 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1296 up the pass1 processing of e2fsck. This is a read-only feature
1297 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1298 mount filesystems that have had this feature enabled.
1300 Severity : enhancement
1302 Description: Improve multi-block allocation algorithm to avoid fragmentation
1303 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1304 allocation locality and avoid filesystem fragmentation.
1306 ------------------------------------------------------------------------------
1308 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1310 * Support for kernels:
1311 2.6.9-42.0.3.EL (RHEL 4)
1312 2.6.5-7.276 (SLES 9)
1313 2.4.21-47.0.1.EL (RHEL 3)
1314 2.6.12.6 vanilla (kernel.org)
1315 2.6.16.21-0.8 (SLES10)
1316 * Recommended e2fsprogs version: 1.39.cfs2-0
1318 * The backwards-compatible /proc/sys/portals symlink has been removed
1319 in this release. Before upgrading, please ensure that you change
1320 any configuration scripts or /etc/sysctl.conf files that access
1321 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1322 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1323 in advance of the upgrade on any system running Lustre 1.4.6 or
1324 newer, since /proc/sys/lnet was added in that version.
1325 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1329 Frequency : only when quota is used
1331 Description: avoid scanning export list for quota master
1332 Details : Change the algorithms to avoid scanning export list in order
1333 to improve the efficiency.
1336 Frequency : MDS failover only, very rarely
1338 Description: "went back in time" messages on mds failover
1339 Details : The greatest transno may be lost when the current operation
1340 finishes with an error (transno==0) and the client's last_rcvd
1341 record is over-written. Save the greatest transno in the
1342 mds_last_transno for this case.
1345 Frequency : always for specific kernels and striping counts
1347 Description: client may get "Matching packet too big" without ACL support
1348 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1349 when trying to access files in certain configurations. The
1350 clients should in fact be denied when mounting because they do
1351 not understand ACLs.
1354 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1356 Description: many clients connecting with IO in progress causes connect timeouts
1357 Details : Avoid synchronous journal commits to avoid delays caused by many
1358 clients connecting/disconnecting when bulk IO is in progress.
1359 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1360 OST_IO_PORTAL to avoid delays behind potentially many pending
1364 Frequency : occasionally with multiple writers to a single file
1366 Description: shared writes to file may result in wrong size reported by stat()
1367 Details : Allow growing of kms when extent lock is cancelled
1370 Frequency : always with random mmap IO to multi-striped file
1372 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1373 Details : Only if the hole is at the end of OST object so that kms is too
1374 small. Fix is to increase kms accordingly in ll_nopage.
1377 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1379 Description: writes to a missing object would leak memory on the OST
1380 Details : If there is an inconsistency between the MDS and OST filesystems,
1381 such that the MDS references an object that doesn't exist, writes
1382 to that object will leak memory due to incorrect cleanup in the
1383 error handling path, eventually running out of memory on the OST.
1388 Description: Creating too long symlink causes lustre errors
1389 Details : Check symlink and name lengths before sending requests to MDS.
1392 Frequency : only if flock is enabled (not on by default)
1394 Description: posix locks not released on fd closure on 2.6.9+
1395 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1396 kernels, this caused such locks not to be released on fd close and
1397 then assertions on fs unmount about still used locks.
1400 Frequency : MDS failover only, very rarely
1402 Description: clients may get ASSERTION(granted_lock != NULL)
1403 Details : When request was taking a long time, and a client was resending
1404 a getattr by name lock request. The were multiple lock requests
1405 with the same client lock handle and
1406 mds_getattr_name->fixup_handle_for_resent_request found one of the
1407 lock handles but later failed with ASSERTION(granted_lock != NULL).
1412 Description: handle->h_buffer_credits > 0, assertion failure
1413 Details : h_buffer_credits is zero after truncate, causing assertion
1414 failure. This patch extends the transaction or creates a new
1418 Frequency : NFS re-export or patchless client
1419 Bugzilla : 11179, 10796
1420 Description: Crash on NFS re-export node (__d_move)
1421 Details : We do not want to hash the dentry if we don't have a lock.
1422 But if this dentry is later used in d_move, we'd hit uninitialised
1423 list head d_hash, so we just do this to init d_hash field but
1424 leave dentry unhashed.
1427 Frequency : NFS re-export or patchless client
1429 Description: NFS exports has problem with symbolic link
1430 Details : lustre client didn't properly install dentry when re-exported
1431 to NFS or running patchless client.
1434 Frequency : NFS re-export or patchless client
1436 Description: Various nfs/patchless fixes.
1437 Details : fixes reuse disconected alias for lookup process - this fixes
1438 warning "find_exported_dentry: npd != pd",
1439 fix permission error with open files at nfs.
1440 fix apply umask when do revalidate.
1443 Frequency : occasional
1445 Description: Crash on NFS re-export node
1446 Details : calling clear_page() on the wrong pointer triggered oops in
1447 generic_mapping_read().
1450 Frequency : rarely, using O_DIRECT IO
1452 Description: unaligned directio crashes client with LASSERT
1453 Details : check for unaligned buffers before trying any requests.
1456 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1458 Description: stale data returned from RAID cache
1459 Details : If only a small amount of IO is done to the RAID device before
1460 reading it again it is possible to get stale data from the RAID
1461 cache instead of reading it from disk.
1464 Frequency : always for sles10 kernel
1466 Description: sles10 support
1467 Details : ll_follow_link: compile fixes and using of nd_set_link
1468 under newer kernels.
1471 Frequency : depends on arch, kernel and compiler version, always on sles10
1474 Description: recursive or deep enough symlinks cause stack overflow
1475 Details : getting rid of large stack-allocated variable in
1479 Frequency : depends on hardware
1481 Description: lustre write performance loss in the SLES10 kernel
1482 Details : the performance loss is caused by using of write barriers in the
1483 ext3 code. The SLES10 kernel turns barrier support on by
1484 default. The fix is to undo that change for ldiskfs.
1486 ------------------------------------------------------------------------------
1488 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1490 * Support for kernels:
1491 2.6.9-42.0.3EL (RHEL 4)
1492 2.6.5-7.276 (SLES 9)
1493 2.4.21-47.0.1.EL (RHEL 3)
1494 2.6.12.6 vanilla (kernel.org)
1498 Frequency : quota enabled and large files being deleted
1500 Description: releasing more than 4GB of quota at once hangs OST
1501 Details : If a user deletes more than 4GB of files on a single OST it
1502 will cause the OST to spin in an infinite loop. Release
1503 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1508 Description: statfs data retrieved from /proc may be stale or zero
1509 Details : When reading per-device statfs data from /proc, in the
1510 {kbytes,files}_{total,free,avail} files, it may appear
1511 as zero or be out of date.
1514 Frequency : systems with MD RAID1 external journal devices
1516 Description: lconf's call to blkid is confused by RAID1 journal devices
1517 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1518 of returning all block devices that match the journal UUID.
1521 Frequency : always, for aggregate stripe size over 4GB
1523 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1524 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1525 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1528 Frequency : always if "lfs find" used on a local file/directory
1530 Description: "lfs find" segfaults if used on a local file/directory
1531 Details : The case where a directory component was not specified wasn't
1532 handled correctly. Handle this properly.
1535 Frequency : always on ppc64
1537 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1538 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1539 perform endianness conversion.
1542 Frequency : rarely (truncate to non-zero file size after write under load)
1543 Bugzilla : 10730, 10687
1544 Description: Files padded with zeros to next 4K multiple
1545 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1546 it is possible that files that are truncated to a non-zero size
1547 immediately after being written are filled with zero bytes beyond
1548 the truncated size. No file data is lost.
1550 Severity : enhancement
1552 Description: Allow recovery/failover for liblustre clients.
1553 Details : liblustre clients were unaware of failover configurations until
1556 Severity : enhancement
1558 Description: user file locks should fail when not mounting with flock option
1559 Details : Set up an error-returning stub in ll_file_operations.lock field
1560 to prevent incorrect behaviour when client is mounted without
1561 flock option. Also, set up properly f_op->flock field for
1565 Frequency : always on ia64
1567 Description: "lfs df" loops on printing out MDS statfs information
1568 Details : The obd_ioctl_data was not initialized and in some systems
1569 this caused a failure during the ioctl that did not return
1570 an error. Initialize the struct and return an error on failure.
1573 Frequency : SLES 9 only
1575 Description: Error of copying files with lustre special EAs as root
1576 Details : Client side always return success for setxattr call for lustre
1577 special xattr (currently only "trusted.lov").
1580 Frequency : rarely on clusters with both ia64+i386 clients
1582 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1583 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1584 clients are doing concurrent writes to the same file it is
1585 possible that the ia64 clients may LASSERT because the OST
1586 extent locks are not PAGE_SIZE aligned. Ensure that grown
1587 locks are always aligned on the request boundary.
1590 Frequency : specific use, occasional
1592 Description: Overwriting in use executable truncates on-disk binary image
1593 Details : If one node attempts to overwrite an executable in use by
1594 another node, we now correctly return ETXTBSY instead of
1595 truncating the file.
1597 Severity : enhancement
1599 Description: Async OSC create to avoid the blocking unnecessarily.
1600 Details : If a OST has no remain object, system will block on the creating
1601 when need to create a new object on this OST. Now, ways use
1602 pre-created objects when available, instead of blocking on an
1603 empty osc while others are not empty. If we must block, we block
1604 for the shortest possible period of time.
1609 Description: chmod on Lustre root is propagated to other clients
1610 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1611 invalid by the follow_mount time.
1614 Frequency : liblustre clients only
1616 Description: Race in 'instant cancel' lock handling could lead to such locks
1617 never to be granted in case of SMP MDS
1618 Details : Do not destroy not yet granted but cbpending locks in
1622 Frequency : replay/resend of open
1624 Description: non null lock assetion failure in mds_intent_policy
1625 Details : Trying to replay/resend lockless open requests resulted in
1626 mds_open() returning 0 with no lock. Now it sets a flag if
1627 a lock is going to be returned.
1629 Severity : enhancement
1631 Description: Checksum enhancements
1632 Details : New checksum enhancements allow for resending RPCs that failed
1635 Severity : enhancement
1637 Description: Tunables on number of dirty pages in cacche
1638 Details : Allow to set limit on number of dirty pages cached.
1643 Description: client crash on unmount - lock still has references
1644 Details : In some error handling cases it was possible to leak a lock
1645 reference on a client while accessing a file. This was not
1646 harmful to the client during operation, but would cause the
1647 client to crash when the filesystem is unmounted.
1650 Frequency : specific case, rare
1652 Description: ETXTBSY on mds though file not in use by client
1653 Details : ETXTBSY is no longer incorrectly returned when attempting to
1654 chmod or chown a directory that the user previously tried to
1655 execute or a currently-executing binary.
1658 Frequency : extremely rare except on liblustre-based clients
1660 Description: Lustre space not freed when files are deleted
1661 Details : Clean up open-unlinked files after client eviction. Previously
1662 the unlink was skipped and the files remained as orphans.
1667 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1668 Details : In some cases it was possible to send a blocking callback to
1669 a client doing a glimpse, even though that client didn't get
1670 a lock granted. When the glimpse lock is cancelled on the OST
1671 the freed lock is left on the waiting list and corrupted the list.
1674 Frequency : all core dumps
1676 Description: Broke core dumps to lustre
1677 Details : Negative dentry may be unhashed if parent does not have UPDATE
1678 lock, but some callers, e.g. do_coredump, expect dentry to be
1679 hashed after successful create, hash it in ll_create_it.
1681 ------------------------------------------------------------------------------
1683 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1685 * Support for kernels:
1686 2.6.9-42.0.2.EL (RHEL 4)
1687 2.6.5-7.276 (SLES 9)
1688 2.4.21-40.EL (RHEL 3)
1689 2.6.12.6 vanilla (kernel.org)
1693 Frequency : always on RHEL 3
1695 Description: Number of open files grows over time
1696 Details : The number of open files grows over time, whether or not
1697 Lustre is started. This was due to a filp leak introduced
1698 by one of our kernel patches.
1700 ------------------------------------------------------------------------------
1702 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1704 * Support for kernels:
1705 2.6.9-42.EL (RHEL 4)
1706 2.6.5-7.276 (SLES 9)
1707 2.4.21-40.EL (RHEL 3)
1708 2.6.12.6 vanilla (kernel.org)
1713 Bugzilla : 5719, 9635, 9792, 9684
1714 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1715 Details : If a server is under heavy load and cannot reply to new
1716 connection requests before the client resends the (re)connect,
1717 the connection handling code can behave badly if two service
1718 threads are concurrently handing separate (re)connections from
1719 the same client. Add better locking to the connection handling
1720 code, and ensure that only a single connection will be processed
1721 for a given client UUID, even if the lock is dropped.
1723 Severity : enhancement
1725 Description: add TCP zero-copy support to kernel
1726 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1727 sends if the hardware supports scatter-gather and checksumming.
1728 This allows socklnd to do client-write and server-read more
1729 efficiently and reduce CPU utilization from skbuf copying.
1732 Frequency : only if NFS exporting from client
1734 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1735 Details : If setting attributes on a file created under NFS that had
1736 never been opened it would be possible to oops the client
1737 if the file had no objects.
1740 Frequency : always for liblustre
1742 Description: liblustre client does MDS+OSTs setattr RPC for each write
1743 Details : When doing a write from a liblustre client, the client
1744 incorrectly issued an RPC to the MDS and each OST the file was
1745 striped over in order to update the timestamps. When writing
1746 with small chunks and many clients this could overwhelm the MDS
1747 with RPCs. In all cases it would slow down the write because
1748 these RPCs are unnecessary.
1750 Severity : enhancement
1752 Description: allow number of MDS service threads to be changed at module load
1753 Details : It is now possible to change the number of MDS service threads
1754 running. Adding "options mds mds_num_threads={N}" to the MDS's
1755 /etc/modprobe.conf will set the number of threads for the next
1756 time Lustre is restarted (assuming the "mds" module is also
1757 reloaded at that time). The default number of threads will
1758 stay the same, 32 for most systems.
1763 Description: OST crash if filesystem is unformatted or corrupt
1764 Details : If an OST is started on a device that has never been formatted
1765 or if the filesystem is corrupt and cannot even mount then the
1766 error handling cleanup routines would dereference a NULL pointer.
1771 Description: NULL pointer deref in llap_from_page.
1772 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1773 incorrect) mapping if the page was truncated/reclaimed while it was
1774 searched for. Check for this condition and skip such pages when
1775 doing readahead. Introduce extra check to llap_from_page() to
1776 verify page->mapping->host is non-NULL (so page is not anonymous).
1779 Frequency : Sometimes when using sys_sendfile
1781 Description: "page not covered by a lock" warnings from ll_readpage
1782 Details : sendfile called ll_readpage without right page locks present.
1783 Now we introduced ll_file_sendfile that does necessary locking
1784 around call to generic_file_sendfile() much like we do in
1788 Frequency : with certain MDS communication failures at client mount time
1790 Description: NULL pointer deref after failed client mount
1791 Details : a client connection request may delayed by the network layer
1792 and not be sent until after the PTLRPC layer has timed out the
1793 request. If the client fails the mount immediately it will try
1794 to clean up before the network times out the request. Add a
1795 reference from the request import to the obd device and delay
1796 the cleanup until the network drops the request.
1799 Frequency : occasionally during client (re)connect
1801 Description: assertion failure during client (re)connect
1802 Details : processing a client connection request may be delayed by the
1803 client or server longer than the client connect timeout. This
1804 causes the client to resend the connection request. If the
1805 original connection request is replied in this interval, the
1806 client may trip an assertion failure in ptlrpc_connect_interpret()
1807 which thought it would be the only running connect process.
1810 Frequency : only with obd_echo servers and clients that are rebooted
1812 Description: kernel BUG accessing uninitialized data structure
1813 Details : When running an obd_echo server it did not start the ping_evictor
1814 thread, and when a client was evicted an uninitialized data
1815 structure was accessed. Start the ping_evictor in the RPC
1816 service startup instead of the OBD startup.
1818 Severity : enhancement
1819 Bugzilla : 10193 (patchless)
1820 Description: Remove dependency on various unexported kernel interfaces.
1821 Details : No longer need reparent_to_init, exit_mm, exit_files,
1822 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1825 Frequency : rare (only users of deprecated and unsupported LDAP config)
1827 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1828 Details : LDAP apparently contains 'lustreName' attributes instead of
1829 'name'. A simple remapping of the name is sufficient.
1832 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1834 Description: waiting_locks_callback trips kernel BUG if client is evicted
1835 Details : Running with the dump_on_timeout debug flag turned on makes
1836 it possible that the waiting_locks_callback() can try to dump
1837 the Lustre kernel debug logs from an interrupt handler. Defer
1838 this log dumping to the expired_lock_main() thread.
1840 Severity : enhancement
1842 Description: Support NFS exporting on 2.6 kernels.
1843 Details : Implement non-rawops metadata methods for NFS server to use without
1844 changing NFS server code.
1847 Frequency : very rare (synthetic metadata workload only)
1849 Description: two racing renames might cause an MDS thread to deadlock
1850 Details : Running the "racer" program may cause one MDS thread to rename
1851 a file from being the source of a rename to being the target of
1852 a rename at exactly the same time that another thread is doing
1853 so, and the second thread has already enqueued these locks after
1854 doing a lookup of the target and is trying to relock them in
1855 order. Ensure that we don't try to re-lock the same resource.
1858 Frequency : only very large systems with liblustre clients
1860 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1861 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1862 all OSTs in parallel. This allows the eviction of stale liblustre
1863 clients to proceed much faster than if they were done in series,
1864 and also offers similar improvements for other set_info RPCs.
1869 Description: excessive CPU usage during initial read phase on client
1870 Details : During the initial read phase on a client, it would agressively
1871 retry readahead on the file, consuming too much CPU and impacting
1872 performance (since 1.4.5.8). Improve the readahead algorithm
1873 to avoid this, and also improve some other common cases (read
1874 of small files in particular, where "small" is files smaller than
1875 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1880 Description: MDS crash when receiving packet with unknown intent.
1881 Details : Do not LBUG in unknown intent case, just return -EFAULT
1883 Severity : enhancement
1884 Bugzilla : 9293, 9385
1885 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1886 Details : Do not serialize getattr (non-intent version) and statfs.
1889 Frequency : occasional, when OST network is overloaded/intermittent
1891 Description: client evicted by OST after bulk IO timeout
1892 Details : If a client sends a bulk IO request (read or write) the OST
1893 may evict the client if it is unresposive to its data GET/PUT
1894 request. This is incorrect if the network is overloaded (takes
1895 too long to transfer the RPC data) or dropped the OST GET/PUT
1896 request. There is no need to evict the client at all, since
1897 the pinger and/or lock callbacks will handle this, and the
1898 client can restart the bulk request.
1901 Frequency : Always when mmapping file with no objects
1903 Description: client crashes when mmapping file with no objects
1904 Details : Check that we actually have objects in a file before doing any
1905 operations on objects in ll_vm_open, ll_vm_close and
1911 Description: Request leak when working with deleted CWD
1912 Details : Introduce advanced request refcount tracking for requests
1913 referenced from lustre intent.
1915 Severity : Enhancement
1917 Description: Cache open file handles on client.
1918 Details : MDS now will return special lock along with openhandle, if
1919 requested and client is allowed to hold openhandle, even if unused,
1920 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1921 closing files for every read/write openration.
1923 Severity : Enhancement
1925 Description: Cache open negative dentries on client when possible.
1926 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1927 negative dentries on lock revocation.
1932 Description: Remounting a client read-only wasn't possible with a zconf mount
1933 Details : It wasn't possible to remount a client read-only with llmount.
1935 Severity : enhancement
1936 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1937 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1938 MPICH is included to add Lustre-specific ADIO interfaces.
1939 This is based closely on the UFS ADIO layer and only differs
1940 in file creation, in order to allow the OST striping to be set.
1941 This is user-contributed code and not supported by CFS.
1946 Description: extended inode attributes (immutable, append-only) work improperly
1947 when 2.4 and 2.6 kernels are used on client/server or vice versa
1948 Details : Introduce kernel-independent values for these flags.
1950 Severity : enhancement
1953 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1954 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1955 granularity. Now, Fractional MB granularity is be supported,
1956 this is very useful for low memory system.
1958 Severity : enhancement
1960 Description: Getattr by fid
1961 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1962 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1963 have locks and avoids getattr rpc requests for it on every stat.
1966 Frequency : Always, for filesystems larger than 2TB
1968 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1969 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1970 filesystems larger than 2TB to be mounted with mballoc enabled.
1973 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1975 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1976 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1977 filesystems larger than 2TB is mounted then the kernel will
1978 silently corrupt the start of the filesystem. CONFIG_LBD is
1979 enabled for all CFS-supported kernels, but the possibility of
1980 this happening with a modified kernel config exists.
1982 Severity : enhancement
1984 Description: add client O_DIRECT support for 2.6 kernels
1985 Details : It is now possible to do O_DIRECT reads and writes to files
1986 in the Lustre client mountpoint on 2.6 kernel clients.
1988 Severity : enhancement
1990 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1991 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1992 parallel, not waiting for response from every OST before sending
1993 a rpc to the next OST.
1998 Description: setattr vs write race when updating file timestamps
1999 Details : Client processes that update a file timestamp into the past
2000 right after writing to the file (e.g. tar) it is possible that
2001 the updated file modification time can be reset to the current
2002 time due to a race between processing the setattr and write RPC.
2004 Severity : enhancement
2006 Description: Bring 'lfs find' closer in line with regular Linux find.
2007 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2008 -print0 options and obtains all the needed info through the lustre
2011 Severity : enhancement
2013 Description: support up to 1024 configured devices on one node
2014 Details : change obd_dev array from statically allocated to dynamically
2015 allocated structs as they are first used to reduce memory usage
2020 Description: Flush dirty partially truncated pages during truncate
2021 Details : Immediatelly flush partially truncated pages in filter_setattr,
2022 this way we completely avoid having any pages in page cache on OST
2023 and can retire ugly workarounds during writes to flush such pages.
2028 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2029 Details : protect inode from truncation within vfs_unlink() context
2030 just take a reference before calling vfs_unlink() and release it
2031 when parent's i_sem is free.
2036 Description: last_id value checked outside lock on OST caused LASSERT failure
2037 Details : If there were multiple MDS->OST object precreate requests in
2038 flight, it was possible that the OST's last object id was checked
2039 outside a lock and incorrectly tripped an assertion. Move checks
2040 inside locks, and discard old precreate requests.
2043 Frequency : always, if extents are used on OSTs
2045 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2046 Details : OSTs using the extents format would not zero the high 16 bits of
2047 the index physical block number. This is not a problem for any
2048 OST filesystems smaller than 16TB, and no kernels support ext3
2049 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2050 new/modified files) and can be fixed for existing filesystems
2051 with e2fsprogs-1.39-cfs1.
2056 Description: import connection selection may be incorrect if timer wraps
2057 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2058 import connections to be ignored if the 32-bit jiffies counter
2059 wraps. Use a 64-bit jiffies counter.
2062 Frequency : very large clusters immediately after boot
2064 Description: LNET request buffers exhausted under heavy short-term load
2065 Details : If a large number of client requests are generated on a service
2066 that has previously never seen so many requests it is possible
2067 that the request buffer growth cannot keep up with the spike in
2068 demand. Instead of dropping incoming requests, they are held in
2069 the LND until the RPC service can accept more requests.
2072 Frequency : Sometimes during replay
2074 Description: Assertion failure in ll_local_open after replay.
2075 Details : If replay happened on an open request reply before we were able
2076 to set replay handler, reply will become not swabbed tripping the
2077 assertion in ll_local_open. Now we set the handler right after
2078 recognising of open request
2081 Frequency : very rare
2083 Description: kernel reports "badness in vsnprintf"
2084 Details : Reading from the "recovery_status" /proc file in small chunks
2085 may cause a negative length in lprocfs_obd_rd_recovery_status()
2086 call to vsnprintf() (which is otherwise harmless). Exit early
2087 if there is no more space in the output buffer.
2089 Severity : enhancement
2091 Description: clear OBD RPC statistics by writing to them
2092 Details : It is now possible to clear the OBD RPC statistics by writing
2093 to the "stats" file.
2098 Description: Client mtime is not the same on different clients after utimes
2099 Details : In some cases, the client was using the utimes() syscall on
2100 a file cached on another node. The clients now validate the
2101 ctime from the MDS + OSTs to determine which one is right.
2106 Description: Inability to activate failout mode
2107 Details : lconf script incorrectly assumed that in pythong string's numeric
2108 value is used in comparisons.
2111 Frequency : always with multiple stripes per file
2113 Description: Inefficient object allocation for mutli-stripe files
2114 Details : When selecting which OSTs to stripe files over, for files with
2115 a stripe count that divides evenly into the number of OSTs,
2116 the MDS is always picking the same starting OST for each file.
2117 Return the OST selection heuristic to the original design.
2122 Description: mount failures may take full timeout to return an error
2123 Details : Under some heavy load conditions it is possible that a
2124 failed mount can wait for the full obd_timeout interval,
2125 possibly several minutes, before reporting an error.
2126 Instead return an error as soon as the status is known.
2128 Frequency : quota enabled and large files being deleted
2130 Description: releasing more than 4GB of quota at once hangs OST
2131 Details : If a user deletes more than 4GB of files on a single OST it
2132 will cause the OST to spin in an infinite loop. Release
2133 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2138 Description: statfs data retrieved from /proc may be stale or zero
2139 Details : When reading per-device statfs data from /proc, in the
2140 {kbytes,files}_{total,free,avail} files, it may appear
2141 as zero or be out of date.
2144 Frequency : systems with MD RAID1 external journal devices
2146 Description: lconf's call to blkid is confused by RAID1 journal devices
2147 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2148 of returning all block devices that match the journal UUID.
2151 Frequency : always, for aggregate stripe size over 4GB
2153 Description: assertion fails when trying to use 4GB stripe size
2154 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2155 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2158 Frequency : always on ppc64
2160 Description: the first write on an ext3 filesystem with mballoc got stuck
2161 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2162 perform endianness conversion.
2164 ------------------------------------------------------------------------------
2166 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2168 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2169 INTEROPERATE with older versions automatically. Please read the
2170 user documentation before upgrading any part of a live system.
2171 * WARNING: Lustre networking configuration changes are required with
2172 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2175 * Support for kernels:
2176 2.6.9-22.0.2.EL (RHEL 4)
2177 2.6.5-7.244 (SLES 9)
2178 2.6.12.6 vanilla (kernel.org)
2181 Severity : enhancement
2182 Bugzilla : 7981/8208
2183 Description: Introduced Lustre Networking (LNET)
2184 Details : LNET is new networking infrastructure for Lustre, it includes
2185 a reorganized network configuration mode (see the user
2186 documentation for full details) as well as support for routing
2187 between different network fabrics. Lustre Networking Devices
2188 (LNDs) for the supported network fabrics have also been
2189 created for this new infrastructure.
2191 Severity : enhancement
2192 Description: Introduced Access control lists
2193 Details : clients can set ACLs on files and directories in order to have
2194 more fine-grained permissions than the standard Unix UGO+RWX.
2195 The MDS must be started with the "-o acl" mount option.
2197 Severity : enhancement
2198 Description: Introduced filesystem quotas
2199 Details : Administrators may now establish per-user quotas on the
2202 Severity : enhancement
2204 Description: Configuration change for the XT3
2205 The PTLLND is now used to run Lustre over Portals on the XT3
2206 The configure option(s) --with-cray-portals are no longer used.
2207 Rather --with-portals=<path-to-portals-includes> is used to
2208 enable building on the XT3. In addition to enable XT3 specific
2209 features the option --enable-cray-xt3 must be used.
2214 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2215 Details : A race between allocating a new llcd and re-getting the llcd_lock
2216 allowed another thread to grab newly-allocated llcd.
2218 Severity : enhancement
2220 Description: 2.6 OST async journal commit and locking fix to improve performance
2221 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2222 2.6 kernels are now async as they already were in 2.4 kernels so
2223 that they can commit concurrently with the network bulk transfer.
2224 For block-allocated files the filter allocation semaphore is held
2225 to avoid filesystem fragmentation during allocation. BKL lock
2226 removed for 2.6 xattr operations where it is no longer needed.
2231 Description: lconf incorrectly determined whether two IP networks could talk
2232 Details : In some more complicated routing and multiple-network
2233 configurations, lconf will avoid trying to make a network
2234 connection to a disjoint part of the IP space. It was doing the
2235 math incorrectly for one set of cases.
2240 Description: Fix for potential infinite loop processing records in an llog.
2241 Details : If an llog record is corrupted/zeroed, it is possible to loop
2242 forever in llog_process(). Validate the llog record length
2243 and skip the remainder of the block on error.
2246 Frequency : occasional (liblustre only)
2248 Description: liblustre could not open files whose last component is a symlink
2249 Details : sysio_path_walk() would incorrectly pass the open intent to
2250 intermediate path components.
2253 Frequency : rare (liblustre only with non-standard tuning)
2254 Bugzilla : 7201 (7350)
2255 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2256 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2257 after it has been released, only do this for referenced locks
2259 Severity : enhancement
2261 Description: specifying an (invalid) directory default stripe_size of -1
2262 would reset the directory default striping
2263 Details : stripe_size -1 was used internally to signal directory stripe
2264 removal, now use "all default" to signal dir stripe removal
2265 as a directory striping of "all default" is not useful
2268 Frequency : common for large clusters running liblustre clients
2270 Description: doing an ls when liblustre clients are running is slow
2271 Details : sending a glimpse AST to a liblustre client waits for every AST
2272 to time out, as liblustre clients will not respond. Since they
2273 cannot cache data we refresh the OST lock LVB from disk instead.
2275 Severity : enhancement
2277 Description: doing an ls at the same time as file IO can be slow
2278 Details : enqueue and other "small" requests can be blocked behind many
2279 large IO requests. Create a new OST IO portal for non-IO
2280 requests so they can be processed faster.
2283 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2285 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2286 created file with mode 000
2287 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2288 allow an ftruncate() as a non-root user to a file with mode 000.
2289 Lustre now respects this flag to disable mode checks when
2290 truncating a file owned by the user
2293 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2296 Description: Revoking locks from clients that went dead or catatonic might take
2298 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2299 cancellation of such locks instant on servers without waiting for
2300 any reply from clients. Clients drops these locks when cancel
2301 notification from server is received without replying.
2304 Frequency : liblustre-only, when liblustre client dies or becomes busy
2306 Description: Doing ls on Linux clients can take a long time with active
2308 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2309 granting such locks to it in the first place if possible. Locks
2310 are taken by proxy on the OST during the read or write and
2311 dropped immediately afterward. Add connect flags handling, do
2312 not grant locks to liblustre clients for glimpse ASTs.
2314 Severity : enhancement
2316 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2317 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2318 sequential reads and behaves badly with applications doing random
2319 reads. Improve it by reading ahead at least read region, and
2320 avoiding excessive large RPC for small reads.
2322 Severity : enhancement
2324 Description: Creating more than 1000 files for a single job may cause a load
2325 imbalance on the OSTs if there are also a large number of OSTs.
2326 Details : qos_prep_create() uses an OST index reseed value that is an
2327 even multiple of the number of available OSTs so that if the
2328 reseed happens in the middle of the object allocation it will
2329 still utilize the OSTs as uniformly as possible.
2334 Description: OST or MDS may oops in ping_evictor_main()
2335 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2336 but doesn't restart at beginning of obd_exports_timed list
2339 Severity : enhancement
2341 Description: improve by-nid export eviction on the MDS and OST
2342 Details : allow multiple exports with the same NID to be evicted at one
2343 time without re-searching the exports list.
2346 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2348 Description: MDS may oops in groups_free()
2349 Details : in rare race conditions a newly allocated group_info struct is
2350 freed again, and this can be NULL. The 2.4 compatibility code
2351 for groups_free() checked for a NULL pointer, but 2.6 did not.
2354 Frequency : common for liblustre clients doing little filesystem IO
2355 Bugzilla : 9352, 7313
2356 Description: server may evict liblustre clients accessing contended locks
2357 Details : if a client is granted a lock or receives a completion AST
2358 with a blocking AST already set it would not reply to the AST
2359 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2362 Frequency : lfs setstripe, only systems with more than 160 OSTs
2364 Description: unable to set striping with a starting offset beyond OST 160
2365 Details : llapi_create_file() incorrectly limited the starting stripe
2366 index to the maximum single-file stripe count.
2369 Frequency : LDAP users only
2371 Description: lconf did not handle in-kernel recovery with LDAP properly
2372 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2374 Severity : enhancement
2376 Description: bind OST threads to NUMA nodes to improve performance
2377 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2378 node and do their allocations there to localize memory access
2380 Severity : enhancement
2382 Description: llmount can determine client NID directly from Myrinet (GM)
2383 Details : the client NID code from gmnalnid was moved directly into
2384 llmount, removing the need to use this or specifying the
2385 client NID explicitly when mounting GM clients with zeroconf
2388 Frequency : if client is started with down MDS
2390 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2391 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2392 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2393 ptlrpc_import_delayed_req() because we didn't check if the
2394 request was interrupted, and we also didn't break out of the
2395 event loop if there was no timeout
2400 Description: data loss during non-page-aligned writes to a single file from
2401 both multiple nodes and multiple threads on one node at same time
2402 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2403 inconsistency led to false short-reads, that were cached and later
2404 used by ->prepare_write() to fill in partially written page,
2405 leading to data loss.
2408 Frequency : always, if lconf --abort_recovery used
2410 Description: lconf --abort_recovery fails with 'Operation not supported'
2411 Details : lconf was attempting to abort recovery on the MDT device and not
2414 Severity : enhancement
2416 Description: remove cleanup logs
2417 Details : replace lconf-generated cleanup logs with lustre internal
2418 cleanup routines. Eliminates the need for client-cleanup and
2421 Severity : enhancement
2423 Description: add support for EAs (user and system) on lustre filesystems
2424 Details : it is now possible to store extended attributes in the Lustre
2425 client filesystem, and with the user_xattr mount option it
2426 is possible to allow users to store EAs on their files also
2428 Severity : enhancement
2430 Description: Add possibility (config option) to show minimal available OST free
2432 Details : When compiled with --enable-mindf configure option, statfs(2)
2433 (and so, df) will return least minimal free space available from
2434 all OSTs as amount of free space on FS, instead of summary of
2435 free spaces of all OSTs.
2437 Severity : enhancement
2439 Description: do not expand extent locks acquired on OST-side
2440 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2441 by server: they are not cached anyway.
2444 Frequency : when mmap is used/binaries executed from Lustre
2446 Description: Unmmap pages before throwing them away from read cache.
2447 Details : llap_shrink cache now attempts to unmap pages before discarding
2448 them (if unmapping failed - do not discard). SLES9 kernel has
2449 extra checks that trigger if this unmapping is not done first.
2454 Description: lconf didn't resolve symlinks before checking to see whether a
2455 given mountpoint was already in use
2458 Frequency : when migrating failover services
2459 Bugzilla : 6395, 9514
2460 Description: When migrating a subset of services from a node (e.g. failback
2461 from a failover service node) the remaining services would
2462 time out and evict clients.
2463 Details : lconf --force (implied by --failover) sets the global obd_timeout
2464 to 5 seconds in order to quickly disconnect, but this caused
2465 other RPCs to time out too quickly. Do not change the global
2466 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2468 Severity : enhancement
2469 Frequency : if MDS is started with down OST
2470 Bugzilla : 9439,5706
2471 Description: Allow startup/shutdown of an MDS without depending on the
2472 availability of the OSTs.
2473 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2474 Add appropriate locking and lov-osc refcounts for safe
2475 cleaning. Add osc abort_inflight calls in case the
2476 synchronize never started.
2479 Frequency : occasional (Cray XT3 only)
2481 Description: root not authorized to access files in CRAY_PORTALS environment
2482 Details : The client process capabilities were not honoured on the MDS in
2483 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2484 been accessed by an authorized user then root was able to access
2485 the file on the local client also. The root user capabilities
2486 are now allowed on the MDS, as this environment has secure UID.
2489 Frequency : occasional
2491 Description: ldiskfs "too long searching" message happens too often
2492 Details : A debugging message (otherwise harmless) prints too often on
2493 the OST console. This has been reduced to only happen when
2494 there are fragmentation problems on the filesystem.
2499 Description: Division by zero in statfs when all OSCs are inactive
2500 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2501 lov_statfs divides by value returned by lov_get_stripecnt().
2505 Bugzilla : 9489, 3273
2506 Description: First write from each client to each OST was only 4kB in size,
2507 to initialize client writeback cache, which caused sub-optimal
2508 RPCs and poor layout on disk for the first writen file.
2509 Details : Clients now request an initial cache grant at (re)connect time
2510 and so that they can start streaming writes to the cache right
2511 away and always do full-sized RPCs if there is enough data.
2512 If the OST is rebooted the client also re-establishes its grant
2513 so that client cached writes will be honoured under the grant.
2518 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2519 Details : Now I/O RPCs go to different portal number and (presumably) fast
2520 lock requests (and glimses) and other RPCs get their own service
2521 threads pool that should be able to service those RPCs
2524 Severity : enhancement
2526 Description: Ability to exchange lustre version between client and servers and
2527 issue warnings at client side if client is too old. Also for
2528 liblustre clients there is ability to refuse connection of too old
2530 Details : New 'version' field is added to connect data structure that is
2531 filled with version info. That info is later checked by server and
2535 Frequency : rare, liblustre only.
2536 Bugzilla : 9296, 9581
2537 Description: Two simultaneous writes from liblustre at offset within same page
2538 might proceed at the same time overwriting eachother with stale
2540 Details : I/O lock withing llu_file_prwv was released too early, before data
2541 actually was hitting the wire. Extended lock-holding time until
2542 server acknowledges receiving data.
2545 Frequency : extremely rare. Never observed in practice.
2547 Description: avoid generating lustre_handle cookie of 0.
2548 Details : class_handle_hash() generates handle cookies by incrementing
2549 global counter, and can hit 0 occasionaly (this is unlikely, but
2550 not impossible, because initial value of cookie counter is
2551 selected randonly). Value of 0 is used as a sentinel meaning
2552 "unassigned handle" --- avoid it. Also coalesce two critical
2553 sections in this function into one.
2555 Severity : enhancement
2557 Description: allow liblustre clients to delegate truncate locking to OST
2558 Details : To avoid overhead of locking, liblustre client instructs OST to
2559 take extent lock in ost_punch() on client's behalf. New connection
2560 flag is added to handle backward compatibility.
2562 Severity : enhancement
2563 Bugzilla : 4928, 7341, 9758
2564 Description: allow number of OST service threads to be specified
2565 Details : a module parameter allows the number of OST service threads
2566 to be specified via "options ost ost_num_threads={N}" in the
2567 OSS's /etc/modules.conf or /etc/modprobe.conf.
2571 Bugzilla : 6146, 9635, 9895
2572 Description: servers crash with bad pointer in target_handle_connect()
2573 Details : In rare cases when a client is reconnecting it was possible that
2574 the connection request was the last reference for that export.
2575 We would temporarily drop the export reference and get a new
2576 one, but this may have been the last reference and the export
2577 was just destroyed. Get new reference before dropping old one.
2579 Severity : enhancement
2580 Frequency : if client is started with failover MDS
2582 Description: Allow multiple MDS hostnames in the mount command
2583 Details : Try to read the configuration from all specified MDS
2584 hostnames during a client mount in case the "primary"
2587 Severity : enhancement
2589 Description: Stop sending data to evicted clients as soon as possible.
2590 Details : Check if the client we are about to send or are sending data to
2591 was evicted already. (Check is done every second of waiting,
2592 for which l_wait_event interface was extended to allow checking
2593 of exit condition at specified intervals).
2596 Frequency : rare, normally only when NFS exporting is done from client
2598 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2600 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2603 Frequency : always, if srand() is called before liblustre initialization
2605 Description: Liblustre uses system PRNG disturbing its usage by user application
2606 Details : Introduce internal to lustre fast and high-quality PRNG for
2607 lustre usage and make liblustre and some other places in generic
2608 lustre code to use it.
2610 Severity : enhancement
2611 Bugzilla : 9477, 9557, 9870
2612 Description: Verify that the MDS configuration logs are updated when xml is
2613 Details : Check if the .xml configuration logs are newer than the config
2614 logs stored on the MDS and report an error if this is the case.
2615 Request --write-conf, or allow starting with --old_conf.
2617 Severity : enhancement
2619 Description: Handle symlinks in the path when checking if Lustre is mounted.
2620 Details : Resolve intermediate symlinks when checking if a client has
2621 mounted a filesystem to avoid duplicate client mounts.
2626 Description: lconf can hit an error exception but still return success.
2627 Details : The lconf command catches the Command error exception at the top
2628 level script context and will exit with the associated exit
2629 status, but doesn't ensure that this exit status is non-zero.
2634 Description: failure of ptlrpc thread startup can cause oops
2635 Details : Starting a ptlrpc service thread can fail if there are a large
2636 number of threads or the server memory is very fragmented.
2637 Handle this without oopsing.
2640 Frequency : always, only if liblustre and non-default acceptor port was used
2642 Description: liblustre cannot connect to servers with non-default acceptor port
2643 Details : tcpnal_set_default_params() was not called and was therefore
2644 ignoring the environment varaible TCPNAL_PORT, as well as other
2645 TCPNAL_ environment variables
2650 Description: two objects could be created on the same OST for a single file
2651 Details : If an OST is down, in some cases it was possible to create two
2652 objects on a single OST for a single file. No problems other
2653 than potential performance impact and spurious error messages.
2657 Bugzilla : 5681, 9562
2658 Description: Client may oops in ll_unhash_aliases
2659 Details : Client dcache may become inconsistent in race condition.
2660 In some cases "getcwd" can fail if the current directory is
2666 Description: Inode refcounting problems in NFS export code
2667 Details : link_raw functions used to call d_instantiate without obtaining
2668 extra inode reference first.
2672 Bugzilla : 9942, 9903
2673 Description: Referencing freed requests leading to crash, memleaks with NFS.
2674 Details : We used to require that call to ll_revalidate_it was always
2675 followed by ll_lookup_it. Also with revalidate_special() it is
2676 possible to call ll_revalidate_it() twice for the same dentry
2677 even if first occurence returned success. This fix changes semantic
2678 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2679 reference on a request referred from the intent.
2680 ll_intent_release() then releases such a request.
2683 Frequency : rare, normally benchmark loads only
2685 Description: unlinked inodes were kept in memory on the client
2686 Details : If a client is repeatedly creating and unlinking files it
2687 can accumulate a lot of stale inodes in the inode slab cache.
2688 If there is no other client load running this can cause the
2689 client node to run out of memory. Instead flush old inodes
2690 from client cache that have the same inode number as a new inode.
2693 Frequency : SLES9 2.6.5 kernel and long filenames only
2694 Bugzilla : 9969, 10379
2695 Description: utime reports stale NFS file handle
2696 Details : SLES9 uses out-of-dentry names in some cases, which confused
2697 the lustre dentry revalidation. Change it to always use the
2701 Frequency : rare, unless heavy write-truncate concurrency is continuous
2702 Bugzilla : 4180, 6984, 7171, 9963, 9331
2703 Description: OST becomes very slow and/or deadlocked during object unlink
2704 Details : filter_destroy() was holding onto the parent directory lock
2705 while truncating+unlinking objects. For very large objects this
2706 may block other threads for a long time and slow overall OST
2707 responsiveness. It may also be possible to get a lock ordering
2708 deadlock in this case, or run out of journal credits because of
2709 the combined truncate+unlink. Solution is to do object truncate
2710 first in one transaction without parent lock, and then do the
2711 final unlink in a new transaction with the parent lock. This
2712 reduces the lock hold time dramatically.
2715 Frequency : rare, 2.4 kernels only
2717 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2718 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2719 before unmounting their filesystem in order to avoid deadlock.
2720 The kernel_locked() function in 2.4 kernels only checks whether
2721 the kernel lock is held, not whether it is this process that is
2722 holding it as 2.6 kernels do.
2727 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2728 Details : The client ptlrpc code may be trying to reconnect to a down
2729 server before a previous connection attempt has timed out.
2730 Increase the reconnect interval to be longer than the connection
2731 timeout interval to avoid sending duplicate connections to
2735 Frequency : echo_client brw_test command
2737 Description: fix echo_client to work with OST preallocated code
2738 Details : OST preallocation code (5137) didn't take echo_client IO path
2739 into account: echo_client calls filter methods outside of any
2740 OST thread and, hence, there is no per-thread preallocated
2741 pages and buffers to use. Solution: hijack pga pages for IO. As
2742 a byproduct, this avoids unnecessary data copying.
2746 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2747 Description: Client can oops in mdc_commit_close() after open replay
2748 Details : It was possible for the MDS to return an open request with no
2749 transaction number in mds_finish_transno() if the client was
2750 evicted, but without actually returning an error. Clients
2751 would later try to replay that open and may trip an assertion
2752 Simplify the client close codepath, and always return an error
2753 from the MDS in case the open is not successful.
2756 Frequency : rare, 2.6 OSTs only
2758 Description: OST may deadlock under high load on fragmented files
2759 Details : If there was a heavy load and highly-fragmented OST filesystems
2760 it was possible to have all the OST threads deadlock waiting on
2761 allocation of biovecs, because the biovecs were not released
2762 until the entire RPC IO was completed. Instead, release biovecs
2763 as soon as they are complete to ensure forward IO progress.
2765 Severity : enhancement
2767 Description: Support for specifying external journal device at mount
2768 Details : If an OST or MDS device is formatted with an external journal
2769 device, this device major/minor is stored in the ext3 superblock
2770 and may not be valid for failover. Allow detecting and
2771 specifying the external journal at mount time.
2776 Description: Mounting an MDS with pending unlinked files may cause oops
2777 Details : target_finish_recovery() calls mds_postrecov() which returned
2778 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2779 considers this an error and immediately begins cleaning up the
2780 lov, just after starting the mds_lov process
2782 Severity : enhancement
2784 Description: Implement 'lfs df' to report actual free space on per-OST basis
2785 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2786 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2787 usage of objects; '-h' to report in human readable format.
2789 ------------------------------------------------------------------------------
2791 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2798 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2799 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2800 at startup during orphan unlinking. Instead, skip buddy update
2801 before setup, as it will be regenerated after recovery is complete.
2806 Description: If an OST is inactive, its locks might reference stale inodes.
2807 Details : lov_change_cbdata() must iterate over all namespaces, even if
2808 they are inactive to clear inode references from the lock.
2810 Severity : enhancement
2811 Frequency : occasional, if non-standard max_dirty_mb used
2813 Description: Client will block write RPCs if not enough grant
2814 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2815 then the client will block writes while waiting for another RPC
2816 to complete instead of consuming its dirty limit. With change
2817 we get improved performance when max_dirty_mb is small.
2819 Severity : enhancement
2820 Bugzilla : 3389, 6253
2821 Description: Add support for supplementary groups on the MDS.
2822 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2823 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2824 lookups for user supplementary groups into a cache.
2828 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2829 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2830 for liblustre O_EXCL is enforced by MDS.
2834 Description: readdir never returns NULL in liblustre.
2835 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2836 dentry in struct dirent.
2840 Frequency : liblustre only, depends on application IO pattern
2841 Description: liblustre clients evicted if not contacting servers
2842 Details : Don't put liblustre clients into the ping_evictor list, so
2843 they will not be evicted by the pinger ever.
2845 Severity : enhancement
2847 Description: Add ability to evict clients by NID from MDS.
2848 Details : By echoing "nid:$NID" string into
2849 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2850 $NID would be instantly evicted from this MDS and from all active
2851 OSTs connected to it.
2855 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2856 Details : lookup_it_finish() used to query file size from OSTs that was not
2861 Description: service threads change working directory to that of init
2862 Details : Starting lustre service threads may pin the working directory
2863 of the parent thread, making that filesystem busy. Threads
2864 now change to the working directory of init to avoid this.
2868 Frequency : during shutdown only
2869 Description: shutdown with a failed MDS or OST can cause unmount to hang
2870 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2875 Frequency : 2.6 only
2876 Description: chmod/chown may include an extra supplementary group
2877 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2878 supplementary group and if none is specified this is used.
2881 Bugzilla : 5479 (6816)
2883 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2884 Details : If lookup is in progress on a file that is unlinked we might try
2885 to revalidate the inode and fail in revalidate after lookup is
2886 complete and ll_file_open() enqueues the open again but
2887 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2890 Frequency : always, if lconf --abort_recovery used
2892 Description: lconf --abort_recovery fails with 'Operation not supported'
2893 Details : lconf was attempting to abort recovery on the MDT device and not
2896 ------------------------------------------------------------------------------
2898 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2903 Frequency : rare (only unsupported configurations with a node running as an
2905 Bugzilla : 6514, 5137
2906 Description: Mounting a Lustre file system on a node running as an OST could
2908 Details : OSTs now preallocates memory needed to write out data at
2909 startup, instead of when needed, to avoid having to
2910 allocate memory in possibly low memory situations.
2911 Specifically, if the file system is mounted on on OST,
2912 memory pressure could force it to try to write out data,
2913 which it needed to allocate memory to do. Due to the low
2914 memory, it would be unable to do so and the node would
2915 become unresponsive.
2917 Severity : enhancement
2919 Description: Addition of lconf --service command line option
2920 Details : lconf now accepts a '--service <arg>' option, which is
2921 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2923 Severity : enhancement
2925 Description: Failover mode is now the default for OSTs.
2926 Details : By default, OSTs will now run in failover mode. To return to
2927 the old behaviour, add '--failout' to the lmc line for OSTs.
2929 Severity : enhancement
2931 Description: Health checks are now provided for MDS and OSTs
2932 Details : Additional detailed health check information on MSD and OSTs
2933 is now provided through the procfs health_check value.
2936 Frequency : occasional, depends on IO load
2938 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2939 after numerous create/delete cycles
2940 Details : The ext3 inode allocation policy would not allocate new inodes
2941 very well on the OSTs because there are no new directories
2942 being created. Instead we look for groups with free space if
2943 the parent directories are nearly full.
2948 Description: Network or server problems during mount may cause partially
2949 mounted clients instead of returning an error.
2950 Details : The config llog parsing code may overwrite the error return
2951 code during mount error handling, returning success instead
2957 Description: MDS can fail to allocate large reply buffers
2958 Details : After long uptimes the MDS can fail to allocate large reply
2959 buffers (e.g. zconf client mount config records) due to memory
2960 fragmentation or consumption by the buffer cache. Preallocate
2961 some large reply buffers so that these replies can be sent even
2962 under memory pressure.
2966 Frequency : rare (liblustre)
2967 Description: fsx running with liblustre complained that using truncate() to
2968 extend the file doesn't work. This patch corrects that issue.
2969 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2970 ATTR_SIZE and lsm use in llu_setattr_raw.
2974 Frequency : rare, only 2.6 kernels
2975 Description: Unusual file access patterns on the MDS may result in inode
2976 data being lost in very rare circumstances.
2977 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2978 code in the 2.6 kernel caused the inode and/or EA data not to
2979 be read from disk, causing single-file corruption.
2983 Frequency : rare, only 2.6 filesystems using extents
2984 Description: Heavy concurrent write and delete load may cause data corruption.
2985 Details : It was possible under high-load situations to have an extent
2986 metadata block in the block device cache from a just-unlinked
2987 file overwrite a newly-allocated data block. We now unmap any
2988 metadata buffers that alias just-allocated data blocks.
2992 Frequency : filesystems with default stripe_count larger than 77
2993 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2994 Details : lconf specifies an inode size of 4096 bytes when the default
2995 stripe_count is larger than 77. This conflicts with the default
2996 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2997 this case to avoid pinning too much memory for large EAs.
2999 ------------------------------------------------------------------------------
3001 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3006 Frequency : rare (extremely heavy IO load with hundreds of clients)
3008 Description: Client is evicted, gets IO error writing to file
3009 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3010 caused clients to be evicted for AST timeouts. The fixes in
3011 bug 5192 mean we no longer need to have such short AST timeouts
3012 so ldlm_timeout has been increased.
3015 Frequency : occasional during --force or --failover shutdown under load
3016 Bugzilla : 5949, 4834
3017 Description: Server oops/LBUG if stopped with --force or --failover under load
3018 Details : a collection of import/export refcount and cleanup ordering
3019 issues fixed for safer force cleanup
3022 Frequency : only filesystems larger than 120 OSTs
3023 Bugzilla : 5990, 6223
3024 Description: lfs getstripe would oops on a very large filesystem
3025 Details : lov_getconfig used kfree on vmalloc'd memory
3028 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3029 Bugzilla : 6242, 6243
3030 Description: reading from files that had been truncated to a non-zero size
3031 but never opened returned no data
3032 Details : ll_file_read() reads zeros from no-object files to EOF
3037 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3038 Details : There's little harm in aborting MDS/OSS recovery and letting it
3039 try again, so I removed the LASSERT and return an error instead.
3041 Severity : enhancement
3043 Description: New debugging infrastructure for tracking down data corruption
3044 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3045 (b) cover more of the client-side code path, and (c) try to narrow
3046 down where problems occurred
3050 Bugzilla : 3819, 4364, 4397, 6313
3051 Description: Racing close and eviction MDS could cause assertion in mds_close
3052 Details : It was possible to get multiple mfd references during close and
3053 client eviction, leading to one thread referencing a freed mfd.
3055 Severity: : enhancement
3056 Bugzilla : 3262, 6359
3057 Description: Attempts to reconnect to servers are now more aggressive.
3058 Details : This builds on the enhanced upcall-less recovery that was added
3059 in 1.4.2. When trying to reconnect to servers, clients will
3060 now try each server in the failover group every 10 seconds. By
3061 default, clients would previously try one server every 25 seconds.
3066 Description: After recovery, certain operations trigger a failed
3067 assertion on a client.
3068 Details : Failing over an mds, using lconf -d --failover, while a
3069 client was doing a readdir() call would cause the client to
3070 LBUG after recovery completed and the readdir() was resent.
3072 Severity : enhancement
3074 Description: Default groups are now added by lconf
3075 Details : You can now run lconf --group <servicename> without having to
3076 manually add groups with lmc.
3079 Frequency : occasional
3081 Description: Nodes with an elan id of 0 trigger a failed assertion
3084 Frequency : always when accessing e.g. tty/console device nodes
3086 Description: tty and some other devices nodes cannot be used on lustre
3087 Details : file's private_data field is used by device data and lustre
3088 values in there got lost. New field was added to struct file to
3089 store fs-specific private data.
3092 Frequency : when exporting Lustre via NFS
3094 Description: NFSD failed occasionally when looking up a path component
3095 Details : NFSD is looking up ".." which was broken in ext3 directories
3096 that had grown large enough to become hashed.
3099 Frequency : Clusters with multiple interfaces not on the same subnet
3101 Description: Nodes will repeatedly try to reconnect to an interface which it
3102 cannot reach and report an error to the log.
3103 Details : Extra peer list entries will be created by lconf with some peers
3104 unreachable. lconf now validates the peer before adding it.
3107 Frequency : Only if a default stripe is set on the filesystem root.
3109 Description: Setting a default stripe on the filesystem root prevented the
3110 filesystem from being remounted.
3111 Details : The client was sending extra request flags in the root getattr
3112 request and did not allocate a reply buffer for the dir EA.
3115 Frequency : occasional, higher if lots of files are accessed by one client
3116 Bugzilla : 6159, 6097
3117 Description: Client trips assertion regarding lsm mismatch/magic
3118 Details : While revalidating inodes the VFS looks up inodes with ifind()
3119 and in rare cases can find an inode that is being freed.
3120 The ll_test_inode() code will free the lsm during ifind()
3121 when it finds an existing inode and then the VFS later attaches
3122 this free lsm to a new inode.
3126 Bugzilla : 6422, 7030
3127 Description: MDS deadlock between mkdir and client eviction
3128 Details : Creating a new file via mkdir or mknod (starting a transaction
3129 and getting the ns lock) can deadlock with client eviction
3130 (gets ns lock and trying to finish a synchronous transaction).
3133 Frequency : occasional
3134 Description: While starting a server, the fsfilt_ext3 module could not be
3136 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3137 kernels. Previously, lconf would still use the ext3 name
3138 when trying to load modules. Now, it will correctly use
3139 ext3 on 2.4 and ldiskfs on 2.6.
3141 Severity : enhancement
3142 Description: The default stripe count has been changed to 1
3143 Details : The interpretation of the default stripe count (0, to lfs
3144 or lmc) has been changed to mean striping across a single
3145 OST, rather than all available. For general usage we have
3146 found a stripe count of 1 or 2 works best.
3148 Severity : enhancement
3149 Description: Add support for compiling against Cray portals.
3150 Details : Conditional compiling for some areas that are different
3154 Frequency : occasional
3155 Bugzilla : 6409, 6834
3156 Description: Creating files with an explicit stripe count may lead to
3157 a failed assertion on the MDS
3158 Details : If some OSTs are full or unavailable, creating files may
3159 trigger a failed assertion on the MDS. Now, Lustre will
3160 try to use other servers or return an error to the
3164 Frequency : occasional
3166 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3167 caused lock timeout during readahead (since 1.4.2).
3168 Details : Processes doing readahead might match a lock that hasn't been
3169 granted yet if there are overlapping and conflicting lock
3170 requests. The readahead process waits on ungranted lock
3171 (original lock is CBPENDING), while OST waits for that process
3172 to cancel CBPENDING read lock and eventually evicts client.
3174 Severity : enhancement
3176 Description: Initial enabling of flock support for clients
3177 Details : Implements fcntl advisory locking and file status functions.
3178 This feature is provided as an optional mount flag (default
3179 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3180 locking are implemented yet, and those that are are not guaranteed
3181 to be completely correct in production environments.
3182 mount -t lustre -o [flock|noflock] ...
3185 Frequency : occasional
3187 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3188 assertion in the ext3 JBD (journaling) layer.
3189 Details : The b_committed_data struct is protected by the big kernel lock
3190 in 2.4 kernels, serializing journal_commit_transaction() and
3191 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3192 access to this struct. In 2.6 kernels there is finer grained
3193 locking to improve SMP performance of the JBD layer.
3197 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3202 Description: merge and cleanup kernel patches.
3205 -----------------------------------------------------------------------------
3207 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3209 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3210 versions of Lustre. Please update all servers and clients to
3211 version 1.4.2 or later at the same time. You must also run
3212 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3213 to update the configuration logs.
3215 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3216 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3217 - swab llog records of type '0' so we get proper header size/idx (5861)
3218 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3219 - fix rename of one directory over another leaking an inode (5953)
3220 - avoid SetPageDirty on 2.6 (5981)
3221 - don't re-add just-being-destroyed locks to the waiting list (5653)
3222 - when creating new directories, inherit the parent's custom
3223 striping settings if present parent (3048)
3224 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3225 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3226 - don't hold client locks on temporary worklist from l_lru (5666)
3227 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3228 - automatically evict dead clients (5921)
3229 - Update file size properly in create+truncate+fstat case (6196)
3230 - Do not unhash mountpoint dentries, do not allow removal of
3232 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3233 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3234 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3235 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3236 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3237 - Correctly handle reads of files with no objects (6243)
3238 - lctl recover will also mark a device active if deactivate used (5933)
3240 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3241 - allow --write-conf on an MDS with different nettype than client (5619)
3242 - don't write config llogs to MDS for mounts not from that MDS (5617)
3243 - lconf should create multiple TCP connections from a client (5201)
3244 - init scripts are now turned off by default; run chkconfig --on
3245 lustre and chkconfig --on lustrefs to use them
3246 - upcalls are no longer needed for clients to recover to failover
3248 - add --abort-recovery option to lconf to abort recovery on device
3250 - add support for an arbitrary number of OSTs (3026)
3251 - Quota support protocol changes.
3252 - forward compatibility changes to wire structs (6007)
3253 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3254 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3255 - improved "lustre status" script
3256 - initialize blocksize for non-regular files (6062)
3257 - added --disable-server and --disable-client configure options (5782)
3258 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3259 - Vanilla 2.4.29 support
3260 - increase maximum number of obd devices to 520 (6242)
3261 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3262 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3263 SLES 9 2.6.5 kernels
3264 - limit stripes per file to 160 (the maximum EA size) (6093)
3266 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3269 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3270 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3271 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3272 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3273 - small ext3 extents cleanups and fixes (5733)
3274 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3275 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3276 - avoid needless client->OST connect, fix handle mismatch (5317)
3277 - fix DLM error path that led to out-of-sync client, long delays (5779)
3278 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3279 - fix several locking issues related to i_size (5492,5624,5654,5672)
3280 - don't move pending lock onto export if it is already evicted (5683)
3281 - fix kernel oops when creating .foo in unlinked directory (5548)
3282 - fix deadlock in obdfilter statistics vs. object create (5811)
3283 - use time_{before,after} to avoid timer jiffies wrap (5882)
3284 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3285 - Do not leak request if server was not able to process it (5154)
3286 - If mds_open unable to find parent dir, make that negative lookup(5154)
3287 - don't create new directories with extent-mapping (5909, 5936)
3289 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3290 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3291 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3292 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3293 - support for RHEL4 kernel on i686 (5773)
3294 - provide error messages when incompatible logs are encountered (5898)
3296 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3297 * version 1.4.0.10 (1.4.1 release candidate 1)
3299 - don't keep a lock reference when lock is not granted (4238)
3300 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3301 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3302 - drop import inflight refcount on signal_completed_replay error (5255)
3303 - unlock page after async write error during send (3677)
3304 - handle missing objects in filter_preprw_read properly (5265)
3305 - no transno return for symlink open, don't save no-trasno open (3440)
3306 - don't try to complete elan receive that already failed (4012)
3307 - free RPC server reply state on error (5406)
3308 - clean up thread from ptlrpc_start_thread() on error (5160)
3309 - readahead could read extra page into cache that wasn't ejected (5388)
3310 - prevent races in class_attach/setup/cleanup/detach (5260)
3311 - don't dereference de->d_inode after l_dput of de (5458)
3312 - use "int" for stripe value returned from lock_to_stripe (5544)
3313 - mballoc allocation and error-checking fixes in 2.6 (5504)
3314 - block device patches to fix I/O request sizes in 2.6 (5482)
3315 - look up hostnames for IB nals (5602)
3316 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3317 - don't start multiple acceptors for the same port (5277)
3318 - fix incorrect LASSERT in mds_getattr_name (5635)
3319 - export a proc file for general "ping" checking (5628)
3320 - fix "lfs check" to not block when the MDS is down (5628)
3322 - service request history (4965)
3323 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3324 - create an "evict_client" /proc entry on OSTs, like the MDS has
3325 - fix mount usage message, return errors per mount(8) (5168)
3326 - change grep [] to grep "[]" in tests so they work in more UMLs
3327 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3328 - remove ancient LOV_MAGIC_V0 EA support (5047)
3329 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3330 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3331 - allow readahead allocations to fail when low on memory (5383)
3332 - mmap locking landed again, after considerable improvement (2828)
3333 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3335 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3338 - send OST transaction number in read/write reply to free req (4966)
3339 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3340 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3341 - the watchdog thread now runs as interruptible (5246)
3342 - flock/lockf fixes (but it's still disabled, pending 5135)
3343 - don't use EXT3 constants in llite code (5094)
3344 - memory shortage at startup could cause assertion (5176)
3346 - reorganization of lov code
3347 - single portals codebase
3349 - add extents/mballoc support (5025)
3350 - direct I/O reads in the obdfilter (4048)
3351 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3353 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3356 - send OST transaction number in read/write reply to free req (4966)
3357 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3358 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3359 - the watchdog thread now runs as interruptible (5246)
3360 - handle missing objects in filter_preprw_read properly (5265)
3361 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3362 - drop import inflight refcount on signal_completed_replay error (5255)
3363 - unlock page after async write error during send (3677)
3364 - return original error code on reconstructed replies (3761)
3365 - no transno return for symlink open, don't save no-trasno open (3440)
3367 - add pid to ldlm debugging output (4922)
3368 - bump the watchdog timeouts -- we can't handle 30sec yet
3369 - extra debugging for orphan dentry/inode bug (5259)
3371 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3374 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3375 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3376 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3377 - fix NULL dereference in /proc/sys/portals/routes (4827)
3378 - allow failed mdc_close() operations to be interrupted (4561)
3379 - stop precreate on OST before MDS would time out on it (4778)
3380 - don't send partial-page writes before EOF from client (4410)
3381 - discard client grant for sub-page writes on large-page clients (4520)
3382 - don't free dentries not owned by NFS code, check generation (4806)
3383 - fix lsm leak if mds_create_objects() fails (4801)
3384 - limit debug_daemon file size, always print CERROR messages (4789)
3385 - use transno after validating reply (3892)
3386 - process timed out requests if import state changes (3754)
3387 - update mtime on OST during writes, return in glimpse (4829)
3388 - add mkfsoptions to LDAP (4679)
3389 - use ->max_readahead method instead of zapping global ra (5039)
3390 - don't interrupt __l_wait_event() during strace
3392 - add software watchdogs to catch hung threads quickly (4941)
3393 - make lustrefs init script start after nfs is mounted
3394 - fix CWARN/ERROR duplication (4930)
3395 - return async write errors to application if possible (2248)
3396 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3397 - print NAL number in %x format (4645)
3398 - update barely-supported suse-2.4.21-171 series (4842)
3399 - support for sles 9 %post scripts
3400 - support for building 2.6 kernel-source packages
3401 - support for sles km_* packages
3403 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3406 - ignore -ENOENT errors in osc_destroy (3639)
3407 - notify osc create thread that OSC is being cleaned up (4600)
3408 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3409 - reconstruct ost_handle() like mds_handle() (4657)
3410 - create a new thread to do import eviction to avoid deadlock (3969)
3411 - let lconf resolve symlinked-to devices (4629)
3412 - don't unlink "objects" from directory with default EA (4554)
3413 - hold socknal file ref over connect in case target is down (4394)
3414 - allow more than 32000 subdirectories in a single directory (3244)
3415 - fix blocks count for O_DIRECT writes (3751)
3416 - OST returns ENOSPC from object create when no space left (4539)
3417 - don't send truncate RPC if file size isn't changing (4410)
3418 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3419 - bind to privileged port in socknal and tcpnal (3689)
3421 - rate limit CERROR/CWARN console message to avoid overload (4519)
3422 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3423 - basic mmap support (3918)
3424 - kernel patch series update from b1_4 (4711)
3426 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3429 - avoid crash during MDS cleanup with OST shut down (2775)
3430 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3431 - don't use bad inodes on the MDS (3744)
3432 - dynamic object preallocation to improve recovery speed (4236)
3433 - don't hold spinlock over lock dumping or change debug flags (4401)
3434 - don't zero obd_dev when it is force cleaned (3651)
3435 - print grants to console if they go negative (4431)
3436 - "lctl deactivate" will stop automatic recovery attempts (3406)
3437 - look for existing locks in ldlm_handle_enqueue() (3764)
3438 - don't resolve lock handle twice in recovery avoiding race (4401)
3439 - revalidate should check working dir is a directory (4134)
3441 - don't always mark "slow" obdfilter messages as errors (4418)
3443 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3446 - don't close LustreDB during write_conf until it is done (3860)
3447 - fix typo in lconf for_each_profile (3821)
3448 - allow dumping logs from multiple threads at one time (3820)
3449 - don't allow multiple threads in OSC recovery (3812)
3450 - fix debug_size parameters (3864)
3451 - fix mds_postrecov to initialize import for llog ctxt (3121)
3452 - replace config semaphore with spinlock (3306)
3453 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3454 - don't allow enqueue to complete on a destroyed export (3822)
3455 - down write_lock before checking llog header bitmap (3825)
3456 - recover from lock replay timeout (3764)
3457 - up llog sem before sending rpc (3652)
3458 - reduce ns lock hold times when setting kms (3267)
3459 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3460 - fix NULL deref and obd_dev leak on setup error (3312)
3461 - replace some LBUG about llog ops with error handling (3841)
3462 - don't match INVALID dentries from d_lookup and spin (3784)
3463 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3464 - fix invalid assertion in ptlrpc_set_wait (3880)
3466 - add libwrap support for the TCP acceptor (3996)
3467 - add /proc/sys/portals/routes for non-root route listing (3994)
3468 - allow setting MDS UUID in .xml (2580)
3469 - print the stack of a process that LBUGs (4228)
3471 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3474 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3475 - make sure to unset replay flag from failed open requests (3440)
3476 - if default stripe count is 0, use OST count for inode size (3636)
3477 - update parent mtime/ctime on client for create/unlink (2611)
3478 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3479 - free recovery state on server during a forced cleanup (3571)
3480 - unregister_reply for resent reqs (3063)
3481 - loop back devices mounting and status check on 2.6 (3563)
3482 - fix resource-creation race that can provoke i_size == 0 (3513)
3483 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3484 - more debugging for page-accounting assertion (3746)
3485 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3486 - avoid deadlock after precreation failure (3758)
3487 - fix race and lock order deadlock in orphan handling (3450, 3750)
3488 - add validity checks when grabbing inodes from l_ast_data (3599)
3490 - add /proc/.../recovery_status to obdfilter (3428)
3491 - lightweight CDEBUG infrastructure, debug daemon (3668)
3492 - change default OSC RPC parameters to be better on small clusters
3493 - turn off OST read cache for files smaller than 32MB
3494 - install man pages and include them in rpms (3100)
3495 - add new init script for (un)mounting lustre filesystems (2593)
3496 - run chkconfig in %post for init scripts (3701)
3497 - drop scimac NAL (unmaintained)
3499 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3502 - clean kiobufs before and after use (3485)
3503 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3504 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3505 - change init script to fail more clearly if not run as root (1528)
3506 - allow clients to reconnect during replay (1742)
3507 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3508 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3509 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3510 - fix cond_resched (3554)
3511 - only evict unfinished clients after recovery (3515)
3512 - allow bulk resend, prevent data loss (3570)
3513 - dynamic ptlrpc request buffer allocation (2102)
3514 - don't allow unlinking open directory if it isn't empty (2904)
3515 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3516 - remove extraneous obd dereference causing LASSERT failure (3334)
3517 - don't use get_cycles() when creating temp. files on the mds (3156)
3518 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3519 - handle EEXIST for set-stripe, set proper directory name (3336)
3521 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3522 - fix ksocknal_fmb_callback() error messages (2918)
3524 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3527 - don't copy lvb into (possibly NULL) reply on error (2983)
3528 - don't deref dentry after dput, don't free lvb on error (2922)
3529 - use the kms to determine writeback rpc length (2947)
3530 - increment oti_logcookies when osc is inactive (2948)
3531 - update client's i_blocks count via lvb messages (2543)
3532 - handle intent open/close of special files properly (1557)
3533 - mount MDS with errors=remount-ro, like obdfilter (2009)
3534 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3535 - don't use cancelling-locks' kms values (2947)
3536 - use highest lock extent for kms, not last one (2925)
3537 - don't dereference ERR_PTR() dentry in error handling path (3107)
3538 - fix thread race in portals_debug_dumplog() (3122)
3539 - create lprocfs device entries at setup instead of at attach (1519)
3540 - common AST error handler, don't evict client on completion race (3145)
3541 - zero nameidata in detach_mnt in 2.6 (3118)
3542 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3543 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3544 - initialize RPC timeout timer earlier for 2.6 (3219)
3545 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3546 - print nal/nid for unknown nid (3258)
3547 - additional checks for oscc recovery before doing precreate (3284)
3548 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3549 - don't crash in mdc_close for bad permissions on open (3285)
3550 - zero i_rdev for non-device files (3147)
3551 - clear page->private before handing to FS, better assertion (3119)
3552 - tune the read pipeline (3236)
3553 - fix incorrect decref of invalidated dentry (2350)
3554 - provide read-ahead stats and refine rpc in flight stats (3328)
3555 - don't hold journal transaction open across create RPC (3313)
3556 - update atime on MDS at close time (3265)
3557 - close LDAP connection when recovering to avoid server load (3315)
3558 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3559 - don't leak open file on MDS after open resend (3325)
3560 - serialize filter_precreate and filter_destroy_precreated (3329)
3561 - loop device shouldn't call sync_dev() for nul device (3092)
3562 - clear page cache after eviction (2766)
3563 - resynchronize MDS->OST in background (2824)
3564 - refuse to mount the same filesystem twice on same mountpoint (3394)
3565 - allow llmount to create routes for mounting behind routers (3320)
3566 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3567 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3568 - fix rare problem with rename on htree directories (3417)
3570 - allow default OST striping configuration per directory (1414)
3571 - fix compilation for qswnal for 2.6 kernels (3125)
3572 - increase maximum number of MDS request buffers for large systems
3573 - change liblustreapi to be useful for external progs like lfsck (3098)
3574 - increase local configuration timeout for slow disks (3353)
3575 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3577 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3580 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3581 - don't overwrite extent policy data in reply if lock was blocked (2901)
3582 - drop filter export grants atomically with removal from device (2663)
3583 - del obd_self_export from work_list in class_disconnect_exports (2908)
3584 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3585 - swab reply message in mdc_close, other PPC fixes (2464)
3586 - fix destroying of named logs (2325)
3587 - overwrite old logs when running lconf --write_conf (2264)
3588 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3589 - fix race in target_handle_connect (2898)
3590 - mds_reint_create() should take same inode create lock (2926)
3591 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3592 - don't close files for self_export to avoid uninitialized obd (2936)
3593 - allow MDS with the same name as client node (2939)
3594 - hold dentry reference for closed log files for unlink (2325)
3595 - reserve space for all logs during transactions (2059)
3596 - don't evict page beyond end of stripe extent (2925)
3597 - don't oops on a deleted current working directory (2399)
3598 - handle hard links to targets without a parent properly (2517)
3599 - don't dereference NULL lock when racing during eviction (2867)
3600 - don't grow lock extents when lots of conflicting locks (2919)
3602 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3605 - account for cache space usage on clients to avoid data loss (974)
3606 - lfsck support in lustre kernel code (2349)
3607 - reduce journal credits needed for BRW writes (2370)
3608 - orphan handling to avoid losing space on client/server crashes
3609 - ptlrpcd can be blocked, stopping ALL progress (2477)
3610 - use lock value blocks to assist in proper KMS, faster stat (1021)
3611 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3612 - recovery for initial connections (2355)
3613 - fixes for mds_cleanup_orphans (1934)
3614 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3615 - block all file creations until orphan recovery completes (1901)
3616 - client remove rq_connection from request struct (2423)
3617 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3618 - recovery timer race (2670)
3619 - mdc_close recovey bug (2532)
3620 - ptlrpc cleanup bug (2710)
3621 - mds timeout on local locks (2588)
3622 - namespace lock held during RPCs (2431)
3623 - handle interrupted sync write properly (2503)
3624 - don't try to handle a message that hasn't been replied to (2699)
3625 - client assert failure during cleanup after abort recovery (2701)
3626 - leak mdc device after failed mount (2712)
3627 - ptlrpc_check_set allows timedout requests to complete (2714)
3628 - wait for inflight reqs when ptlrpcd finishes (2710)
3629 - make sure unregistered services are removed from the srv_list
3630 - reset bulk XID's when resending them (caught by 1138 test)
3631 - unregister_bulk after timeout
3632 - fix lconf error (2694)
3633 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3634 - readahead locks pages, leaves pending causing memory pressure (2673)
3635 - increase OST request buffers to 4096 on large machines (2729)
3636 - fix up permission of existing directories in simple_mkdir (2661)
3637 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3638 - don't assign transno to errored transactions (2742)
3639 - don't delete objects on OST if given a bogus objid from MDS (2751)
3640 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3641 - if rq_no_resend, then timeout request after recovery (2432)
3642 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3643 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3644 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3645 - MDS recovery shouldn't delete objects that it has given out (2730)
3646 - if enqueue arrives after completion, don't clobber LVB (2819)
3647 - don't unlock pages twice when trigger_group_io returns error (2814)
3648 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3649 - don't write pages to disk if there was an error (1450)
3650 - don't ping imports that have recovery disabled (2676)
3651 - take buffered bytes into account when balancing socknal conn (2817)
3652 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3653 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3654 - remove little-endian swabbing of llog records (1987)
3655 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3656 - retry reposting request buffers if they fail (1191)
3657 - grow extent at grant time to avoid granting a revoked lock (2809)
3658 - lock revoke doesn't evict page if covered by a second lock (2765)
3659 - disable VM readahead to avoid reading outside lock extents (2805)
3661 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3662 - updated kernel patches for hp-2.4.20 kernel (2681)
3664 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3667 - fix truncated write corruption (2366)
3668 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3670 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3671 - ensure len(uuid) < 37 in lmc (1171)
3672 - fix ia64 OOPS in llog_test (2255)
3673 - zero end of page at obdfilter for partial page writes (2648)
3674 - don't leave stale dentries around after renames (bug 2428)
3675 - fix timeouts when evicting a client with a single lock held (2642)
3676 - set deadline for the initial HELLO message to drain (2634)
3677 - print out dotted-quad IP addresses in the socknal (2302)
3679 - additional debugging for MDS client eviction problem (2443)
3680 - fix mkfsoptions support for osts (2603, 2604)
3682 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3685 - add series for the vanilla 2.6.0 kernel
3686 - add series for the vanilla 2.4.24 kernel
3687 - add series for a cray x86/64 UL kernel drop
3688 - fix xattr patches for the vanilla 2.4.19 series
3690 - generate true UUIDs in lmc (1171)
3691 - have portals stack dumping break in UML (2466)
3692 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3693 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3694 - wake callback waiting threads on client eviction (2460)
3695 - Add --ptldebug and --subsystem to lmc (1719)
3696 - update assertion to allow safe interrupt allocation
3697 - set rq_no_resend for cancel requests (2432)
3698 - recalculate ptlrpcd timeout after resend (2494)
3699 - call vfs_rmdir when removing pending directories (2368)
3700 - fix renaming a file to itself (2429)
3701 - lmc creates a default one-stripe lov (2454)
3702 - expand procfs space to handle large clusters (2326)
3703 - increase UML stack to avoid overflow
3704 - update lconf's list of debug and subsystem masks
3705 - fix lfs find --obd (2510)
3706 - /proc tunable for disabling filter read caching (2591)
3707 - stop rpm packages from altering slapd.conf (2301)
3708 - disable nagle in the socknal under 0conf (2578)
3709 - choose mds inode size based on stripe count (2572)
3710 - fix kernel-source rpm problems (2516)
3712 - add --disable-doc to avoid pdf generation (2421)
3713 - update documentation, tests, type-os, comments
3714 - avoid format warnings on ia64
3715 - remove the TOE NAL
3716 - tiny code cleanups by removing unused fields
3718 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3721 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3722 - remove the most blatant lies from BUILDING file (2371)
3723 - change default debug level to reasonable production setting
3724 - reduce client side cache size to reduce cache flush time
3725 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3726 - make TCP zerocopy and pinger support enabled by default (2476)
3727 - sync writes completed after process exits caused crashes (2319)
3728 - maintain correct mount count on the MDS (2356)
3729 - backout 1557, because 2316 wasn't really fixed
3730 - better file I/O statistics gathering in /proc
3731 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3732 - another kernel patch to fix zero-copy TCP function export
3733 - don't take duplicate lock when processing re-sent getattr (2420)
3734 - lctl uses obd_self_export instead of creating new conn (2353)
3735 - MDS/OST recovery case which requires object creation asserted (2425)
3736 - move lfs from /usr/sbin to /usr/bin in packages
3737 - fix race between mds_client_add and mds_client_free (2417)
3738 - use kmalloc instead of slabs in portals (2430)
3739 - don't create duplicate records when a failover MDS is present (2442)
3740 - remove unnecessary mount age check (2332)
3741 - don't remove directory inodes from locks prematurely (2451)
3742 - don't break if MDS service name is the same as hostname (2103)
3743 - fix races in client write RPC generation when cache full (2482)
3745 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3748 - remove now-unused request->rq_obd (278)
3749 - if an allocation fails, print out how much memory we've used (1933)
3750 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3751 - add the "configurable stack size" patch to most series files (1256)
3752 - ability to write large log records, for 100+ OST configs (2306)
3753 - fix NULL deref when filter_prep fails (2314)
3754 - fix operator precedence error in filter_sync
3755 - dynamic allocation of socknal TX descriptors (2315)
3756 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3757 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3758 - ensure that CWARN messages in Portals always get to the syslog
3759 - __init/__exit are not for prototype decls (ldlm_init/exit)
3760 - x86-64 compile warning fixes
3761 - fix gateway LMC keyword conflict (2318)
3762 - fix MDS lock inversions in getattr/reint paths (1844)
3763 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3764 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3765 - DLM race condition prevented some lock evictions (2328)
3766 - ENOMEM detection and retry on socknal sends (2230)
3767 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3768 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3769 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3770 - better error messages when a client is rejected during recovery (1505)
3771 - avoid cancelling locks which were never granted, after failure (2330)
3772 - fix i_sem/journal inversion in mds_client_add (2333)
3773 - fix truncate/getattr lock cycle deadlock (2334)
3774 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3775 - fix two rare exit paths which could leak an l_lock() ref (2321)
3776 - fencepost error in MDS/OST orphan recovery (2226)
3777 - make log record alignment 8 bytes (1988)
3778 - lstripe now fails when requested offset > ost_count (2237)
3779 - ensure that all kernel series have a complete list.h (1607)
3780 - fix crashes in special-file operations (2316)
3781 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3783 - allow configurable automake binary, for testing new versions
3784 - small update to the lfs documentation
3786 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3788 * fix negative export reference count in fsfilt_sync (2312)
3790 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3791 * release candidate 0.9.1
3793 - orphans are moved into the PENDING directory for possible recovery
3794 - replayed opens now open by fid for orphan/rename safety (1042)
3795 - last close of an orphan inode generates a transno (683)
3796 - chdir() and mount() now pin the directory entry (1020)
3797 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3798 - discard very old requests without processing them (1502)
3799 - remove some common, well-understood CERRORs (1505)
3800 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3801 - clear "grant" flags in OST replies until OST grant code lands (1644)
3802 - fix read performance by not clobbering i_blksize on client (1598)
3803 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3804 - make LRU size a /proc tunable, clears locks when reduced (707)
3805 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3806 - server locks take references on exports now (1558)
3807 - build fixes for 2.4.20-rh trees (1663)
3808 - return an error from lov_create if all OSCs are inactive (1751)
3809 - fix import levels when a reconnect happens without a timeout (1597)
3810 - exit early from mds_open if we get a lookup error (1749)
3811 - partial page read at EOF wouldn't wait for disk before sending (1642)
3812 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3813 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3814 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3815 - write/truncate lock inversion (1639)
3816 - Don't auto-load obdclass, portals modules during cleanup (1495)
3817 - fix timestamps from jumping to "now" (1763)
3818 - extra journal assertions (1648)
3819 - add an extra multiunlink test (1771)
3820 - fix read_record/write_record API (1776)
3821 - fix leak of offset_extent, possible incorrect i_size later (1772)
3822 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3823 - Debugging for the kqswnal_get_idle_tx problems (1820)
3824 - Allow recovery to be attempted multiple times (1536)
3825 - Write out MDS last_rcvd file after it is first created (1600)
3826 - Fix tx_descriptor leak in failed transmit situations (1827)
3827 - ext3 journaling fixes for assertion failure after IO error (1871)
3828 - class_export_put() on freed export after completion AST error (1896)
3829 - Fix revalidate looping in VFS (1322)
3830 - Don't access a freed export during MDS_REINT timeout (1521)
3831 - Add open-unlink recovery support on the MDS (1673,1764)
3832 - Return an error if no MDS data was read from last_rcvd (1946)
3833 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3834 - Don't setup a disk device that doesn't match exported UUID (317)
3835 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3836 - avoid committing NULL handle in force close
3837 - local.sh is now a one-stripe LOV configuration
3838 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3839 - fix ext3 htree duplicate directory entry corruption (1516)
3840 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3841 - update to new LOV EA format (2097)
3842 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3844 2003-06-15 Phil Schwan <phil@clusterfs.com>
3847 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3848 - per-import/export recovery handling (958, 931, 959)
3849 - multiple last-rcvd slots, for serving multiple FSes (949)
3850 - connections are again shared between multiple imp/exports (963, 964)
3851 - "umount -f" would hang if any requests needed to be sent (393, 978)
3852 - avoid pinning large req buffer by copying for queued messages (989)
3853 - add "uuid" to "lctl device" command to help upcalls (991)
3854 - "open" RPCs with transnos would confuse recovery counters (1037)
3855 - do proper endian conversion of all wire messages (288, 340, 891)
3856 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3857 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3858 - fix ptlrpc_connection leak in target_handle_connect (1174)
3859 - fix import refcounting bug in OST and MDS cleanup (1134)
3860 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3861 - fix very unlikely obd_types race condition (501)
3862 - remove osc_open hack for echo_client (1187)
3863 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3864 - a failure in read_inode2 leads to deadlock (1139)
3865 - cancel ack-locks as soon as transaction is committed (1072)
3866 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3867 - make sure to commitrw after any preprw to avoid deadlock (1162)
3868 - failing to execute a file in a lustre FS would lock inode (1203)
3869 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3870 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3871 - obd_uuid2tgt would walk off the end of the list (1255)
3872 - on IA64 the osc would give portals incorrect bulk size (1258)
3873 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3874 - fix lock inversion caused by new llite matching code (1282)
3875 - limit the number of dirty pages on a client to 10MB (1286)
3876 - timed out locks were not being corrected cancelled (1289)
3877 - fix O_DIRECT above 4GB on IA-32 (1292)
3878 * major user-visible changes
3879 - fail out/fail over policy now controlled by the upcall (993)
3881 - add OBD_PING to check server availability and failure (954)
3882 - lustre messages are now sent in sending host order (288, 340, 891)
3883 - add eadatalen to MDS getattr reply (340)
3884 - OST read replies may contain second buffer, with per-page status (593)
3886 2003-03-11 Phil Schwan <phil@clusterfs.com>
3889 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3890 - failed open()s could cause deadlock; fixed (867, 869)
3891 - stop cancelling OST locks when files are closed (481)
3892 - overlapping XID spaces caused network corruption (851, 853)
3893 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3894 - setattr_raw added, to do single-RPC, server-side setattrs
3895 - lmc/lconf syntax change for OST UUIDs
3896 - fix crashy race condition between ptlrpc_free_req and osc_close
3897 - don't use request in mdc_enqueue if we hit a timeout (889)
3898 - don't set the inode i_size for regular files from the MDS (896)
3899 - handle out of order completion AST (842)
3900 - don't LBUG if a lock request times out after receiving AST (913)
3901 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3902 - if a bad lock AST arrives, send an error instead of dropping entirely
3903 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3904 - fix leak in bulk IO when only partially completed (899, 900, 926)
3905 - fix O_DIRECT for ia64 (55)
3906 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3907 - C-z after timeout could hang a process forever; fixed (977)
3909 - client-side I/O cache (678, 924, 929, 941, 970)
3911 - READPAGE and SETATTRs which don't take server-side locks get
3914 2003-02-11 Phil Schwan <phil@clusterfs.com>
3917 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3918 - Fix fsfilt_extN_readpage() to read a full page of directory
3919 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3920 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3921 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3922 - Most intent operations were removed, in favour of a new RPC mode
3923 that does a single RPC to the server and bypasses most of the VFS
3924 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3925 - Aggressively cancel local locks on DLM servers
3926 - mds_reint_unlink sends EA to the client if it's the last nlink.
3927 client uses that EA to unlink OST objects.
3928 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3929 - recursive symlinks were fixed (439)
3930 - fixed NULL deref in DEBUG_REQ
3931 - filter_update_lastobjid no longer calls sync, which annoyed extN
3932 - fixed multi-client small-writes to a single file problem (445)
3933 - fixed mtime updates during file writes (607)
3934 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3935 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3936 - fixed Linux OST size reporting problem (444, 656)
3937 - OST now updates object mtime with writes or setattr (607, 619)
3938 - client verifies file size before zeroing page past EOF (445)
3939 - OST now writes last allocated objid to disk with allocation (108)
3940 - LOV on echo now works (409)
3942 - mds_reint_unlink sends a new buffer, with the EA included. this
3943 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3944 set if a regular file was being unlinked, and it was the last link
3945 - use PtlGet from the target for bulk writes (315)
3946 - OST now updates object mtime with writes or setattr (607, 619)
3947 - LDLM now has a grant-time callback to revalidate locked items, if
3949 - Many MDS operations were reorganized to combat race conditions
3951 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3952 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3954 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3957 - Fully reactivate OST imports after reconnection (512, others)
3958 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3959 - More graceful error handling for truncating on dead OST (515)
3960 - Don't error out unless we're actually accessing dead stripes (474)
3961 - Fix garbage sizes when stripes are missing (410)
3962 - LRU counters were broken, causing constant lock purge (433, 432)
3963 - garbage on read from stripes with failed OSTs (441)
3964 - mark OSCs as active before reconnecting during recovery (438)
3965 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3966 - lfind did not preserve OST order in output (443)
3967 - symlinks cause hung clients, incorrect data (439)
3968 - stop dereferencing request after dropping refcount (457)
3969 - don't LASSERT(spin_is_locked) on non-SMP (455)
3970 - fixes for many rename() bugs
3971 - fstat didn't correctly synchronize attributes (399)
3972 - server must handle lock cancellation during blocking AST prep (487)
3973 - bulk descriptors were free()d too soon (511)
3974 - fix paths in lconf, which would load incorrect modules (451, 507)
3975 - fix confusing lconf 'host not found' error message (386)
3976 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3977 - fix race condition in mdc_blocking_ast() for inode access (526)
3978 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3979 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3980 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3981 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3982 - fix lstripe in several ways (kernel side) (527)
3983 - fix request leak in ldlm_cli_enqueue (262)
3984 - incorrect OSC was marked inactive after OST failure
3985 - call mds_fs_cleanup before unmounting filesystem (524)
3986 - fix races between taking ns_lock and ldlm_lock_change_resource
3987 - fix races updating LOV export open file list
3988 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3989 - fix recovery NULL deref in ldlm_cli_cancel_unused
3990 - fix some DLM races by using new hash table for lock handles (419)
3991 - permit the client to specify desired inodes, at replay
3992 - duplicate requests when we queue them for replay reintegration
3993 - fix last_rcvd offset calculation
3994 - sync after each recovered transaction, so we always make progress
3995 - never, not always, ERESTART requests without transnos
3996 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3998 - skip replay if the MDS says that the client is already connected
3999 - don't check for a recovery-enabled export to match lctl's UUID
4000 - don't INC_USE_COUNT for phantom exports
4001 - don't crash when cleaning up phantom exports (567)
4002 - don't double-finish or set replay data for errored mdc_open requests
4003 - abort requests when they time out, so we don't get old replies
4004 - send/receive replies for AST messages again
4005 - if the client says that it doesn't have the lock, cancel it on the
4007 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4008 mark it as destroyed, it will all work out when decref is called
4009 - fix module use counts (22, 581)
4011 - ASTs now expect a reply (server cancels lock on error reply)
4013 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4016 - fix many simultaneous client startup (392)
4017 - fix dentry->d_it clobbering
4018 - credentials weren't being shipped for readdir/getattr operations
4019 - remove invalid assertions triggered during some concurrent MD
4021 - proper Lustre versions added (336, 389)
4022 - fix memory leak for create error case (398)
4023 - fix LOV locking bug that would get cli/srv out of sync
4024 - fix echo client over LOV (409)
4025 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4026 - fix double-O_EXCL intent crash (424)
4027 - avoid sending multiple lock CANCELs (352)
4029 - MDS can do multi-client recovery (modulo bugs in new code)
4031 - many updates, edits, cleanups
4033 2002-11-18 Phil Schwan <phil@clusterfs.com>
4036 - fix null d_it dereference (346)
4037 - fix full OST/dbench hang (333)
4038 - fix permission problem with file removal (286)
4039 - fix removal of OSCs from LOV when they fail
4040 - fix NULL deref during bulk timeout (214)
4041 - fix problems related to multiple filesystems on one MDS (241)
4042 - fixed serious subtle metadata locking bugs
4043 - free locks on clients when inodes are removed due to memory
4045 - fix inode pointer in lock data (285)
4046 - partial support for multiple MDS on a single host (241)
4047 - data locks weren't cancelled at clear_inode time (290, 311)
4048 - intent locks could lead to unbounded lock growth (205)
4049 - added a maximum lock count, an LRU list, and a flusher
4050 - fix multiple rename (365)
4051 - properly abstracted the echo client
4052 - OSC locked 1 byte too many; fixed
4053 - rewrote brw callback code:
4054 - fixed recovery bugs related to LOVs (306)
4055 - fixed too-many-pages-in-one-write crash (191)
4056 - fixed (again) crash in sync_io_timeout (214)
4057 - probably fixed callback-related race (385)
4059 - Add capability to MDS protocol
4060 - LDLM cancellations and callbacks on different portals
4062 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4065 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4066 - defer open object destruction to close time (601981, 138)
4067 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4068 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4069 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4070 - don't call obd_setattr in ll_file_release for destroyed objects
4071 * protocol change to lustre_msg: move |version| and add |flags|
4072 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4073 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4074 * added replay of create, unlink, link and rename operations during
4075 MDS failover; recovery should be much more robust now
4076 * remove failed OSCs from LOVs (only lov_create uses this so far)
4077 * the lustre-HOWTO was brought (more) up to date (582544)
4079 2002-10-23 Phil Schwan <phil@clusterfs.com>
4082 - in-use dentries weren't being reused properly (617851)
4083 - prevent multiple LDLM setup (599178)
4084 - fix LOV size calculations for truncate (617853)
4085 - fix client handling of MDS intent errors (POSIX)
4086 - fix permission bug in lovstripe.c test (624321)
4087 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4088 - truncate past end of file could corrupt data
4089 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4090 - a race in recovery could return ETIMEDOUT to apps (623947)
4091 - building outside the source directory was fixed
4092 * the lustre-HOWTO was brought (more) up to date (582544)
4093 * major progress was made on recovery functionality
4095 2002-10-10 Phil Schwan <phil@clusterfs.com>
4098 - recovery deadlock fix
4099 - rm -rf causes LBUG fix (617817)
4100 - file open by multiple tasks fix (618962)
4101 - directory permissions bugs (602707 and 620007)
4102 - journal_stop fixed with locking (611313)
4103 - O_APPEND failures resolved (618273, perhaps 614459)
4104 - lconf PATH fix (619770)
4105 - IA64 build fix (621450)
4106 - RPC buffer sizes scale with amount of memory
4108 2002-10-01 Phil Schwan <phil@clusterfs.com>
4111 - locks would be cancelled without throwing away data pages,
4112 resulting in inconsistent data (605627)
4113 - inode attributes were not always being refreshed (605627, 612449)
4114 - lconf now continues to cleanup after lctl reports an error
4115 - MDS now enforces user permissions (602707)
4116 - lprocfs cleanup fixed, but not yet enabled (614157)
4117 - fixed infinite server hang, should a client not respond to an AST
4118 - avoid going into recovery if user calls readlink() with a buffer
4119 that's too small (613941)
4120 - AST RPCs no longer require replies (614867) -- this may be changed
4121 - don't crash server if client sends an IOV that's too big (611336)
4122 - fixed lock conversion deadlock (611892)
4123 - fixed the following of symlinks (614622)
4124 * recovery: the server can remove locks from a client that dies, other
4125 clients can make progress
4126 * more extN patch fixes
4127 * compile-time configurable ptlrpc buffer allocations
4129 - collaborative read cache document
4130 - Lustre Lite Performance CDR document-in-progress
4132 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4135 - fix typo in patch-2.4.18
4137 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4140 - clear ptlrpc request each time in handle_incoming_request()
4141 - unlink of files now destroys the object on the OST
4143 2002-09-19 Peter Braam <braam@clusterfs.com>
4145 * add hard link support
4146 * change obdfile creation method
4147 * kernel patch changed
4149 2002-09-19 Peter Braam <braam@clusterfs.com>
4152 - stack overflow bug in extN fixed
4154 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4156 * documentation updates
4157 - add man pages for config tools
4158 - update tests/README to describe testing with new config tools
4159 - finish metadata API descriptions
4160 * bug fixes and cleanups
4161 - statfs workaround for 16TB limit
4162 - LOV stripe allocation improved, can stripe on subset of OSTs
4163 - LOV file size/IO offset was wrong for files > 4GB in size
4164 - object EA data was being dropped, caused files to be unreadable
4165 - memory overflow with non-LOV OST caused memory corruption
4166 - fixed regression tests to work with new config tools, obdfilter
4167 - fixed bug when directory size became larger than 1 block
4168 - fixed bug (for single client case) when PWD was deleted
4169 - invalidate local directory pages when doing intent-based ops
4170 - avoid LDLM oops when lock callback contained bad data
4172 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4174 * documentation updates
4175 * bug fixes and cleanups
4176 - configuration tools
4179 - 64-bit compile warnings
4180 - 64-bit internal statfs data
4182 * test_brw on persistent OST devices
4184 * lprocfs (disabled)
4186 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4188 * documentation updates
4189 * bug fixes and cleanups
4190 * configuration tools
4192 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4195 * many small fixes to 0_5_4
4196 * io/network handling
4197 * thinkos in MDS operations
4199 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4202 * crucial basic fixes to 0.5.3
4203 * IOR, Iozone work over Elan
4206 2002-08-07 Phil Schwan <phil@clusterfs.com>
4207 * version 0_5_3, our first alpha
4208 * we use the new Portals iovs
4209 * documentation updates
4210 * bug fixes and cleanups
4211 * small changes in the DLM wire protocol
4213 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4214 * version 0_5_1 with some initial stability,
4215 * locking on MD and file I/O.
4216 * documentation updates
4217 * several bug fixes since 0.5.0
4218 * small changes in wire protocol
4220 2002-07-18 Phil Schwan <phil@clusterfs.com>
4222 * delivered as Lustre Light Alpha
4223 * fixed a crash after handling invalid MDS requests
4224 * fixed directory pages for architectures with non-4k pages sizes
4226 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4227 * release version v0_4_4
4228 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4230 * Fixes a number of interruption problems with OST operations.
4231 * Update documentation for portals header changes
4232 * Move all wire protocol structs/defines to lustre_idl.h
4233 * Fixes symlink length bug.
4234 * Add tcpdump to repository.
4236 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4237 * release version v0_4_3
4238 * Fixes statfs for inodes on extN.
4239 * Fixes bug in runtests which would delete /etc/hosts.
4240 * Use 64-bit object IDs wherever possible (not into VFS though)
4241 Remove ost_get_info, which is unused by lustre, and out of date.
4243 2002-07-03 Peter Braam <braam@clusterfs.com>
4244 * release version v0_4_2 Fixes a lookup error (type not passed)
4245 * move forward to head of Portals
4246 * move forward to latest Lustre kernel
4248 2002-06-25 Peter Braam <braam@clusterfs.com>
4249 * release version v0_4_1. Hopefully stable on single node use.