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.
42 Remove lock_24kernel from lustre/lvfs/fsfilt_ext3.c.
44 Severity : enhancement
46 Description: Build SLES10 patchless client fails
47 Details : The configure was broken by run ./configure with
48 --with-linux-obj=.... argument for patchless client. When the
49 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
50 can't search header adequately. Use absolute path such as
51 -I($LINUX)/include instead.
55 Description: interrupt oig_wait produce painc on resend.
56 Details : brw_redo_request can be used for resend requests from ptlrpcd and
57 private set, and this produce situation when rq_ptlrpcd_data not
58 copyed to new allocated request and triggered LBUG on assert
59 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
60 ptlrpcd set if request is changed and can be safety changed to use
65 Description: LASSERT_{REQ,REP}SWAB macros are buggy
66 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
67 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
68 these macros and replace them with their difinitions instead.
72 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
73 Details : Update which_patch & target file for SLES10 latest kernel.
76 Description: Remove client patches from SLES 10 kernel.
77 Details : This causes SLES 10 clients to behave as patchless clients
78 even on a Lustre-patched (server) kernel.
80 Severity : enhancement
82 Description: self-adjustable client's lru lists
83 Details : use adaptive algorithm for managing client cached locks lru
84 lists according to current server load, other client's work
85 pattern, memory activities, etc. Both, server and client
86 side namespaces provide number of proc tunables for controlling
89 Severity : enhancement
91 Description: light-weight GSS support
92 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
93 and reduce performance overhead.
95 Severity : enhancement
97 Description: Linux keyring support
98 Details : Support using service of Linux keyring for Lustre GSS internal
99 context refresh/cache mechanism.
103 Description: Fix errors in lfs documentation
104 Details : Fixes man pages
108 Description: don't use GFP_* in generic Lustre code.
109 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
113 Description: obdclass is limited by single OBD_ALLOC(idarray)
114 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
118 Description: warnings with build patchless client with vanila 2.6.19 and up
119 Details : change old ctl_table style and replace ctl_table/ctl_table_header
120 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
124 Frequency : Only for SLES
125 Description: sanity_quota fail test_1
126 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
127 file, which makes TSTID[2] inunique.
131 Frequency : Only for SLES9
132 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
133 Details : Update target/ChangeLog/which_patch .
137 Frequency : Only for test_10 in sanity-quota.sh
138 Description: a bug in quota test code
139 Details : checking a test flag in a wrong place
143 Frequency : Only for test_13 in sanity-quota.sh
144 Description: a bug in quota test script
145 Details : shouldn't check overfull allocated quota
149 Description: jbd statistics
150 Details : Port older jbd statistics patch for sles10
152 Severity : enhancement
154 Description: Add printing inode info into message about error in writepage.
159 Description: reply_lock_interpret crash due to race with it and lock cancel.
160 Details : Do not replay locks that are being cancelled. Do not reference
161 locks by their address during replay, just by their handle.
165 Frequency : When flocks are used.
166 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
167 Details : Flock locks might destroy just granted lock if it could be merged
168 with another existing flock, this is done in completion handler,
169 so teach ldlm_cli_enquque_fini that this is a valid case for
175 Description: Oops in read and write path when failing to allocate lock.
176 Details : Check if lock allocation failed and return error back.
180 Description: lstripe command fails for valid OST index
181 Details : The stripe offset is compared to lov->desc.ld_tgt_count
182 instead of lov->desc.ld_active_tgt_count.
186 Description: sanity.sh failed test 103
187 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
192 Description: df doesn't work properly if diskfs blocksize != 4K
193 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
197 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
198 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
202 Description: obd_health_check_timeout too short
203 Details : set obd_health_check_timeout as 1.5x of obd_timeout
207 Description: llapi_file_create() does not allow some changes
208 Details : add llapi_file_open() that allows specifying the mode and
209 open flags, and also returns an open file handle.
213 Description: merge and cleanup kernel patches.
214 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
218 Description: Add journal checksum support.(Kernel part)
219 Details : The journal checksum feature adds two new flags i.e
220 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
221 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
222 indicates that the commit block contains the checksum for
223 the blocks described by the descriptor blocks. Now commit
224 record can be sent to disk without waiting for descriptor
225 blocks to be written to disk. This behavior is controlled
226 using JBD2_FEATURE_ASYNC_COMMIT flag.
230 Description: OSS needs mutliple precreate threads
231 Details : Add ability to start more than one create thread per OSS.
235 Description: Sanity.sh test_65k failed: setstripe should have succeeded
236 Details : execute lfs setstripe on client
240 Description: mds_obd_create error creating tmp object
241 Details : When the user sets quota on root, llog will be affected and can't
242 create files and write files.
245 Frequency : Always on ia64 patchless client, and possibly others.
247 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
248 Details : This allows the patchless client to be loaded on architectures
253 Description: RedHat Update kernel for RHEL5
254 Details : Add the kernel config file for RHEL5.
258 Description: RedHat Update kernel for RHEL5
259 Details : Modify the kernel config file more closer RHEL5.
263 Description: Build failure against Centos5 (RHEL5)
264 Details : Define PAGE_SIZE when it isn't present.
268 Description: test_7 on sanity-quota.sh can't be used on separate machines
269 Details : add facet to handle it
273 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
274 Details : with NFS, the anon dentry's parent was set to itself in
275 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
276 corresponding dentry other than use rec->ur_name.
278 Severity : enhancement
280 Description: lfs setstripe enhancement
281 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
285 Description: testing performance impact of enabling checksumming
286 Details : enable checksum by default, allow --disable-checksum
287 configure option and "-o nochecksum" mount option
291 Description: lustre support for RHEL5
292 Details : Add support for RHEL5.
296 Description: Client eviction due to failover config
297 Details : after a connection loss, the lustre client should attempt to
298 reconnect to the last active server first before trying the
299 other potential connections.
303 Description: when mds and osts use different quota unit(32bit and 64bit),
304 quota will be released repeatly.
305 Details : void sending multiple quota reqs to mds, which will keep the
306 status between the reqs.
310 Description: osts not allocated evenly to files
311 Details : change the condition to increase offset_idx
314 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
316 Description: Data corruption for OSTs that are formatted larger than 2TB
318 Details : When generating the bio request for lustre file writes the
319 sector number would overflow a temporary variable before being
320 used for the IO. The data reads correctly from Lustre (which
321 will overflow in a similar manner) but other file data or
322 filesystem metadata may be corrupted in some cases.
326 Description: Tune the kernel for good SCSI performance.
327 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
328 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
333 Description: rewrite ext2-derived code in obdclass/uuid.c
334 Details : rewrite inherited code (uuid parsing code from ext2 utils)
335 from scratch preserving functionality.
339 Description: "lfs find -obd UUID" prints directories
340 Details : "lfs find -obd UUID" will return all directory names instead
341 of just file names. It is incorrect because the directories
342 do not reside on the OSTs.
346 Description: use i_size_read and i_size_write in 2.6 port
347 Details : replace inode->i_size access with i_size_read/write()
350 Frequency : when removing large files
352 Description: scheduling issue during removal of large Lustre files
353 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
354 It causes scheduling issues when removing large files (17TB in the
358 Frequency : only with liblustre clients on XT3
360 Description: evictions taking too long
361 Details : allow llrd to evict clients directly on OSTs
364 Frequency : only on ppc
366 Description: /proc/fs/lustre/devices broken on ppc
367 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
368 We should make sure the type of 'index' is loff_t and then cast
369 explicitly as needed below. Do not assign an explicitly cast
373 Frequency : only for rhel5
375 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
376 Details : Modify the target file & which_kernel.
378 Severity : enhancement
380 Description: omit set fsid for export NFS
381 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
382 as FS_REQUIRES_DEV for avoid problems with generate fsid.
386 Frequency : Always, for kernels after 2.6.16
387 Description: Fix warning idr_remove called for id=.. which is not
389 Details : Last kernels save old s_dev before kill super and not allow
390 to restore from callback - restore it before call kill_anon_super.
394 Description: buffer overruns could theoretically occur
395 Details : llapi_semantic_traverse() modifies the "path" argument by
396 appending values to the end of the origin string, and a
397 overrun may occur. Adding buffer overrun check in liblustreapi.
401 Description: Fix error on 'ls .' at the top of the Lustre mount.
402 Details : Don't revalidate dentry if it is a root dentry.
406 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
407 Details : Modify vm-tunables-rhel4.patch.
411 Description: Kernel config for 2.6.18-vanilla.
412 Details : Modify targets/2.6-vanilla.target.in.
413 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
414 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
415 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
416 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
420 Description: improve handling recoverable errors
421 Details : if request processig with error which can be recoverable on server
422 request should be resend, otherwise page released from cache and
427 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
428 Details : Modify target file & which_patch.
429 A flaw was found in the IA32 system call emulation provided
430 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
431 value could be stored in the %RAX register, which could trigger an
432 out-of-bounds system call table access. An untrusted local user
433 could exploit this flaw to run code in the kernel
434 (ie a root privilege escalation). (CVE-2007-4573).
438 Description: O_DIRECT bypasses client statistics.
439 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
440 read_ahead_stats were updated. copied the stats section from
441 osc_send_oap_rpc() into async_internal().
445 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
449 Description: change order of libsysio includes
450 Details : '#include sysio.h' should always come before '#include xtio.h'
454 Description: handle "serious error: objid * already exists" more gracefully
455 Details : If LAST_ID value on disk is smaller than the objects existing in
456 the O/0/d* directories, it indicates disk corruption and causes an
457 LBUG(). If the object is 0-length, then we should use the existing
458 object. This will help to avoid a full fsck in most cases.
460 Severity : enhancement
462 Description: adapt the lustre_config script to support the upgrade case
463 Details : Add "-u" option for lustre_config script to support upgrading 1.4
464 server targets to 1.6 in parallel.
468 Description: To avoid grant space > avaible space when the disk is almost
469 full. Without this patch you might see the error "grant XXXX >
470 available" or some LBUG about grant, when the disk is almost
472 Details : In filter_check_grant, for non_grant cache write, we should
473 check the left space by if (*left > ungranted + bytes), instead
474 of (*left > ungranted), because only we are sure the left space
475 is enough for another "bytes", then the ungrant space should be
476 increase. In client, we should update cl_avail_grant only there
477 is OBD_MD_FLGRANT in the reply.
481 Description: Update RHEL 4 kernel to fix local root privilege escalation.
482 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
483 described in CVE-2007-4573. This problem could allow untrusted
484 local users to gain root access.
488 Frequency : when using O_DIRECT and quotas
490 Description: Incorrect file ownership on O_DIRECT output files
491 Details : block usage reported by 'lfs quota' does not take into account
492 files that have been written with O_DIRECT.
494 --------------------------------------------------------------------------------
496 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
498 * Support for kernels:
499 2.6.5-7.283 (SLES 9),
500 2.6.9-55.EL (RHEL 4),
501 2.6.16.46-0.14 (SLES 10),
502 2.6.18.8 vanilla (kernel.org)
503 * Client support for unpatched kernels:
504 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
505 2.6.16 - 2.6.22 vanilla (kernel.org)
506 * Due to recently discovered recovery problems, we do not recommend
507 using patchless RHEL 4 clients with this or any earlier release.
508 * Recommended e2fsprogs version: 1.39.cfs8
509 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
510 * Starting with this release, the ldiskfs backing filesystem required
511 by Lustre is now in its own package, lustre-ldiskfs. This package
512 should be installed. It is versioned separately from Lustre and
513 may be released separately in future.
517 Description: block reactivating mgc import until all deactivates complete
518 Details : Fix race when failing back MDT/MGS to itself (testing)
521 Frequency : at statup only
523 Description: mds_lov_synchronize race leads to various problems
524 Details : simultaneous MDT->OST connections at startup can cause the
525 sync to abort, leaving the OSC in a bad state.
527 Severity : enhancement
529 Description: add optional extra BUILD_VERSION info
530 Details : add a new environment variable (namely LUSTRE_VERS) which allows
531 to override the lustre version.
534 Frequency : 2.6.18 servers only
536 Description: ll_kern_mount() doesn't release the module reference
537 Details : The ldiskfs module reference count never drops down to 0
538 because ll_kern_mount() doesn't release the module reference.
543 Description: server LBUG when using old ost_num_threads parameter
544 Details : Accept the old ost_num_threads parameter but warn that it
545 is deprecated, and fix an off-by-one error that caused an LBUG.
550 Description: Transient SCSI error results in persistent IO issue
551 Details : iobuf->dr_error is not reinitialized to 0 between two
555 Frequency : sometimes when underlying device returns I/O errors
557 Description: OSTs not going read-only during write failures
558 Details : OSTs are not remounted read-only when the journal commit threads
559 get I/O errors because fsfilt_ext3 calls journal_start/stop()
560 instead of the ext3 wrappers.
564 Description: poor connect scaling with increasing client count
565 Details : Don't run filter_grant_sanity_check for more than 100 exports
566 to improve scaling for large numbers of clients.
569 Frequency : SLES10 only
571 Description: sanity-quota.sh quotacheck failed: rc = -22
572 Details : Quotas cannot be enabled on SLES10.
575 Frequency : liblustre clients only
577 Description: getdirentries does not give error when run on compute nodes
578 Details : getdirentries does not fail when the size specified as an argument
579 is too small to contain at least one entry
581 Severity : enhancement
583 Description: Add LNET router traceability for debug purposes
584 Details : If a checksum failure occurs with a router as part of the
585 IO path, the NID of the last router that forwarded the bulk data
586 is printed so it can be identified.
591 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
592 Details : Due to a race condition, liblustre clients were occasionally
595 Severity : enhancement
597 Description: lfs setstripe use optional parameters instead of postional
600 Severity : enhancement
602 Description: Nanosecond timestamp support for ldiskfs
603 Details : The on-disk ldiskfs filesystem has added support for nanosecond
604 resolution timestamps. There is not yet support for this at
605 the Lustre filesystem level.
608 Frequency : during server recovery
610 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
611 Details : request with rq_no_resend flag not awake l_wait_event if they get a
615 Frequency : nfs export on patchless client
617 Description: connectathon hang when test nfs export over patchless client
618 Details : Disconnected dentry cannot be found with lookup, so we do not need
619 to unhash it or make it invalid
622 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
623 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
624 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
629 Description: open req refcounting wrong on reconnect
630 Details : If reconnect happened between getting open reply from server and
631 call to mdc_set_replay_data in ll_file_open, we will schedule
632 replay for unreferenced request that we are about to free.
633 Subsequent close will crash in variety of ways.
634 Check that request is still eligible for replay in
635 mdc_set_replay_data().
640 Description: disable writes to filesystem when reading health_check file
641 Details : the default for reading the health_check proc file has changed
642 to NOT do a journal transaction and write to disk, because this
643 can cause reads of the /proc file to hang and block HA state
644 checking on a healthy but otherwise heavily loaded system. It
645 is possible to return to the previous behaviour during configure
646 with --enable-health-write.
648 Severity : enhancement
650 Description: 64-bit inode version
651 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
652 to the inode. This will be required for version-based recovery.
657 Description: MDS fails to start if a duplicate client export is detected
658 Details : in some rare cases it was possible for a client to connect to
659 an MDS multiple times. Upon recovery the MDS would detect this
660 and fail during startup. Handle this more gracefully.
662 Severity : enhancement
664 Description: Add -o localflock option to simulate old noflock
666 Details : This will achieve local-only flock/fcntl locks
672 Description: log_commit_thread vs filter_destroy race leads to crash
673 Details : Take import reference before releasing llog record semaphore
678 Description: Wrong request locking in request set processing
679 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
680 imp_delayed_list, in this place should be used imp_lock.
683 Frequency : when reconnection
685 Description: Grant Leak when osc reconnect to OST
686 Details : When osc reconnect ost, OST(filter) should check whether it
687 should grant more space to client by comparing fed_grant and
688 cl_avail_grant, and return the granted space to client instead
689 of "new granted" space, because client will call osc_init_grant
690 to update the client grant space info.
693 Frequency : when client reconnect to OST
695 Description: Grant Leak when osc do resend and replay bulk write
696 Details : When osc reconnect to OST, OST(filter)should clear grant info of
697 bulk write request, because the grant info while be sync between
698 OSC and OST when reconnect, and we should ignore the grant info
699 these of resend/replay write req.
704 Description: Grant space more than avaiable left space sometimes.
705 Details : When then OST is about to be full, if two bulk writing from
706 different clients came to OST. Accord the avaliable space of the
707 OST, the first req should be permitted, and the second one
708 should be denied by ENOSPC. But if the seconde arrived before
709 the first one is commited. The OST might wrongly permit second
710 writing, which will cause grant space > avaiable space.
713 Frequency : when client is evicted
715 Description: Grant might be wrongly erased when osc is evicted by OST
716 Details : when the import is evicted by server, it will fork another
717 thread ptlrpc_invalidate_import_thread to invalidate the
718 import, where the grant will be set to 0. While the original
719 thread will update the grant it got when connecting. So if
720 the former happened latter, the grant will be wrongly errased
721 because of this race.
726 Description: Checking Stale with correct fid
727 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
728 to check whether it is stale, because sometimes, we want the
729 enqueue happened anyway, and op_data.fid2 will not be initialized.
731 Severity : enhancement
733 Description: update patchless client
734 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
737 Frequency : only with 2.4 kernel
739 Description: random memory corruption
740 Details : size of struct ll_inode_info is to big for union inode.u and this
741 can be cause of random memory corruption.
746 Description: Memory leak in recovery
747 Details : Lov_mds_md was not free in an error handler in mds_create_object.
748 It should also check obd_fail before fsfilt_start, otherwise if
749 fsfilt_start return -EROFS,(failover mds during mds recovery).
750 then the req will return with repmsg->transno = 0 and rc = EROFS.
751 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
752 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
753 be freed no matter whether fsfilt_commit success or not.
756 Frequency : only with huge count clients
758 Description: Prevents from taking the superblock lock in llap_from_page for
760 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
761 for llap_from_page call in ll_removepage prevents from taking the
762 superblock lock for a soon died page.
767 Description: Not check open intent error before release open handle
768 Details : in some rare cases, the open intent error is not checked before
769 release open handle, which may cause
770 ASSERTION(open_req->rq_transno != 0), because it tries to release
771 the failed open handle.
776 Description: Set cat log bitmap only after create log success.
777 Details : in some rare cases, the cat log bitmap is set too early. and it
778 should be set only after create log success.
782 Description: Accessing a block bevice can re-enable I/O when Lustre is
783 tearing down a device.
784 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
789 Description: service threads may hog cpus when there are a lot of requests
791 Details : Insert cond_resched to give other threads a chance to use some of
797 Description: the cat log was not initialized in recovery
798 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
799 unlink log on mds will not be initialized until mds post
800 recovery. And also in mds post recovery, the unlink log will
801 initialization will be done asynchronausly, so there will be race
802 between add unlink log and unlink log initialization.
806 Description: brw_stats were being printed incorrectly
807 Details : brw_stats were being printed as log2 but all of them were not
808 recorded as log2. Also remove some code duplication arising from
809 filter_tally_{read,write}.
813 Frequency : rare, only in recovery.
814 Description: ASSERTION(req->rq_type != LI_POISON) failed
815 Details : imp_lock should be held while iterating over imp_sending_list for
816 prevent destroy request after get timeout in ptlrpc_queue_wait.
820 Description: replay-single.sh test 52 fails
821 Details : A lock's skiplist need to be cleanup when it being unlinked
822 from its resource list.
826 Description: Short directio read returns full requested size rather than
828 Details : Direct I/O operations should return actual amount of bytes
829 transferred rather than requested size.
831 Severity : enhancement
833 Description: metadata RPC reduction (e.g. for rm performance)
834 Details : decrease the amount of synchronous RPC between clients and servers
835 by canceling conflicing lock before the operation on the client side
836 and packing thier handles into the main operation RPC to server.
838 Severity : enhancement
840 Description: Async OSC create to avoid the blocking unnecessarily.
841 Details : If a OST has no remain object, system will block on the creating
842 when need to create a new object on this OST. Now, ways use
843 pre-created objects when available, instead of blocking on an
844 empty osc while others are not empty. If we must block, we block
845 for the shortest possible period of time.
847 --------------------------------------------------------------------------------
849 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
854 Frequency : on some architectures
856 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
857 Details : Uninitialized flags sometimes cause configuration commands to
861 Frequency : patchless clients only
863 Description: missing __iget() symbol export
864 Details : The __iget() symbol export is missing. To avoid the need for
865 this on patchless clients the deathrow inode reaper is turned
866 off, and we depend on the VM to clean up old inodes. This
867 dependency was during via the fix for bug 12181.
869 --------------------------------------------------------------------------------
871 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
873 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
874 INTEROPERATE with older versions automatically. In many cases a
875 special upgrade step is needed. Please read the
876 user documentation before upgrading any part of a live system.
877 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
878 version will not interoperate with 1.6 betas before beta5 (1.5.95).
879 * WARNING: Lustre configuration and startup changes are required with
880 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
886 Severity : enhancement
888 Description: MountConf
889 Details : Lustre configuration is now managed via mkfs and mount
890 commands instead of lmc and lconf. New obd types (MGS, MGC)
891 are added for dynamic configuration management. See
892 https://mail.clusterfs.com/wikis/lustre/MountConf for
895 Severity : enhancement
897 Description: dynamic OST addition
898 Details : OSTs can now be added to a live filesystem
900 Severity : enhancement
902 Description: startup order invariance
903 Details : MDTs and OSTs can be started in any order. Clients only
904 require the MDT to complete startup.
906 Severity : enhancement
908 Description: parallel, asynchronous orphan cleanup
909 Details : orphan cleanup is now performed in separate threads for each
910 OST, allowing parallel non-blocking operation.
912 Severity : enhancement
914 Description: optimized stripe assignment
915 Details : stripe assignments are now made based on ost space available,
916 ost previous usage, and OSS previous usage, in order to try
917 to optimize storage space and networking resources.
919 Severity : enhancement
921 Description: Permanently set tunables
922 Details : All writable /proc/fs/lustre tunables can now be permanently
923 set on a per-server basis, at mkfs time or on a live
926 Severity : enhancement
928 Description: Lustre message v2
929 Details : Add lustre message format v2.
931 Severity : enhancement
933 Description: client OST exclusion list
934 Details : Clients can be started with a list of OSTs that should be
935 declared "inactive" for known non-responsive OSTs.
939 Description: SPEC SFS validation failure on NFS v2 over lustre.
940 Details : Changes the blocksize for regular files to be 2x RPC size,
941 and not depend on stripe size.
943 Severity : enhancement
945 Description: Multiple MD RPCs in flight.
946 Details : Further unserialise some read-only MDS RPCs - learn about intents.
947 To avoid overly-overloading MDS, introduce a limit on number of
948 MDS RPCs in flight for a single client and add /proc controls
949 to adjust this limit.
951 Severity : enhancement
953 Description: client read/write statistics
954 Details : Add client read/write call usage stats for performance
955 analysis of user processes.
956 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
957 file access. extents_stats shows chunk size distribution.
958 extents_stats_per_process show chunk size distribution per
961 Severity : enhancement
963 Description: mds statistics
964 Details : Add detailed mds operations statistics in
965 /proc/fs/lustre/mds/*/stats.
969 Description: Failure of copying files with lustre special EAs.
970 Details : Client side always return success for setxattr call for lustre
971 special xattr (currently only "trusted.lov").
976 Description: Refcount LNET uuids
977 Details : The global LNET uuid list grew linearly with every startup;
978 refcount repeated list entries instead of always adding to
981 Severity : enhancement
983 Description: Dynamic service threads
984 Details : Within a small range, start extra service threads
985 automatically when the request queue builds up.
988 Frequency : mixed-endian client/server environments
990 Description: mixed-endian crashes
991 Details : The new msg_v2 system had some failures in mixed-endian
994 Severity : enhancement
996 Description: Easy OST removal
997 Details : OSTs can be permanently deactivated with e.g. 'lctl
998 conf_param lustre-OST0001.osc.active=0'
1000 Severity : enhancement
1002 Description: MGS proc entries
1003 Details : Added basic proc entries for the MGS showing what filesystems
1006 Severity : enhancement
1008 Description: provide MGS failover
1009 Details : Added config lock reacquisition after MGS server failover.
1011 Severity : enhancement
1013 Description: add Linux 2.4 support
1014 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1018 Description: a large application tries to do I/O to the same resource and dies
1019 in the middle of it.
1020 Details : Check the req->rq_arrival time after the call to
1021 ost_brw_lock_get(), but before we do anything about
1022 processing it & sending the BULK transfer request. This
1023 should help move old stale pending locks off the queue as
1024 quickly as obd_timeout.
1027 Frequency : when an incorrect nid is specified during startup
1029 Description: ptlrpc connect to non-existant node causes kernel crash
1030 Details : LNET can't be re-entered from an event callback, which
1031 happened when we expire a message after the export has been
1032 cleaned up. Instead, hand the zombie cleanup off to another
1035 Severity : enhancement
1037 Description: plain/inodebits lock performance improvement
1038 Details : Grouping plain/inodebits in granted list by their request modes
1039 and bits policy, thus improving the performance of search through
1043 Frequency : only if OST filesystem is corrupted
1045 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1046 Details : for a short time RPCs with bulk IO are in the replay list,
1047 but replay of bulk IOs is unimplemented. If the OST filesystem
1048 is corrupted due to disk cache incoherency and then replay is
1049 started it is possible to trip an assertion. Avoid putting
1050 committed RPCs into the replay list at all to avoid this issue.
1053 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1055 Description: System hang on startup
1056 Details : This bug allowed the liblustre (e.g. catamount) client to
1057 return to the app before handling all startup RPCs. This
1058 could leave the node unresponsive to lustre network traffic
1059 and manifested as a server ptllnd timeout.
1061 Severity : enhancement
1063 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1064 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1065 boolean to control whether to print peer debug info when a
1066 client's RPC times out.
1069 Frequency : only for kernels with patches from Lustre below 1.4.3
1071 Description: Remove old rdonly API
1072 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1075 Frequency : only for devices with external journals
1077 Description: Set external device read-only also
1078 Details : During a commanded failover stop, we set the disk device
1079 read-only while the server shuts down. We now also set any
1080 external journal device read-only at the same time.
1083 Frequency : when upgrading from 1.4 while trying to change parameters
1085 Description: The wrong (new) MDC name was used when setting parameters for
1086 upgraded MDT's. Also allows changing of OSC (and MDC)
1087 parameters if --writeconf is specified at tunefs upgrade time.
1090 Frequency : when setting specific ost indicies
1092 Description: QOS code breaks on skipped indicies
1093 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1094 created with --index need not be sequential.
1096 Severity : enhancement
1098 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1099 Details : The uninit_groups feature works in conjunction with the kernel
1100 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1101 up the pass1 processing of e2fsck. This is a read-only feature
1102 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1103 mount filesystems that have had this feature enabled.
1105 Severity : enhancement
1107 Description: Improve multi-block allocation algorithm to avoid fragmentation
1108 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1109 allocation locality and avoid filesystem fragmentation.
1111 ------------------------------------------------------------------------------
1113 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1115 * Support for kernels:
1116 2.4.21-47.0.1.EL (RHEL 3)
1117 2.6.5-7.283 (SLES 9)
1118 2.6.9-42.0.10.EL (RHEL 4)
1119 2.6.12.6 vanilla (kernel.org)
1120 2.6.16.27-0.9 (SLES 10)
1121 * Recommended e2fsprogs version: 1.39.cfs5
1123 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1127 Frequency : occasional, depends on client load and configuration
1128 Bugzilla : 12181, 12203
1129 Description: data loss for recently-modified files
1131 Details : In some cases it is possible that recently written or created
1132 files may not be written to disk in a timely manner (this should
1133 normally be within 30s unless client IO load is very high).
1134 The problem appears as zero-length files or files that are a
1135 multiple of 1MB in size after a client crash or client eviction
1136 that are missing data at the end of the file.
1138 This problem is more likely to be hit on clients where files are
1139 repeatedly created and unlinked in the same directory, clients
1140 have a large amount of RAM, have many CPUs, the filesystem has
1141 many OSTs, the clients are rebooted frequently, and/or the files
1142 are not accessed by other nodes after being written.
1144 The presence of the problem can be detected by looking at
1145 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1146 smaller than the second (nr_unused) then dirty files will not
1147 be flushed automatically to disk. "sync; sleep 10" should be
1148 run several times on the node before unmounting it to update
1149 Lustre (this is also safe to run on nodes without this problem).
1151 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1152 that can cause this same problem, so customers using that kernel
1153 also need to update the kernel in addition to Lustre. In order
1154 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1156 It is normal that files written just before a client crash (less
1157 than 30s) may not yet have been flushed to disk, even for local
1161 Frequency : frequent on thin XT3 nodes
1163 Description: UUID collision on thin XT3 Linux nodes
1164 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1165 randomly, since we relied on an insufficiently-seeded PRNG.
1170 Description: OSS hangs after "All ost request buffers busy"
1171 Details : A deadlock between quota and journal operations caused OSS
1172 hangs after printing "All ost request buffers busy."
1175 Frequency : always on liblustre builds
1177 Description: Cleanup compiler warnings on liblustre
1180 Frequency : always on liblustre builds on XT3
1182 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1186 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1187 > 32000 subdirectories
1188 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1189 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1190 subdirectory count crosses 32000. This will aid e2fsck to
1191 correctly handle more than 32000 subdirectories.
1194 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1196 Description: System hang on startup
1197 Details : This bug allowed the liblustre (e.g. catamount) client to
1198 return to the app before handling all startup RPCs. This
1199 could leave the node unresponsive to lustre network traffic
1200 and manifested as a server ptllnd timeout.
1202 Severity : enhancement
1204 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1205 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1206 boolean to control whether to print peer debug info when a
1207 client's RPC times out.
1212 Description: make O_SYNC working on 2.6 kernels
1213 Details : 2.6 kernels use different method for mark pages for write,
1214 so need add a code to lustre for O_SYNC work.
1219 Description: Failure to close file and release space on NFS
1220 Details : Put inode details into lock acquired in ll_intent_file_open.
1221 Use mdc_intent_lock in ll_intent_open to properly
1222 detect all kind of errors unhandled by mdc_enqueue.
1227 Description: proc file read during shutdown sometimes raced obd removal,
1229 Details : Add lock to prevent obd access after proc file removal.
1232 Frequency : Only for files larger than 4GB on 32-bit clients.
1234 Description: improperly doing page alignment of locks
1235 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1236 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1241 Description: RPCs being resent when they shouldn't be
1242 Details : Some RPCs that should not be resent are being resent. This
1243 can cause inconsistencies in the RPC state machine. Do not
1244 resend such requests.
1247 Frequency : rare, only with NFS export
1249 Description: Crash on NFS re-export node
1250 Details : under very unusual load conditions an assertion is hit in
1251 ll_intent_file_open()
1254 Frequency : only if OST filesystem is corrupted
1256 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1257 Details : for a short time RPCs with bulk IO are in the replay list,
1258 but replay of bulk IOs is unimplemented. If the OST filesystem
1259 is corrupted due to disk cache incoherency and then replay is
1260 started it is possible to trip an assertion. Avoid putting
1261 committed RPCs into the replay list at all to avoid this issue.
1266 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1267 Details : Large single O_DIRECT read and write calls can fail to allocate
1268 a sufficiently large buffer to process the request. In case of
1269 allocation failure the allocation is retried with a smaller
1270 buffer and broken into smaller requests.
1272 Severity : enhancement
1274 Description: Add -o localflock option to simulate old noflock behaviour.
1275 Details : This will achieve local-only flock/fcntl locks coherentness.
1280 Description: versioning check is incomplete
1281 Details : Checking the version difference of client vs. server, report
1282 error if the gap is too big.
1287 Description: add support PG_writeback bit
1288 Details : add support for PG_writeback bit for Lustre, for more carefull
1289 work with page cache in 2.6 kernel. This also fix some deadlocks
1290 and remove hack for work O_SYNC with 2.6 kernel.
1292 Severity : enhancement
1294 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1295 Details : The uninit_groups feature works in conjunction with the kernel
1296 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1297 up the pass1 processing of e2fsck. This is a read-only feature
1298 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1299 mount filesystems that have had this feature enabled.
1301 Severity : enhancement
1303 Description: Improve multi-block allocation algorithm to avoid fragmentation
1304 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1305 allocation locality and avoid filesystem fragmentation.
1307 ------------------------------------------------------------------------------
1309 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1311 * Support for kernels:
1312 2.6.9-42.0.3.EL (RHEL 4)
1313 2.6.5-7.276 (SLES 9)
1314 2.4.21-47.0.1.EL (RHEL 3)
1315 2.6.12.6 vanilla (kernel.org)
1316 2.6.16.21-0.8 (SLES10)
1317 * Recommended e2fsprogs version: 1.39.cfs2-0
1319 * The backwards-compatible /proc/sys/portals symlink has been removed
1320 in this release. Before upgrading, please ensure that you change
1321 any configuration scripts or /etc/sysctl.conf files that access
1322 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1323 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1324 in advance of the upgrade on any system running Lustre 1.4.6 or
1325 newer, since /proc/sys/lnet was added in that version.
1326 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1330 Frequency : only when quota is used
1332 Description: avoid scanning export list for quota master
1333 Details : Change the algorithms to avoid scanning export list in order
1334 to improve the efficiency.
1337 Frequency : MDS failover only, very rarely
1339 Description: "went back in time" messages on mds failover
1340 Details : The greatest transno may be lost when the current operation
1341 finishes with an error (transno==0) and the client's last_rcvd
1342 record is over-written. Save the greatest transno in the
1343 mds_last_transno for this case.
1346 Frequency : always for specific kernels and striping counts
1348 Description: client may get "Matching packet too big" without ACL support
1349 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1350 when trying to access files in certain configurations. The
1351 clients should in fact be denied when mounting because they do
1352 not understand ACLs.
1355 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1357 Description: many clients connecting with IO in progress causes connect timeouts
1358 Details : Avoid synchronous journal commits to avoid delays caused by many
1359 clients connecting/disconnecting when bulk IO is in progress.
1360 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1361 OST_IO_PORTAL to avoid delays behind potentially many pending
1365 Frequency : occasionally with multiple writers to a single file
1367 Description: shared writes to file may result in wrong size reported by stat()
1368 Details : Allow growing of kms when extent lock is cancelled
1371 Frequency : always with random mmap IO to multi-striped file
1373 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1374 Details : Only if the hole is at the end of OST object so that kms is too
1375 small. Fix is to increase kms accordingly in ll_nopage.
1378 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1380 Description: writes to a missing object would leak memory on the OST
1381 Details : If there is an inconsistency between the MDS and OST filesystems,
1382 such that the MDS references an object that doesn't exist, writes
1383 to that object will leak memory due to incorrect cleanup in the
1384 error handling path, eventually running out of memory on the OST.
1389 Description: Creating too long symlink causes lustre errors
1390 Details : Check symlink and name lengths before sending requests to MDS.
1393 Frequency : only if flock is enabled (not on by default)
1395 Description: posix locks not released on fd closure on 2.6.9+
1396 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1397 kernels, this caused such locks not to be released on fd close and
1398 then assertions on fs unmount about still used locks.
1401 Frequency : MDS failover only, very rarely
1403 Description: clients may get ASSERTION(granted_lock != NULL)
1404 Details : When request was taking a long time, and a client was resending
1405 a getattr by name lock request. The were multiple lock requests
1406 with the same client lock handle and
1407 mds_getattr_name->fixup_handle_for_resent_request found one of the
1408 lock handles but later failed with ASSERTION(granted_lock != NULL).
1413 Description: handle->h_buffer_credits > 0, assertion failure
1414 Details : h_buffer_credits is zero after truncate, causing assertion
1415 failure. This patch extends the transaction or creates a new
1419 Frequency : NFS re-export or patchless client
1420 Bugzilla : 11179, 10796
1421 Description: Crash on NFS re-export node (__d_move)
1422 Details : We do not want to hash the dentry if we don't have a lock.
1423 But if this dentry is later used in d_move, we'd hit uninitialised
1424 list head d_hash, so we just do this to init d_hash field but
1425 leave dentry unhashed.
1428 Frequency : NFS re-export or patchless client
1430 Description: NFS exports has problem with symbolic link
1431 Details : lustre client didn't properly install dentry when re-exported
1432 to NFS or running patchless client.
1435 Frequency : NFS re-export or patchless client
1437 Description: Various nfs/patchless fixes.
1438 Details : fixes reuse disconected alias for lookup process - this fixes
1439 warning "find_exported_dentry: npd != pd",
1440 fix permission error with open files at nfs.
1441 fix apply umask when do revalidate.
1444 Frequency : occasional
1446 Description: Crash on NFS re-export node
1447 Details : calling clear_page() on the wrong pointer triggered oops in
1448 generic_mapping_read().
1451 Frequency : rarely, using O_DIRECT IO
1453 Description: unaligned directio crashes client with LASSERT
1454 Details : check for unaligned buffers before trying any requests.
1457 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1459 Description: stale data returned from RAID cache
1460 Details : If only a small amount of IO is done to the RAID device before
1461 reading it again it is possible to get stale data from the RAID
1462 cache instead of reading it from disk.
1465 Frequency : always for sles10 kernel
1467 Description: sles10 support
1468 Details : ll_follow_link: compile fixes and using of nd_set_link
1469 under newer kernels.
1472 Frequency : depends on arch, kernel and compiler version, always on sles10
1475 Description: recursive or deep enough symlinks cause stack overflow
1476 Details : getting rid of large stack-allocated variable in
1480 Frequency : depends on hardware
1482 Description: lustre write performance loss in the SLES10 kernel
1483 Details : the performance loss is caused by using of write barriers in the
1484 ext3 code. The SLES10 kernel turns barrier support on by
1485 default. The fix is to undo that change for ldiskfs.
1487 ------------------------------------------------------------------------------
1489 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1491 * Support for kernels:
1492 2.6.9-42.0.3EL (RHEL 4)
1493 2.6.5-7.276 (SLES 9)
1494 2.4.21-47.0.1.EL (RHEL 3)
1495 2.6.12.6 vanilla (kernel.org)
1499 Frequency : quota enabled and large files being deleted
1501 Description: releasing more than 4GB of quota at once hangs OST
1502 Details : If a user deletes more than 4GB of files on a single OST it
1503 will cause the OST to spin in an infinite loop. Release
1504 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1509 Description: statfs data retrieved from /proc may be stale or zero
1510 Details : When reading per-device statfs data from /proc, in the
1511 {kbytes,files}_{total,free,avail} files, it may appear
1512 as zero or be out of date.
1515 Frequency : systems with MD RAID1 external journal devices
1517 Description: lconf's call to blkid is confused by RAID1 journal devices
1518 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1519 of returning all block devices that match the journal UUID.
1522 Frequency : always, for aggregate stripe size over 4GB
1524 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1525 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1526 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1529 Frequency : always if "lfs find" used on a local file/directory
1531 Description: "lfs find" segfaults if used on a local file/directory
1532 Details : The case where a directory component was not specified wasn't
1533 handled correctly. Handle this properly.
1536 Frequency : always on ppc64
1538 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1539 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1540 perform endianness conversion.
1543 Frequency : rarely (truncate to non-zero file size after write under load)
1544 Bugzilla : 10730, 10687
1545 Description: Files padded with zeros to next 4K multiple
1546 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1547 it is possible that files that are truncated to a non-zero size
1548 immediately after being written are filled with zero bytes beyond
1549 the truncated size. No file data is lost.
1551 Severity : enhancement
1553 Description: Allow recovery/failover for liblustre clients.
1554 Details : liblustre clients were unaware of failover configurations until
1557 Severity : enhancement
1559 Description: user file locks should fail when not mounting with flock option
1560 Details : Set up an error-returning stub in ll_file_operations.lock field
1561 to prevent incorrect behaviour when client is mounted without
1562 flock option. Also, set up properly f_op->flock field for
1566 Frequency : always on ia64
1568 Description: "lfs df" loops on printing out MDS statfs information
1569 Details : The obd_ioctl_data was not initialized and in some systems
1570 this caused a failure during the ioctl that did not return
1571 an error. Initialize the struct and return an error on failure.
1574 Frequency : SLES 9 only
1576 Description: Error of copying files with lustre special EAs as root
1577 Details : Client side always return success for setxattr call for lustre
1578 special xattr (currently only "trusted.lov").
1581 Frequency : rarely on clusters with both ia64+i386 clients
1583 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1584 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1585 clients are doing concurrent writes to the same file it is
1586 possible that the ia64 clients may LASSERT because the OST
1587 extent locks are not PAGE_SIZE aligned. Ensure that grown
1588 locks are always aligned on the request boundary.
1591 Frequency : specific use, occasional
1593 Description: Overwriting in use executable truncates on-disk binary image
1594 Details : If one node attempts to overwrite an executable in use by
1595 another node, we now correctly return ETXTBSY instead of
1596 truncating the file.
1598 Severity : enhancement
1600 Description: Async OSC create to avoid the blocking unnecessarily.
1601 Details : If a OST has no remain object, system will block on the creating
1602 when need to create a new object on this OST. Now, ways use
1603 pre-created objects when available, instead of blocking on an
1604 empty osc while others are not empty. If we must block, we block
1605 for the shortest possible period of time.
1610 Description: chmod on Lustre root is propagated to other clients
1611 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1612 invalid by the follow_mount time.
1615 Frequency : liblustre clients only
1617 Description: Race in 'instant cancel' lock handling could lead to such locks
1618 never to be granted in case of SMP MDS
1619 Details : Do not destroy not yet granted but cbpending locks in
1623 Frequency : replay/resend of open
1625 Description: non null lock assetion failure in mds_intent_policy
1626 Details : Trying to replay/resend lockless open requests resulted in
1627 mds_open() returning 0 with no lock. Now it sets a flag if
1628 a lock is going to be returned.
1630 Severity : enhancement
1632 Description: Checksum enhancements
1633 Details : New checksum enhancements allow for resending RPCs that failed
1636 Severity : enhancement
1638 Description: Tunables on number of dirty pages in cacche
1639 Details : Allow to set limit on number of dirty pages cached.
1644 Description: client crash on unmount - lock still has references
1645 Details : In some error handling cases it was possible to leak a lock
1646 reference on a client while accessing a file. This was not
1647 harmful to the client during operation, but would cause the
1648 client to crash when the filesystem is unmounted.
1651 Frequency : specific case, rare
1653 Description: ETXTBSY on mds though file not in use by client
1654 Details : ETXTBSY is no longer incorrectly returned when attempting to
1655 chmod or chown a directory that the user previously tried to
1656 execute or a currently-executing binary.
1659 Frequency : extremely rare except on liblustre-based clients
1661 Description: Lustre space not freed when files are deleted
1662 Details : Clean up open-unlinked files after client eviction. Previously
1663 the unlink was skipped and the files remained as orphans.
1668 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1669 Details : In some cases it was possible to send a blocking callback to
1670 a client doing a glimpse, even though that client didn't get
1671 a lock granted. When the glimpse lock is cancelled on the OST
1672 the freed lock is left on the waiting list and corrupted the list.
1675 Frequency : all core dumps
1677 Description: Broke core dumps to lustre
1678 Details : Negative dentry may be unhashed if parent does not have UPDATE
1679 lock, but some callers, e.g. do_coredump, expect dentry to be
1680 hashed after successful create, hash it in ll_create_it.
1682 ------------------------------------------------------------------------------
1684 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1686 * Support for kernels:
1687 2.6.9-42.0.2.EL (RHEL 4)
1688 2.6.5-7.276 (SLES 9)
1689 2.4.21-40.EL (RHEL 3)
1690 2.6.12.6 vanilla (kernel.org)
1694 Frequency : always on RHEL 3
1696 Description: Number of open files grows over time
1697 Details : The number of open files grows over time, whether or not
1698 Lustre is started. This was due to a filp leak introduced
1699 by one of our kernel patches.
1701 ------------------------------------------------------------------------------
1703 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1705 * Support for kernels:
1706 2.6.9-42.EL (RHEL 4)
1707 2.6.5-7.276 (SLES 9)
1708 2.4.21-40.EL (RHEL 3)
1709 2.6.12.6 vanilla (kernel.org)
1714 Bugzilla : 5719, 9635, 9792, 9684
1715 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1716 Details : If a server is under heavy load and cannot reply to new
1717 connection requests before the client resends the (re)connect,
1718 the connection handling code can behave badly if two service
1719 threads are concurrently handing separate (re)connections from
1720 the same client. Add better locking to the connection handling
1721 code, and ensure that only a single connection will be processed
1722 for a given client UUID, even if the lock is dropped.
1724 Severity : enhancement
1726 Description: add TCP zero-copy support to kernel
1727 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1728 sends if the hardware supports scatter-gather and checksumming.
1729 This allows socklnd to do client-write and server-read more
1730 efficiently and reduce CPU utilization from skbuf copying.
1733 Frequency : only if NFS exporting from client
1735 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1736 Details : If setting attributes on a file created under NFS that had
1737 never been opened it would be possible to oops the client
1738 if the file had no objects.
1741 Frequency : always for liblustre
1743 Description: liblustre client does MDS+OSTs setattr RPC for each write
1744 Details : When doing a write from a liblustre client, the client
1745 incorrectly issued an RPC to the MDS and each OST the file was
1746 striped over in order to update the timestamps. When writing
1747 with small chunks and many clients this could overwhelm the MDS
1748 with RPCs. In all cases it would slow down the write because
1749 these RPCs are unnecessary.
1751 Severity : enhancement
1753 Description: allow number of MDS service threads to be changed at module load
1754 Details : It is now possible to change the number of MDS service threads
1755 running. Adding "options mds mds_num_threads={N}" to the MDS's
1756 /etc/modprobe.conf will set the number of threads for the next
1757 time Lustre is restarted (assuming the "mds" module is also
1758 reloaded at that time). The default number of threads will
1759 stay the same, 32 for most systems.
1764 Description: OST crash if filesystem is unformatted or corrupt
1765 Details : If an OST is started on a device that has never been formatted
1766 or if the filesystem is corrupt and cannot even mount then the
1767 error handling cleanup routines would dereference a NULL pointer.
1772 Description: NULL pointer deref in llap_from_page.
1773 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1774 incorrect) mapping if the page was truncated/reclaimed while it was
1775 searched for. Check for this condition and skip such pages when
1776 doing readahead. Introduce extra check to llap_from_page() to
1777 verify page->mapping->host is non-NULL (so page is not anonymous).
1780 Frequency : Sometimes when using sys_sendfile
1782 Description: "page not covered by a lock" warnings from ll_readpage
1783 Details : sendfile called ll_readpage without right page locks present.
1784 Now we introduced ll_file_sendfile that does necessary locking
1785 around call to generic_file_sendfile() much like we do in
1789 Frequency : with certain MDS communication failures at client mount time
1791 Description: NULL pointer deref after failed client mount
1792 Details : a client connection request may delayed by the network layer
1793 and not be sent until after the PTLRPC layer has timed out the
1794 request. If the client fails the mount immediately it will try
1795 to clean up before the network times out the request. Add a
1796 reference from the request import to the obd device and delay
1797 the cleanup until the network drops the request.
1800 Frequency : occasionally during client (re)connect
1802 Description: assertion failure during client (re)connect
1803 Details : processing a client connection request may be delayed by the
1804 client or server longer than the client connect timeout. This
1805 causes the client to resend the connection request. If the
1806 original connection request is replied in this interval, the
1807 client may trip an assertion failure in ptlrpc_connect_interpret()
1808 which thought it would be the only running connect process.
1811 Frequency : only with obd_echo servers and clients that are rebooted
1813 Description: kernel BUG accessing uninitialized data structure
1814 Details : When running an obd_echo server it did not start the ping_evictor
1815 thread, and when a client was evicted an uninitialized data
1816 structure was accessed. Start the ping_evictor in the RPC
1817 service startup instead of the OBD startup.
1819 Severity : enhancement
1820 Bugzilla : 10193 (patchless)
1821 Description: Remove dependency on various unexported kernel interfaces.
1822 Details : No longer need reparent_to_init, exit_mm, exit_files,
1823 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1826 Frequency : rare (only users of deprecated and unsupported LDAP config)
1828 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1829 Details : LDAP apparently contains 'lustreName' attributes instead of
1830 'name'. A simple remapping of the name is sufficient.
1833 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1835 Description: waiting_locks_callback trips kernel BUG if client is evicted
1836 Details : Running with the dump_on_timeout debug flag turned on makes
1837 it possible that the waiting_locks_callback() can try to dump
1838 the Lustre kernel debug logs from an interrupt handler. Defer
1839 this log dumping to the expired_lock_main() thread.
1841 Severity : enhancement
1843 Description: Support NFS exporting on 2.6 kernels.
1844 Details : Implement non-rawops metadata methods for NFS server to use without
1845 changing NFS server code.
1848 Frequency : very rare (synthetic metadata workload only)
1850 Description: two racing renames might cause an MDS thread to deadlock
1851 Details : Running the "racer" program may cause one MDS thread to rename
1852 a file from being the source of a rename to being the target of
1853 a rename at exactly the same time that another thread is doing
1854 so, and the second thread has already enqueued these locks after
1855 doing a lookup of the target and is trying to relock them in
1856 order. Ensure that we don't try to re-lock the same resource.
1859 Frequency : only very large systems with liblustre clients
1861 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1862 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1863 all OSTs in parallel. This allows the eviction of stale liblustre
1864 clients to proceed much faster than if they were done in series,
1865 and also offers similar improvements for other set_info RPCs.
1870 Description: excessive CPU usage during initial read phase on client
1871 Details : During the initial read phase on a client, it would agressively
1872 retry readahead on the file, consuming too much CPU and impacting
1873 performance (since 1.4.5.8). Improve the readahead algorithm
1874 to avoid this, and also improve some other common cases (read
1875 of small files in particular, where "small" is files smaller than
1876 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1881 Description: MDS crash when receiving packet with unknown intent.
1882 Details : Do not LBUG in unknown intent case, just return -EFAULT
1884 Severity : enhancement
1885 Bugzilla : 9293, 9385
1886 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1887 Details : Do not serialize getattr (non-intent version) and statfs.
1890 Frequency : occasional, when OST network is overloaded/intermittent
1892 Description: client evicted by OST after bulk IO timeout
1893 Details : If a client sends a bulk IO request (read or write) the OST
1894 may evict the client if it is unresposive to its data GET/PUT
1895 request. This is incorrect if the network is overloaded (takes
1896 too long to transfer the RPC data) or dropped the OST GET/PUT
1897 request. There is no need to evict the client at all, since
1898 the pinger and/or lock callbacks will handle this, and the
1899 client can restart the bulk request.
1902 Frequency : Always when mmapping file with no objects
1904 Description: client crashes when mmapping file with no objects
1905 Details : Check that we actually have objects in a file before doing any
1906 operations on objects in ll_vm_open, ll_vm_close and
1912 Description: Request leak when working with deleted CWD
1913 Details : Introduce advanced request refcount tracking for requests
1914 referenced from lustre intent.
1916 Severity : Enhancement
1918 Description: Cache open file handles on client.
1919 Details : MDS now will return special lock along with openhandle, if
1920 requested and client is allowed to hold openhandle, even if unused,
1921 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1922 closing files for every read/write openration.
1924 Severity : Enhancement
1926 Description: Cache open negative dentries on client when possible.
1927 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1928 negative dentries on lock revocation.
1933 Description: Remounting a client read-only wasn't possible with a zconf mount
1934 Details : It wasn't possible to remount a client read-only with llmount.
1936 Severity : enhancement
1937 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1938 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1939 MPICH is included to add Lustre-specific ADIO interfaces.
1940 This is based closely on the UFS ADIO layer and only differs
1941 in file creation, in order to allow the OST striping to be set.
1942 This is user-contributed code and not supported by CFS.
1947 Description: extended inode attributes (immutable, append-only) work improperly
1948 when 2.4 and 2.6 kernels are used on client/server or vice versa
1949 Details : Introduce kernel-independent values for these flags.
1951 Severity : enhancement
1954 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1955 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1956 granularity. Now, Fractional MB granularity is be supported,
1957 this is very useful for low memory system.
1959 Severity : enhancement
1961 Description: Getattr by fid
1962 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1963 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1964 have locks and avoids getattr rpc requests for it on every stat.
1967 Frequency : Always, for filesystems larger than 2TB
1969 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1970 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1971 filesystems larger than 2TB to be mounted with mballoc enabled.
1974 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1976 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1977 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1978 filesystems larger than 2TB is mounted then the kernel will
1979 silently corrupt the start of the filesystem. CONFIG_LBD is
1980 enabled for all CFS-supported kernels, but the possibility of
1981 this happening with a modified kernel config exists.
1983 Severity : enhancement
1985 Description: add client O_DIRECT support for 2.6 kernels
1986 Details : It is now possible to do O_DIRECT reads and writes to files
1987 in the Lustre client mountpoint on 2.6 kernel clients.
1989 Severity : enhancement
1991 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1992 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1993 parallel, not waiting for response from every OST before sending
1994 a rpc to the next OST.
1999 Description: setattr vs write race when updating file timestamps
2000 Details : Client processes that update a file timestamp into the past
2001 right after writing to the file (e.g. tar) it is possible that
2002 the updated file modification time can be reset to the current
2003 time due to a race between processing the setattr and write RPC.
2005 Severity : enhancement
2007 Description: Bring 'lfs find' closer in line with regular Linux find.
2008 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2009 -print0 options and obtains all the needed info through the lustre
2012 Severity : enhancement
2014 Description: support up to 1024 configured devices on one node
2015 Details : change obd_dev array from statically allocated to dynamically
2016 allocated structs as they are first used to reduce memory usage
2021 Description: Flush dirty partially truncated pages during truncate
2022 Details : Immediatelly flush partially truncated pages in filter_setattr,
2023 this way we completely avoid having any pages in page cache on OST
2024 and can retire ugly workarounds during writes to flush such pages.
2029 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2030 Details : protect inode from truncation within vfs_unlink() context
2031 just take a reference before calling vfs_unlink() and release it
2032 when parent's i_sem is free.
2037 Description: last_id value checked outside lock on OST caused LASSERT failure
2038 Details : If there were multiple MDS->OST object precreate requests in
2039 flight, it was possible that the OST's last object id was checked
2040 outside a lock and incorrectly tripped an assertion. Move checks
2041 inside locks, and discard old precreate requests.
2044 Frequency : always, if extents are used on OSTs
2046 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2047 Details : OSTs using the extents format would not zero the high 16 bits of
2048 the index physical block number. This is not a problem for any
2049 OST filesystems smaller than 16TB, and no kernels support ext3
2050 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2051 new/modified files) and can be fixed for existing filesystems
2052 with e2fsprogs-1.39-cfs1.
2057 Description: import connection selection may be incorrect if timer wraps
2058 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2059 import connections to be ignored if the 32-bit jiffies counter
2060 wraps. Use a 64-bit jiffies counter.
2063 Frequency : very large clusters immediately after boot
2065 Description: LNET request buffers exhausted under heavy short-term load
2066 Details : If a large number of client requests are generated on a service
2067 that has previously never seen so many requests it is possible
2068 that the request buffer growth cannot keep up with the spike in
2069 demand. Instead of dropping incoming requests, they are held in
2070 the LND until the RPC service can accept more requests.
2073 Frequency : Sometimes during replay
2075 Description: Assertion failure in ll_local_open after replay.
2076 Details : If replay happened on an open request reply before we were able
2077 to set replay handler, reply will become not swabbed tripping the
2078 assertion in ll_local_open. Now we set the handler right after
2079 recognising of open request
2082 Frequency : very rare
2084 Description: kernel reports "badness in vsnprintf"
2085 Details : Reading from the "recovery_status" /proc file in small chunks
2086 may cause a negative length in lprocfs_obd_rd_recovery_status()
2087 call to vsnprintf() (which is otherwise harmless). Exit early
2088 if there is no more space in the output buffer.
2090 Severity : enhancement
2092 Description: clear OBD RPC statistics by writing to them
2093 Details : It is now possible to clear the OBD RPC statistics by writing
2094 to the "stats" file.
2099 Description: Client mtime is not the same on different clients after utimes
2100 Details : In some cases, the client was using the utimes() syscall on
2101 a file cached on another node. The clients now validate the
2102 ctime from the MDS + OSTs to determine which one is right.
2107 Description: Inability to activate failout mode
2108 Details : lconf script incorrectly assumed that in pythong string's numeric
2109 value is used in comparisons.
2112 Frequency : always with multiple stripes per file
2114 Description: Inefficient object allocation for mutli-stripe files
2115 Details : When selecting which OSTs to stripe files over, for files with
2116 a stripe count that divides evenly into the number of OSTs,
2117 the MDS is always picking the same starting OST for each file.
2118 Return the OST selection heuristic to the original design.
2123 Description: mount failures may take full timeout to return an error
2124 Details : Under some heavy load conditions it is possible that a
2125 failed mount can wait for the full obd_timeout interval,
2126 possibly several minutes, before reporting an error.
2127 Instead return an error as soon as the status is known.
2129 Frequency : quota enabled and large files being deleted
2131 Description: releasing more than 4GB of quota at once hangs OST
2132 Details : If a user deletes more than 4GB of files on a single OST it
2133 will cause the OST to spin in an infinite loop. Release
2134 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2139 Description: statfs data retrieved from /proc may be stale or zero
2140 Details : When reading per-device statfs data from /proc, in the
2141 {kbytes,files}_{total,free,avail} files, it may appear
2142 as zero or be out of date.
2145 Frequency : systems with MD RAID1 external journal devices
2147 Description: lconf's call to blkid is confused by RAID1 journal devices
2148 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2149 of returning all block devices that match the journal UUID.
2152 Frequency : always, for aggregate stripe size over 4GB
2154 Description: assertion fails when trying to use 4GB stripe size
2155 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2156 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2159 Frequency : always on ppc64
2161 Description: the first write on an ext3 filesystem with mballoc got stuck
2162 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2163 perform endianness conversion.
2165 ------------------------------------------------------------------------------
2167 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2169 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2170 INTEROPERATE with older versions automatically. Please read the
2171 user documentation before upgrading any part of a live system.
2172 * WARNING: Lustre networking configuration changes are required with
2173 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2176 * Support for kernels:
2177 2.6.9-22.0.2.EL (RHEL 4)
2178 2.6.5-7.244 (SLES 9)
2179 2.6.12.6 vanilla (kernel.org)
2182 Severity : enhancement
2183 Bugzilla : 7981/8208
2184 Description: Introduced Lustre Networking (LNET)
2185 Details : LNET is new networking infrastructure for Lustre, it includes
2186 a reorganized network configuration mode (see the user
2187 documentation for full details) as well as support for routing
2188 between different network fabrics. Lustre Networking Devices
2189 (LNDs) for the supported network fabrics have also been
2190 created for this new infrastructure.
2192 Severity : enhancement
2193 Description: Introduced Access control lists
2194 Details : clients can set ACLs on files and directories in order to have
2195 more fine-grained permissions than the standard Unix UGO+RWX.
2196 The MDS must be started with the "-o acl" mount option.
2198 Severity : enhancement
2199 Description: Introduced filesystem quotas
2200 Details : Administrators may now establish per-user quotas on the
2203 Severity : enhancement
2205 Description: Configuration change for the XT3
2206 The PTLLND is now used to run Lustre over Portals on the XT3
2207 The configure option(s) --with-cray-portals are no longer used.
2208 Rather --with-portals=<path-to-portals-includes> is used to
2209 enable building on the XT3. In addition to enable XT3 specific
2210 features the option --enable-cray-xt3 must be used.
2215 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2216 Details : A race between allocating a new llcd and re-getting the llcd_lock
2217 allowed another thread to grab newly-allocated llcd.
2219 Severity : enhancement
2221 Description: 2.6 OST async journal commit and locking fix to improve performance
2222 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2223 2.6 kernels are now async as they already were in 2.4 kernels so
2224 that they can commit concurrently with the network bulk transfer.
2225 For block-allocated files the filter allocation semaphore is held
2226 to avoid filesystem fragmentation during allocation. BKL lock
2227 removed for 2.6 xattr operations where it is no longer needed.
2232 Description: lconf incorrectly determined whether two IP networks could talk
2233 Details : In some more complicated routing and multiple-network
2234 configurations, lconf will avoid trying to make a network
2235 connection to a disjoint part of the IP space. It was doing the
2236 math incorrectly for one set of cases.
2241 Description: Fix for potential infinite loop processing records in an llog.
2242 Details : If an llog record is corrupted/zeroed, it is possible to loop
2243 forever in llog_process(). Validate the llog record length
2244 and skip the remainder of the block on error.
2247 Frequency : occasional (liblustre only)
2249 Description: liblustre could not open files whose last component is a symlink
2250 Details : sysio_path_walk() would incorrectly pass the open intent to
2251 intermediate path components.
2254 Frequency : rare (liblustre only with non-standard tuning)
2255 Bugzilla : 7201 (7350)
2256 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2257 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2258 after it has been released, only do this for referenced locks
2260 Severity : enhancement
2262 Description: specifying an (invalid) directory default stripe_size of -1
2263 would reset the directory default striping
2264 Details : stripe_size -1 was used internally to signal directory stripe
2265 removal, now use "all default" to signal dir stripe removal
2266 as a directory striping of "all default" is not useful
2269 Frequency : common for large clusters running liblustre clients
2271 Description: doing an ls when liblustre clients are running is slow
2272 Details : sending a glimpse AST to a liblustre client waits for every AST
2273 to time out, as liblustre clients will not respond. Since they
2274 cannot cache data we refresh the OST lock LVB from disk instead.
2276 Severity : enhancement
2278 Description: doing an ls at the same time as file IO can be slow
2279 Details : enqueue and other "small" requests can be blocked behind many
2280 large IO requests. Create a new OST IO portal for non-IO
2281 requests so they can be processed faster.
2284 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2286 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2287 created file with mode 000
2288 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2289 allow an ftruncate() as a non-root user to a file with mode 000.
2290 Lustre now respects this flag to disable mode checks when
2291 truncating a file owned by the user
2294 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2297 Description: Revoking locks from clients that went dead or catatonic might take
2299 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2300 cancellation of such locks instant on servers without waiting for
2301 any reply from clients. Clients drops these locks when cancel
2302 notification from server is received without replying.
2305 Frequency : liblustre-only, when liblustre client dies or becomes busy
2307 Description: Doing ls on Linux clients can take a long time with active
2309 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2310 granting such locks to it in the first place if possible. Locks
2311 are taken by proxy on the OST during the read or write and
2312 dropped immediately afterward. Add connect flags handling, do
2313 not grant locks to liblustre clients for glimpse ASTs.
2315 Severity : enhancement
2317 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2318 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2319 sequential reads and behaves badly with applications doing random
2320 reads. Improve it by reading ahead at least read region, and
2321 avoiding excessive large RPC for small reads.
2323 Severity : enhancement
2325 Description: Creating more than 1000 files for a single job may cause a load
2326 imbalance on the OSTs if there are also a large number of OSTs.
2327 Details : qos_prep_create() uses an OST index reseed value that is an
2328 even multiple of the number of available OSTs so that if the
2329 reseed happens in the middle of the object allocation it will
2330 still utilize the OSTs as uniformly as possible.
2335 Description: OST or MDS may oops in ping_evictor_main()
2336 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2337 but doesn't restart at beginning of obd_exports_timed list
2340 Severity : enhancement
2342 Description: improve by-nid export eviction on the MDS and OST
2343 Details : allow multiple exports with the same NID to be evicted at one
2344 time without re-searching the exports list.
2347 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2349 Description: MDS may oops in groups_free()
2350 Details : in rare race conditions a newly allocated group_info struct is
2351 freed again, and this can be NULL. The 2.4 compatibility code
2352 for groups_free() checked for a NULL pointer, but 2.6 did not.
2355 Frequency : common for liblustre clients doing little filesystem IO
2356 Bugzilla : 9352, 7313
2357 Description: server may evict liblustre clients accessing contended locks
2358 Details : if a client is granted a lock or receives a completion AST
2359 with a blocking AST already set it would not reply to the AST
2360 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2363 Frequency : lfs setstripe, only systems with more than 160 OSTs
2365 Description: unable to set striping with a starting offset beyond OST 160
2366 Details : llapi_create_file() incorrectly limited the starting stripe
2367 index to the maximum single-file stripe count.
2370 Frequency : LDAP users only
2372 Description: lconf did not handle in-kernel recovery with LDAP properly
2373 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2375 Severity : enhancement
2377 Description: bind OST threads to NUMA nodes to improve performance
2378 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2379 node and do their allocations there to localize memory access
2381 Severity : enhancement
2383 Description: llmount can determine client NID directly from Myrinet (GM)
2384 Details : the client NID code from gmnalnid was moved directly into
2385 llmount, removing the need to use this or specifying the
2386 client NID explicitly when mounting GM clients with zeroconf
2389 Frequency : if client is started with down MDS
2391 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2392 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2393 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2394 ptlrpc_import_delayed_req() because we didn't check if the
2395 request was interrupted, and we also didn't break out of the
2396 event loop if there was no timeout
2401 Description: data loss during non-page-aligned writes to a single file from
2402 both multiple nodes and multiple threads on one node at same time
2403 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2404 inconsistency led to false short-reads, that were cached and later
2405 used by ->prepare_write() to fill in partially written page,
2406 leading to data loss.
2409 Frequency : always, if lconf --abort_recovery used
2411 Description: lconf --abort_recovery fails with 'Operation not supported'
2412 Details : lconf was attempting to abort recovery on the MDT device and not
2415 Severity : enhancement
2417 Description: remove cleanup logs
2418 Details : replace lconf-generated cleanup logs with lustre internal
2419 cleanup routines. Eliminates the need for client-cleanup and
2422 Severity : enhancement
2424 Description: add support for EAs (user and system) on lustre filesystems
2425 Details : it is now possible to store extended attributes in the Lustre
2426 client filesystem, and with the user_xattr mount option it
2427 is possible to allow users to store EAs on their files also
2429 Severity : enhancement
2431 Description: Add possibility (config option) to show minimal available OST free
2433 Details : When compiled with --enable-mindf configure option, statfs(2)
2434 (and so, df) will return least minimal free space available from
2435 all OSTs as amount of free space on FS, instead of summary of
2436 free spaces of all OSTs.
2438 Severity : enhancement
2440 Description: do not expand extent locks acquired on OST-side
2441 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2442 by server: they are not cached anyway.
2445 Frequency : when mmap is used/binaries executed from Lustre
2447 Description: Unmmap pages before throwing them away from read cache.
2448 Details : llap_shrink cache now attempts to unmap pages before discarding
2449 them (if unmapping failed - do not discard). SLES9 kernel has
2450 extra checks that trigger if this unmapping is not done first.
2455 Description: lconf didn't resolve symlinks before checking to see whether a
2456 given mountpoint was already in use
2459 Frequency : when migrating failover services
2460 Bugzilla : 6395, 9514
2461 Description: When migrating a subset of services from a node (e.g. failback
2462 from a failover service node) the remaining services would
2463 time out and evict clients.
2464 Details : lconf --force (implied by --failover) sets the global obd_timeout
2465 to 5 seconds in order to quickly disconnect, but this caused
2466 other RPCs to time out too quickly. Do not change the global
2467 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2469 Severity : enhancement
2470 Frequency : if MDS is started with down OST
2471 Bugzilla : 9439,5706
2472 Description: Allow startup/shutdown of an MDS without depending on the
2473 availability of the OSTs.
2474 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2475 Add appropriate locking and lov-osc refcounts for safe
2476 cleaning. Add osc abort_inflight calls in case the
2477 synchronize never started.
2480 Frequency : occasional (Cray XT3 only)
2482 Description: root not authorized to access files in CRAY_PORTALS environment
2483 Details : The client process capabilities were not honoured on the MDS in
2484 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2485 been accessed by an authorized user then root was able to access
2486 the file on the local client also. The root user capabilities
2487 are now allowed on the MDS, as this environment has secure UID.
2490 Frequency : occasional
2492 Description: ldiskfs "too long searching" message happens too often
2493 Details : A debugging message (otherwise harmless) prints too often on
2494 the OST console. This has been reduced to only happen when
2495 there are fragmentation problems on the filesystem.
2500 Description: Division by zero in statfs when all OSCs are inactive
2501 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2502 lov_statfs divides by value returned by lov_get_stripecnt().
2506 Bugzilla : 9489, 3273
2507 Description: First write from each client to each OST was only 4kB in size,
2508 to initialize client writeback cache, which caused sub-optimal
2509 RPCs and poor layout on disk for the first writen file.
2510 Details : Clients now request an initial cache grant at (re)connect time
2511 and so that they can start streaming writes to the cache right
2512 away and always do full-sized RPCs if there is enough data.
2513 If the OST is rebooted the client also re-establishes its grant
2514 so that client cached writes will be honoured under the grant.
2519 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2520 Details : Now I/O RPCs go to different portal number and (presumably) fast
2521 lock requests (and glimses) and other RPCs get their own service
2522 threads pool that should be able to service those RPCs
2525 Severity : enhancement
2527 Description: Ability to exchange lustre version between client and servers and
2528 issue warnings at client side if client is too old. Also for
2529 liblustre clients there is ability to refuse connection of too old
2531 Details : New 'version' field is added to connect data structure that is
2532 filled with version info. That info is later checked by server and
2536 Frequency : rare, liblustre only.
2537 Bugzilla : 9296, 9581
2538 Description: Two simultaneous writes from liblustre at offset within same page
2539 might proceed at the same time overwriting eachother with stale
2541 Details : I/O lock withing llu_file_prwv was released too early, before data
2542 actually was hitting the wire. Extended lock-holding time until
2543 server acknowledges receiving data.
2546 Frequency : extremely rare. Never observed in practice.
2548 Description: avoid generating lustre_handle cookie of 0.
2549 Details : class_handle_hash() generates handle cookies by incrementing
2550 global counter, and can hit 0 occasionaly (this is unlikely, but
2551 not impossible, because initial value of cookie counter is
2552 selected randonly). Value of 0 is used as a sentinel meaning
2553 "unassigned handle" --- avoid it. Also coalesce two critical
2554 sections in this function into one.
2556 Severity : enhancement
2558 Description: allow liblustre clients to delegate truncate locking to OST
2559 Details : To avoid overhead of locking, liblustre client instructs OST to
2560 take extent lock in ost_punch() on client's behalf. New connection
2561 flag is added to handle backward compatibility.
2563 Severity : enhancement
2564 Bugzilla : 4928, 7341, 9758
2565 Description: allow number of OST service threads to be specified
2566 Details : a module parameter allows the number of OST service threads
2567 to be specified via "options ost ost_num_threads={N}" in the
2568 OSS's /etc/modules.conf or /etc/modprobe.conf.
2572 Bugzilla : 6146, 9635, 9895
2573 Description: servers crash with bad pointer in target_handle_connect()
2574 Details : In rare cases when a client is reconnecting it was possible that
2575 the connection request was the last reference for that export.
2576 We would temporarily drop the export reference and get a new
2577 one, but this may have been the last reference and the export
2578 was just destroyed. Get new reference before dropping old one.
2580 Severity : enhancement
2581 Frequency : if client is started with failover MDS
2583 Description: Allow multiple MDS hostnames in the mount command
2584 Details : Try to read the configuration from all specified MDS
2585 hostnames during a client mount in case the "primary"
2588 Severity : enhancement
2590 Description: Stop sending data to evicted clients as soon as possible.
2591 Details : Check if the client we are about to send or are sending data to
2592 was evicted already. (Check is done every second of waiting,
2593 for which l_wait_event interface was extended to allow checking
2594 of exit condition at specified intervals).
2597 Frequency : rare, normally only when NFS exporting is done from client
2599 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2601 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2604 Frequency : always, if srand() is called before liblustre initialization
2606 Description: Liblustre uses system PRNG disturbing its usage by user application
2607 Details : Introduce internal to lustre fast and high-quality PRNG for
2608 lustre usage and make liblustre and some other places in generic
2609 lustre code to use it.
2611 Severity : enhancement
2612 Bugzilla : 9477, 9557, 9870
2613 Description: Verify that the MDS configuration logs are updated when xml is
2614 Details : Check if the .xml configuration logs are newer than the config
2615 logs stored on the MDS and report an error if this is the case.
2616 Request --write-conf, or allow starting with --old_conf.
2618 Severity : enhancement
2620 Description: Handle symlinks in the path when checking if Lustre is mounted.
2621 Details : Resolve intermediate symlinks when checking if a client has
2622 mounted a filesystem to avoid duplicate client mounts.
2627 Description: lconf can hit an error exception but still return success.
2628 Details : The lconf command catches the Command error exception at the top
2629 level script context and will exit with the associated exit
2630 status, but doesn't ensure that this exit status is non-zero.
2635 Description: failure of ptlrpc thread startup can cause oops
2636 Details : Starting a ptlrpc service thread can fail if there are a large
2637 number of threads or the server memory is very fragmented.
2638 Handle this without oopsing.
2641 Frequency : always, only if liblustre and non-default acceptor port was used
2643 Description: liblustre cannot connect to servers with non-default acceptor port
2644 Details : tcpnal_set_default_params() was not called and was therefore
2645 ignoring the environment varaible TCPNAL_PORT, as well as other
2646 TCPNAL_ environment variables
2651 Description: two objects could be created on the same OST for a single file
2652 Details : If an OST is down, in some cases it was possible to create two
2653 objects on a single OST for a single file. No problems other
2654 than potential performance impact and spurious error messages.
2658 Bugzilla : 5681, 9562
2659 Description: Client may oops in ll_unhash_aliases
2660 Details : Client dcache may become inconsistent in race condition.
2661 In some cases "getcwd" can fail if the current directory is
2667 Description: Inode refcounting problems in NFS export code
2668 Details : link_raw functions used to call d_instantiate without obtaining
2669 extra inode reference first.
2673 Bugzilla : 9942, 9903
2674 Description: Referencing freed requests leading to crash, memleaks with NFS.
2675 Details : We used to require that call to ll_revalidate_it was always
2676 followed by ll_lookup_it. Also with revalidate_special() it is
2677 possible to call ll_revalidate_it() twice for the same dentry
2678 even if first occurence returned success. This fix changes semantic
2679 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2680 reference on a request referred from the intent.
2681 ll_intent_release() then releases such a request.
2684 Frequency : rare, normally benchmark loads only
2686 Description: unlinked inodes were kept in memory on the client
2687 Details : If a client is repeatedly creating and unlinking files it
2688 can accumulate a lot of stale inodes in the inode slab cache.
2689 If there is no other client load running this can cause the
2690 client node to run out of memory. Instead flush old inodes
2691 from client cache that have the same inode number as a new inode.
2694 Frequency : SLES9 2.6.5 kernel and long filenames only
2695 Bugzilla : 9969, 10379
2696 Description: utime reports stale NFS file handle
2697 Details : SLES9 uses out-of-dentry names in some cases, which confused
2698 the lustre dentry revalidation. Change it to always use the
2702 Frequency : rare, unless heavy write-truncate concurrency is continuous
2703 Bugzilla : 4180, 6984, 7171, 9963, 9331
2704 Description: OST becomes very slow and/or deadlocked during object unlink
2705 Details : filter_destroy() was holding onto the parent directory lock
2706 while truncating+unlinking objects. For very large objects this
2707 may block other threads for a long time and slow overall OST
2708 responsiveness. It may also be possible to get a lock ordering
2709 deadlock in this case, or run out of journal credits because of
2710 the combined truncate+unlink. Solution is to do object truncate
2711 first in one transaction without parent lock, and then do the
2712 final unlink in a new transaction with the parent lock. This
2713 reduces the lock hold time dramatically.
2716 Frequency : rare, 2.4 kernels only
2718 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2719 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2720 before unmounting their filesystem in order to avoid deadlock.
2721 The kernel_locked() function in 2.4 kernels only checks whether
2722 the kernel lock is held, not whether it is this process that is
2723 holding it as 2.6 kernels do.
2728 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2729 Details : The client ptlrpc code may be trying to reconnect to a down
2730 server before a previous connection attempt has timed out.
2731 Increase the reconnect interval to be longer than the connection
2732 timeout interval to avoid sending duplicate connections to
2736 Frequency : echo_client brw_test command
2738 Description: fix echo_client to work with OST preallocated code
2739 Details : OST preallocation code (5137) didn't take echo_client IO path
2740 into account: echo_client calls filter methods outside of any
2741 OST thread and, hence, there is no per-thread preallocated
2742 pages and buffers to use. Solution: hijack pga pages for IO. As
2743 a byproduct, this avoids unnecessary data copying.
2747 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2748 Description: Client can oops in mdc_commit_close() after open replay
2749 Details : It was possible for the MDS to return an open request with no
2750 transaction number in mds_finish_transno() if the client was
2751 evicted, but without actually returning an error. Clients
2752 would later try to replay that open and may trip an assertion
2753 Simplify the client close codepath, and always return an error
2754 from the MDS in case the open is not successful.
2757 Frequency : rare, 2.6 OSTs only
2759 Description: OST may deadlock under high load on fragmented files
2760 Details : If there was a heavy load and highly-fragmented OST filesystems
2761 it was possible to have all the OST threads deadlock waiting on
2762 allocation of biovecs, because the biovecs were not released
2763 until the entire RPC IO was completed. Instead, release biovecs
2764 as soon as they are complete to ensure forward IO progress.
2766 Severity : enhancement
2768 Description: Support for specifying external journal device at mount
2769 Details : If an OST or MDS device is formatted with an external journal
2770 device, this device major/minor is stored in the ext3 superblock
2771 and may not be valid for failover. Allow detecting and
2772 specifying the external journal at mount time.
2777 Description: Mounting an MDS with pending unlinked files may cause oops
2778 Details : target_finish_recovery() calls mds_postrecov() which returned
2779 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2780 considers this an error and immediately begins cleaning up the
2781 lov, just after starting the mds_lov process
2783 Severity : enhancement
2785 Description: Implement 'lfs df' to report actual free space on per-OST basis
2786 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2787 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2788 usage of objects; '-h' to report in human readable format.
2790 ------------------------------------------------------------------------------
2792 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2799 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2800 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2801 at startup during orphan unlinking. Instead, skip buddy update
2802 before setup, as it will be regenerated after recovery is complete.
2807 Description: If an OST is inactive, its locks might reference stale inodes.
2808 Details : lov_change_cbdata() must iterate over all namespaces, even if
2809 they are inactive to clear inode references from the lock.
2811 Severity : enhancement
2812 Frequency : occasional, if non-standard max_dirty_mb used
2814 Description: Client will block write RPCs if not enough grant
2815 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2816 then the client will block writes while waiting for another RPC
2817 to complete instead of consuming its dirty limit. With change
2818 we get improved performance when max_dirty_mb is small.
2820 Severity : enhancement
2821 Bugzilla : 3389, 6253
2822 Description: Add support for supplementary groups on the MDS.
2823 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2824 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2825 lookups for user supplementary groups into a cache.
2829 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2830 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2831 for liblustre O_EXCL is enforced by MDS.
2835 Description: readdir never returns NULL in liblustre.
2836 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2837 dentry in struct dirent.
2841 Frequency : liblustre only, depends on application IO pattern
2842 Description: liblustre clients evicted if not contacting servers
2843 Details : Don't put liblustre clients into the ping_evictor list, so
2844 they will not be evicted by the pinger ever.
2846 Severity : enhancement
2848 Description: Add ability to evict clients by NID from MDS.
2849 Details : By echoing "nid:$NID" string into
2850 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2851 $NID would be instantly evicted from this MDS and from all active
2852 OSTs connected to it.
2856 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2857 Details : lookup_it_finish() used to query file size from OSTs that was not
2862 Description: service threads change working directory to that of init
2863 Details : Starting lustre service threads may pin the working directory
2864 of the parent thread, making that filesystem busy. Threads
2865 now change to the working directory of init to avoid this.
2869 Frequency : during shutdown only
2870 Description: shutdown with a failed MDS or OST can cause unmount to hang
2871 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2876 Frequency : 2.6 only
2877 Description: chmod/chown may include an extra supplementary group
2878 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2879 supplementary group and if none is specified this is used.
2882 Bugzilla : 5479 (6816)
2884 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2885 Details : If lookup is in progress on a file that is unlinked we might try
2886 to revalidate the inode and fail in revalidate after lookup is
2887 complete and ll_file_open() enqueues the open again but
2888 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2891 Frequency : always, if lconf --abort_recovery used
2893 Description: lconf --abort_recovery fails with 'Operation not supported'
2894 Details : lconf was attempting to abort recovery on the MDT device and not
2897 ------------------------------------------------------------------------------
2899 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2904 Frequency : rare (only unsupported configurations with a node running as an
2906 Bugzilla : 6514, 5137
2907 Description: Mounting a Lustre file system on a node running as an OST could
2909 Details : OSTs now preallocates memory needed to write out data at
2910 startup, instead of when needed, to avoid having to
2911 allocate memory in possibly low memory situations.
2912 Specifically, if the file system is mounted on on OST,
2913 memory pressure could force it to try to write out data,
2914 which it needed to allocate memory to do. Due to the low
2915 memory, it would be unable to do so and the node would
2916 become unresponsive.
2918 Severity : enhancement
2920 Description: Addition of lconf --service command line option
2921 Details : lconf now accepts a '--service <arg>' option, which is
2922 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2924 Severity : enhancement
2926 Description: Failover mode is now the default for OSTs.
2927 Details : By default, OSTs will now run in failover mode. To return to
2928 the old behaviour, add '--failout' to the lmc line for OSTs.
2930 Severity : enhancement
2932 Description: Health checks are now provided for MDS and OSTs
2933 Details : Additional detailed health check information on MSD and OSTs
2934 is now provided through the procfs health_check value.
2937 Frequency : occasional, depends on IO load
2939 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2940 after numerous create/delete cycles
2941 Details : The ext3 inode allocation policy would not allocate new inodes
2942 very well on the OSTs because there are no new directories
2943 being created. Instead we look for groups with free space if
2944 the parent directories are nearly full.
2949 Description: Network or server problems during mount may cause partially
2950 mounted clients instead of returning an error.
2951 Details : The config llog parsing code may overwrite the error return
2952 code during mount error handling, returning success instead
2958 Description: MDS can fail to allocate large reply buffers
2959 Details : After long uptimes the MDS can fail to allocate large reply
2960 buffers (e.g. zconf client mount config records) due to memory
2961 fragmentation or consumption by the buffer cache. Preallocate
2962 some large reply buffers so that these replies can be sent even
2963 under memory pressure.
2967 Frequency : rare (liblustre)
2968 Description: fsx running with liblustre complained that using truncate() to
2969 extend the file doesn't work. This patch corrects that issue.
2970 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2971 ATTR_SIZE and lsm use in llu_setattr_raw.
2975 Frequency : rare, only 2.6 kernels
2976 Description: Unusual file access patterns on the MDS may result in inode
2977 data being lost in very rare circumstances.
2978 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2979 code in the 2.6 kernel caused the inode and/or EA data not to
2980 be read from disk, causing single-file corruption.
2984 Frequency : rare, only 2.6 filesystems using extents
2985 Description: Heavy concurrent write and delete load may cause data corruption.
2986 Details : It was possible under high-load situations to have an extent
2987 metadata block in the block device cache from a just-unlinked
2988 file overwrite a newly-allocated data block. We now unmap any
2989 metadata buffers that alias just-allocated data blocks.
2993 Frequency : filesystems with default stripe_count larger than 77
2994 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2995 Details : lconf specifies an inode size of 4096 bytes when the default
2996 stripe_count is larger than 77. This conflicts with the default
2997 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2998 this case to avoid pinning too much memory for large EAs.
3000 ------------------------------------------------------------------------------
3002 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3007 Frequency : rare (extremely heavy IO load with hundreds of clients)
3009 Description: Client is evicted, gets IO error writing to file
3010 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3011 caused clients to be evicted for AST timeouts. The fixes in
3012 bug 5192 mean we no longer need to have such short AST timeouts
3013 so ldlm_timeout has been increased.
3016 Frequency : occasional during --force or --failover shutdown under load
3017 Bugzilla : 5949, 4834
3018 Description: Server oops/LBUG if stopped with --force or --failover under load
3019 Details : a collection of import/export refcount and cleanup ordering
3020 issues fixed for safer force cleanup
3023 Frequency : only filesystems larger than 120 OSTs
3024 Bugzilla : 5990, 6223
3025 Description: lfs getstripe would oops on a very large filesystem
3026 Details : lov_getconfig used kfree on vmalloc'd memory
3029 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3030 Bugzilla : 6242, 6243
3031 Description: reading from files that had been truncated to a non-zero size
3032 but never opened returned no data
3033 Details : ll_file_read() reads zeros from no-object files to EOF
3038 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3039 Details : There's little harm in aborting MDS/OSS recovery and letting it
3040 try again, so I removed the LASSERT and return an error instead.
3042 Severity : enhancement
3044 Description: New debugging infrastructure for tracking down data corruption
3045 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3046 (b) cover more of the client-side code path, and (c) try to narrow
3047 down where problems occurred
3051 Bugzilla : 3819, 4364, 4397, 6313
3052 Description: Racing close and eviction MDS could cause assertion in mds_close
3053 Details : It was possible to get multiple mfd references during close and
3054 client eviction, leading to one thread referencing a freed mfd.
3056 Severity: : enhancement
3057 Bugzilla : 3262, 6359
3058 Description: Attempts to reconnect to servers are now more aggressive.
3059 Details : This builds on the enhanced upcall-less recovery that was added
3060 in 1.4.2. When trying to reconnect to servers, clients will
3061 now try each server in the failover group every 10 seconds. By
3062 default, clients would previously try one server every 25 seconds.
3067 Description: After recovery, certain operations trigger a failed
3068 assertion on a client.
3069 Details : Failing over an mds, using lconf -d --failover, while a
3070 client was doing a readdir() call would cause the client to
3071 LBUG after recovery completed and the readdir() was resent.
3073 Severity : enhancement
3075 Description: Default groups are now added by lconf
3076 Details : You can now run lconf --group <servicename> without having to
3077 manually add groups with lmc.
3080 Frequency : occasional
3082 Description: Nodes with an elan id of 0 trigger a failed assertion
3085 Frequency : always when accessing e.g. tty/console device nodes
3087 Description: tty and some other devices nodes cannot be used on lustre
3088 Details : file's private_data field is used by device data and lustre
3089 values in there got lost. New field was added to struct file to
3090 store fs-specific private data.
3093 Frequency : when exporting Lustre via NFS
3095 Description: NFSD failed occasionally when looking up a path component
3096 Details : NFSD is looking up ".." which was broken in ext3 directories
3097 that had grown large enough to become hashed.
3100 Frequency : Clusters with multiple interfaces not on the same subnet
3102 Description: Nodes will repeatedly try to reconnect to an interface which it
3103 cannot reach and report an error to the log.
3104 Details : Extra peer list entries will be created by lconf with some peers
3105 unreachable. lconf now validates the peer before adding it.
3108 Frequency : Only if a default stripe is set on the filesystem root.
3110 Description: Setting a default stripe on the filesystem root prevented the
3111 filesystem from being remounted.
3112 Details : The client was sending extra request flags in the root getattr
3113 request and did not allocate a reply buffer for the dir EA.
3116 Frequency : occasional, higher if lots of files are accessed by one client
3117 Bugzilla : 6159, 6097
3118 Description: Client trips assertion regarding lsm mismatch/magic
3119 Details : While revalidating inodes the VFS looks up inodes with ifind()
3120 and in rare cases can find an inode that is being freed.
3121 The ll_test_inode() code will free the lsm during ifind()
3122 when it finds an existing inode and then the VFS later attaches
3123 this free lsm to a new inode.
3127 Bugzilla : 6422, 7030
3128 Description: MDS deadlock between mkdir and client eviction
3129 Details : Creating a new file via mkdir or mknod (starting a transaction
3130 and getting the ns lock) can deadlock with client eviction
3131 (gets ns lock and trying to finish a synchronous transaction).
3134 Frequency : occasional
3135 Description: While starting a server, the fsfilt_ext3 module could not be
3137 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3138 kernels. Previously, lconf would still use the ext3 name
3139 when trying to load modules. Now, it will correctly use
3140 ext3 on 2.4 and ldiskfs on 2.6.
3142 Severity : enhancement
3143 Description: The default stripe count has been changed to 1
3144 Details : The interpretation of the default stripe count (0, to lfs
3145 or lmc) has been changed to mean striping across a single
3146 OST, rather than all available. For general usage we have
3147 found a stripe count of 1 or 2 works best.
3149 Severity : enhancement
3150 Description: Add support for compiling against Cray portals.
3151 Details : Conditional compiling for some areas that are different
3155 Frequency : occasional
3156 Bugzilla : 6409, 6834
3157 Description: Creating files with an explicit stripe count may lead to
3158 a failed assertion on the MDS
3159 Details : If some OSTs are full or unavailable, creating files may
3160 trigger a failed assertion on the MDS. Now, Lustre will
3161 try to use other servers or return an error to the
3165 Frequency : occasional
3167 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3168 caused lock timeout during readahead (since 1.4.2).
3169 Details : Processes doing readahead might match a lock that hasn't been
3170 granted yet if there are overlapping and conflicting lock
3171 requests. The readahead process waits on ungranted lock
3172 (original lock is CBPENDING), while OST waits for that process
3173 to cancel CBPENDING read lock and eventually evicts client.
3175 Severity : enhancement
3177 Description: Initial enabling of flock support for clients
3178 Details : Implements fcntl advisory locking and file status functions.
3179 This feature is provided as an optional mount flag (default
3180 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3181 locking are implemented yet, and those that are are not guaranteed
3182 to be completely correct in production environments.
3183 mount -t lustre -o [flock|noflock] ...
3186 Frequency : occasional
3188 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3189 assertion in the ext3 JBD (journaling) layer.
3190 Details : The b_committed_data struct is protected by the big kernel lock
3191 in 2.4 kernels, serializing journal_commit_transaction() and
3192 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3193 access to this struct. In 2.6 kernels there is finer grained
3194 locking to improve SMP performance of the JBD layer.
3198 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3203 Description: merge and cleanup kernel patches.
3206 -----------------------------------------------------------------------------
3208 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3210 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3211 versions of Lustre. Please update all servers and clients to
3212 version 1.4.2 or later at the same time. You must also run
3213 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3214 to update the configuration logs.
3216 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3217 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3218 - swab llog records of type '0' so we get proper header size/idx (5861)
3219 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3220 - fix rename of one directory over another leaking an inode (5953)
3221 - avoid SetPageDirty on 2.6 (5981)
3222 - don't re-add just-being-destroyed locks to the waiting list (5653)
3223 - when creating new directories, inherit the parent's custom
3224 striping settings if present parent (3048)
3225 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3226 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3227 - don't hold client locks on temporary worklist from l_lru (5666)
3228 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3229 - automatically evict dead clients (5921)
3230 - Update file size properly in create+truncate+fstat case (6196)
3231 - Do not unhash mountpoint dentries, do not allow removal of
3233 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3234 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3235 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3236 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3237 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3238 - Correctly handle reads of files with no objects (6243)
3239 - lctl recover will also mark a device active if deactivate used (5933)
3241 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3242 - allow --write-conf on an MDS with different nettype than client (5619)
3243 - don't write config llogs to MDS for mounts not from that MDS (5617)
3244 - lconf should create multiple TCP connections from a client (5201)
3245 - init scripts are now turned off by default; run chkconfig --on
3246 lustre and chkconfig --on lustrefs to use them
3247 - upcalls are no longer needed for clients to recover to failover
3249 - add --abort-recovery option to lconf to abort recovery on device
3251 - add support for an arbitrary number of OSTs (3026)
3252 - Quota support protocol changes.
3253 - forward compatibility changes to wire structs (6007)
3254 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3255 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3256 - improved "lustre status" script
3257 - initialize blocksize for non-regular files (6062)
3258 - added --disable-server and --disable-client configure options (5782)
3259 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3260 - Vanilla 2.4.29 support
3261 - increase maximum number of obd devices to 520 (6242)
3262 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3263 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3264 SLES 9 2.6.5 kernels
3265 - limit stripes per file to 160 (the maximum EA size) (6093)
3267 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3270 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3271 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3272 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3273 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3274 - small ext3 extents cleanups and fixes (5733)
3275 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3276 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3277 - avoid needless client->OST connect, fix handle mismatch (5317)
3278 - fix DLM error path that led to out-of-sync client, long delays (5779)
3279 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3280 - fix several locking issues related to i_size (5492,5624,5654,5672)
3281 - don't move pending lock onto export if it is already evicted (5683)
3282 - fix kernel oops when creating .foo in unlinked directory (5548)
3283 - fix deadlock in obdfilter statistics vs. object create (5811)
3284 - use time_{before,after} to avoid timer jiffies wrap (5882)
3285 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3286 - Do not leak request if server was not able to process it (5154)
3287 - If mds_open unable to find parent dir, make that negative lookup(5154)
3288 - don't create new directories with extent-mapping (5909, 5936)
3290 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3291 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3292 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3293 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3294 - support for RHEL4 kernel on i686 (5773)
3295 - provide error messages when incompatible logs are encountered (5898)
3297 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3298 * version 1.4.0.10 (1.4.1 release candidate 1)
3300 - don't keep a lock reference when lock is not granted (4238)
3301 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3302 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3303 - drop import inflight refcount on signal_completed_replay error (5255)
3304 - unlock page after async write error during send (3677)
3305 - handle missing objects in filter_preprw_read properly (5265)
3306 - no transno return for symlink open, don't save no-trasno open (3440)
3307 - don't try to complete elan receive that already failed (4012)
3308 - free RPC server reply state on error (5406)
3309 - clean up thread from ptlrpc_start_thread() on error (5160)
3310 - readahead could read extra page into cache that wasn't ejected (5388)
3311 - prevent races in class_attach/setup/cleanup/detach (5260)
3312 - don't dereference de->d_inode after l_dput of de (5458)
3313 - use "int" for stripe value returned from lock_to_stripe (5544)
3314 - mballoc allocation and error-checking fixes in 2.6 (5504)
3315 - block device patches to fix I/O request sizes in 2.6 (5482)
3316 - look up hostnames for IB nals (5602)
3317 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3318 - don't start multiple acceptors for the same port (5277)
3319 - fix incorrect LASSERT in mds_getattr_name (5635)
3320 - export a proc file for general "ping" checking (5628)
3321 - fix "lfs check" to not block when the MDS is down (5628)
3323 - service request history (4965)
3324 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3325 - create an "evict_client" /proc entry on OSTs, like the MDS has
3326 - fix mount usage message, return errors per mount(8) (5168)
3327 - change grep [] to grep "[]" in tests so they work in more UMLs
3328 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3329 - remove ancient LOV_MAGIC_V0 EA support (5047)
3330 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3331 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3332 - allow readahead allocations to fail when low on memory (5383)
3333 - mmap locking landed again, after considerable improvement (2828)
3334 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3336 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3339 - send OST transaction number in read/write reply to free req (4966)
3340 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3341 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3342 - the watchdog thread now runs as interruptible (5246)
3343 - flock/lockf fixes (but it's still disabled, pending 5135)
3344 - don't use EXT3 constants in llite code (5094)
3345 - memory shortage at startup could cause assertion (5176)
3347 - reorganization of lov code
3348 - single portals codebase
3350 - add extents/mballoc support (5025)
3351 - direct I/O reads in the obdfilter (4048)
3352 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3354 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3357 - send OST transaction number in read/write reply to free req (4966)
3358 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3359 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3360 - the watchdog thread now runs as interruptible (5246)
3361 - handle missing objects in filter_preprw_read properly (5265)
3362 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3363 - drop import inflight refcount on signal_completed_replay error (5255)
3364 - unlock page after async write error during send (3677)
3365 - return original error code on reconstructed replies (3761)
3366 - no transno return for symlink open, don't save no-trasno open (3440)
3368 - add pid to ldlm debugging output (4922)
3369 - bump the watchdog timeouts -- we can't handle 30sec yet
3370 - extra debugging for orphan dentry/inode bug (5259)
3372 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3375 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3376 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3377 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3378 - fix NULL dereference in /proc/sys/portals/routes (4827)
3379 - allow failed mdc_close() operations to be interrupted (4561)
3380 - stop precreate on OST before MDS would time out on it (4778)
3381 - don't send partial-page writes before EOF from client (4410)
3382 - discard client grant for sub-page writes on large-page clients (4520)
3383 - don't free dentries not owned by NFS code, check generation (4806)
3384 - fix lsm leak if mds_create_objects() fails (4801)
3385 - limit debug_daemon file size, always print CERROR messages (4789)
3386 - use transno after validating reply (3892)
3387 - process timed out requests if import state changes (3754)
3388 - update mtime on OST during writes, return in glimpse (4829)
3389 - add mkfsoptions to LDAP (4679)
3390 - use ->max_readahead method instead of zapping global ra (5039)
3391 - don't interrupt __l_wait_event() during strace
3393 - add software watchdogs to catch hung threads quickly (4941)
3394 - make lustrefs init script start after nfs is mounted
3395 - fix CWARN/ERROR duplication (4930)
3396 - return async write errors to application if possible (2248)
3397 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3398 - print NAL number in %x format (4645)
3399 - update barely-supported suse-2.4.21-171 series (4842)
3400 - support for sles 9 %post scripts
3401 - support for building 2.6 kernel-source packages
3402 - support for sles km_* packages
3404 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3407 - ignore -ENOENT errors in osc_destroy (3639)
3408 - notify osc create thread that OSC is being cleaned up (4600)
3409 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3410 - reconstruct ost_handle() like mds_handle() (4657)
3411 - create a new thread to do import eviction to avoid deadlock (3969)
3412 - let lconf resolve symlinked-to devices (4629)
3413 - don't unlink "objects" from directory with default EA (4554)
3414 - hold socknal file ref over connect in case target is down (4394)
3415 - allow more than 32000 subdirectories in a single directory (3244)
3416 - fix blocks count for O_DIRECT writes (3751)
3417 - OST returns ENOSPC from object create when no space left (4539)
3418 - don't send truncate RPC if file size isn't changing (4410)
3419 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3420 - bind to privileged port in socknal and tcpnal (3689)
3422 - rate limit CERROR/CWARN console message to avoid overload (4519)
3423 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3424 - basic mmap support (3918)
3425 - kernel patch series update from b1_4 (4711)
3427 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3430 - avoid crash during MDS cleanup with OST shut down (2775)
3431 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3432 - don't use bad inodes on the MDS (3744)
3433 - dynamic object preallocation to improve recovery speed (4236)
3434 - don't hold spinlock over lock dumping or change debug flags (4401)
3435 - don't zero obd_dev when it is force cleaned (3651)
3436 - print grants to console if they go negative (4431)
3437 - "lctl deactivate" will stop automatic recovery attempts (3406)
3438 - look for existing locks in ldlm_handle_enqueue() (3764)
3439 - don't resolve lock handle twice in recovery avoiding race (4401)
3440 - revalidate should check working dir is a directory (4134)
3442 - don't always mark "slow" obdfilter messages as errors (4418)
3444 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3447 - don't close LustreDB during write_conf until it is done (3860)
3448 - fix typo in lconf for_each_profile (3821)
3449 - allow dumping logs from multiple threads at one time (3820)
3450 - don't allow multiple threads in OSC recovery (3812)
3451 - fix debug_size parameters (3864)
3452 - fix mds_postrecov to initialize import for llog ctxt (3121)
3453 - replace config semaphore with spinlock (3306)
3454 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3455 - don't allow enqueue to complete on a destroyed export (3822)
3456 - down write_lock before checking llog header bitmap (3825)
3457 - recover from lock replay timeout (3764)
3458 - up llog sem before sending rpc (3652)
3459 - reduce ns lock hold times when setting kms (3267)
3460 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3461 - fix NULL deref and obd_dev leak on setup error (3312)
3462 - replace some LBUG about llog ops with error handling (3841)
3463 - don't match INVALID dentries from d_lookup and spin (3784)
3464 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3465 - fix invalid assertion in ptlrpc_set_wait (3880)
3467 - add libwrap support for the TCP acceptor (3996)
3468 - add /proc/sys/portals/routes for non-root route listing (3994)
3469 - allow setting MDS UUID in .xml (2580)
3470 - print the stack of a process that LBUGs (4228)
3472 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3475 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3476 - make sure to unset replay flag from failed open requests (3440)
3477 - if default stripe count is 0, use OST count for inode size (3636)
3478 - update parent mtime/ctime on client for create/unlink (2611)
3479 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3480 - free recovery state on server during a forced cleanup (3571)
3481 - unregister_reply for resent reqs (3063)
3482 - loop back devices mounting and status check on 2.6 (3563)
3483 - fix resource-creation race that can provoke i_size == 0 (3513)
3484 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3485 - more debugging for page-accounting assertion (3746)
3486 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3487 - avoid deadlock after precreation failure (3758)
3488 - fix race and lock order deadlock in orphan handling (3450, 3750)
3489 - add validity checks when grabbing inodes from l_ast_data (3599)
3491 - add /proc/.../recovery_status to obdfilter (3428)
3492 - lightweight CDEBUG infrastructure, debug daemon (3668)
3493 - change default OSC RPC parameters to be better on small clusters
3494 - turn off OST read cache for files smaller than 32MB
3495 - install man pages and include them in rpms (3100)
3496 - add new init script for (un)mounting lustre filesystems (2593)
3497 - run chkconfig in %post for init scripts (3701)
3498 - drop scimac NAL (unmaintained)
3500 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3503 - clean kiobufs before and after use (3485)
3504 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3505 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3506 - change init script to fail more clearly if not run as root (1528)
3507 - allow clients to reconnect during replay (1742)
3508 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3509 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3510 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3511 - fix cond_resched (3554)
3512 - only evict unfinished clients after recovery (3515)
3513 - allow bulk resend, prevent data loss (3570)
3514 - dynamic ptlrpc request buffer allocation (2102)
3515 - don't allow unlinking open directory if it isn't empty (2904)
3516 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3517 - remove extraneous obd dereference causing LASSERT failure (3334)
3518 - don't use get_cycles() when creating temp. files on the mds (3156)
3519 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3520 - handle EEXIST for set-stripe, set proper directory name (3336)
3522 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3523 - fix ksocknal_fmb_callback() error messages (2918)
3525 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3528 - don't copy lvb into (possibly NULL) reply on error (2983)
3529 - don't deref dentry after dput, don't free lvb on error (2922)
3530 - use the kms to determine writeback rpc length (2947)
3531 - increment oti_logcookies when osc is inactive (2948)
3532 - update client's i_blocks count via lvb messages (2543)
3533 - handle intent open/close of special files properly (1557)
3534 - mount MDS with errors=remount-ro, like obdfilter (2009)
3535 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3536 - don't use cancelling-locks' kms values (2947)
3537 - use highest lock extent for kms, not last one (2925)
3538 - don't dereference ERR_PTR() dentry in error handling path (3107)
3539 - fix thread race in portals_debug_dumplog() (3122)
3540 - create lprocfs device entries at setup instead of at attach (1519)
3541 - common AST error handler, don't evict client on completion race (3145)
3542 - zero nameidata in detach_mnt in 2.6 (3118)
3543 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3544 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3545 - initialize RPC timeout timer earlier for 2.6 (3219)
3546 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3547 - print nal/nid for unknown nid (3258)
3548 - additional checks for oscc recovery before doing precreate (3284)
3549 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3550 - don't crash in mdc_close for bad permissions on open (3285)
3551 - zero i_rdev for non-device files (3147)
3552 - clear page->private before handing to FS, better assertion (3119)
3553 - tune the read pipeline (3236)
3554 - fix incorrect decref of invalidated dentry (2350)
3555 - provide read-ahead stats and refine rpc in flight stats (3328)
3556 - don't hold journal transaction open across create RPC (3313)
3557 - update atime on MDS at close time (3265)
3558 - close LDAP connection when recovering to avoid server load (3315)
3559 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3560 - don't leak open file on MDS after open resend (3325)
3561 - serialize filter_precreate and filter_destroy_precreated (3329)
3562 - loop device shouldn't call sync_dev() for nul device (3092)
3563 - clear page cache after eviction (2766)
3564 - resynchronize MDS->OST in background (2824)
3565 - refuse to mount the same filesystem twice on same mountpoint (3394)
3566 - allow llmount to create routes for mounting behind routers (3320)
3567 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3568 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3569 - fix rare problem with rename on htree directories (3417)
3571 - allow default OST striping configuration per directory (1414)
3572 - fix compilation for qswnal for 2.6 kernels (3125)
3573 - increase maximum number of MDS request buffers for large systems
3574 - change liblustreapi to be useful for external progs like lfsck (3098)
3575 - increase local configuration timeout for slow disks (3353)
3576 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3578 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3581 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3582 - don't overwrite extent policy data in reply if lock was blocked (2901)
3583 - drop filter export grants atomically with removal from device (2663)
3584 - del obd_self_export from work_list in class_disconnect_exports (2908)
3585 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3586 - swab reply message in mdc_close, other PPC fixes (2464)
3587 - fix destroying of named logs (2325)
3588 - overwrite old logs when running lconf --write_conf (2264)
3589 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3590 - fix race in target_handle_connect (2898)
3591 - mds_reint_create() should take same inode create lock (2926)
3592 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3593 - don't close files for self_export to avoid uninitialized obd (2936)
3594 - allow MDS with the same name as client node (2939)
3595 - hold dentry reference for closed log files for unlink (2325)
3596 - reserve space for all logs during transactions (2059)
3597 - don't evict page beyond end of stripe extent (2925)
3598 - don't oops on a deleted current working directory (2399)
3599 - handle hard links to targets without a parent properly (2517)
3600 - don't dereference NULL lock when racing during eviction (2867)
3601 - don't grow lock extents when lots of conflicting locks (2919)
3603 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3606 - account for cache space usage on clients to avoid data loss (974)
3607 - lfsck support in lustre kernel code (2349)
3608 - reduce journal credits needed for BRW writes (2370)
3609 - orphan handling to avoid losing space on client/server crashes
3610 - ptlrpcd can be blocked, stopping ALL progress (2477)
3611 - use lock value blocks to assist in proper KMS, faster stat (1021)
3612 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3613 - recovery for initial connections (2355)
3614 - fixes for mds_cleanup_orphans (1934)
3615 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3616 - block all file creations until orphan recovery completes (1901)
3617 - client remove rq_connection from request struct (2423)
3618 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3619 - recovery timer race (2670)
3620 - mdc_close recovey bug (2532)
3621 - ptlrpc cleanup bug (2710)
3622 - mds timeout on local locks (2588)
3623 - namespace lock held during RPCs (2431)
3624 - handle interrupted sync write properly (2503)
3625 - don't try to handle a message that hasn't been replied to (2699)
3626 - client assert failure during cleanup after abort recovery (2701)
3627 - leak mdc device after failed mount (2712)
3628 - ptlrpc_check_set allows timedout requests to complete (2714)
3629 - wait for inflight reqs when ptlrpcd finishes (2710)
3630 - make sure unregistered services are removed from the srv_list
3631 - reset bulk XID's when resending them (caught by 1138 test)
3632 - unregister_bulk after timeout
3633 - fix lconf error (2694)
3634 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3635 - readahead locks pages, leaves pending causing memory pressure (2673)
3636 - increase OST request buffers to 4096 on large machines (2729)
3637 - fix up permission of existing directories in simple_mkdir (2661)
3638 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3639 - don't assign transno to errored transactions (2742)
3640 - don't delete objects on OST if given a bogus objid from MDS (2751)
3641 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3642 - if rq_no_resend, then timeout request after recovery (2432)
3643 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3644 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3645 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3646 - MDS recovery shouldn't delete objects that it has given out (2730)
3647 - if enqueue arrives after completion, don't clobber LVB (2819)
3648 - don't unlock pages twice when trigger_group_io returns error (2814)
3649 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3650 - don't write pages to disk if there was an error (1450)
3651 - don't ping imports that have recovery disabled (2676)
3652 - take buffered bytes into account when balancing socknal conn (2817)
3653 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3654 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3655 - remove little-endian swabbing of llog records (1987)
3656 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3657 - retry reposting request buffers if they fail (1191)
3658 - grow extent at grant time to avoid granting a revoked lock (2809)
3659 - lock revoke doesn't evict page if covered by a second lock (2765)
3660 - disable VM readahead to avoid reading outside lock extents (2805)
3662 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3663 - updated kernel patches for hp-2.4.20 kernel (2681)
3665 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3668 - fix truncated write corruption (2366)
3669 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3671 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3672 - ensure len(uuid) < 37 in lmc (1171)
3673 - fix ia64 OOPS in llog_test (2255)
3674 - zero end of page at obdfilter for partial page writes (2648)
3675 - don't leave stale dentries around after renames (bug 2428)
3676 - fix timeouts when evicting a client with a single lock held (2642)
3677 - set deadline for the initial HELLO message to drain (2634)
3678 - print out dotted-quad IP addresses in the socknal (2302)
3680 - additional debugging for MDS client eviction problem (2443)
3681 - fix mkfsoptions support for osts (2603, 2604)
3683 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3686 - add series for the vanilla 2.6.0 kernel
3687 - add series for the vanilla 2.4.24 kernel
3688 - add series for a cray x86/64 UL kernel drop
3689 - fix xattr patches for the vanilla 2.4.19 series
3691 - generate true UUIDs in lmc (1171)
3692 - have portals stack dumping break in UML (2466)
3693 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3694 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3695 - wake callback waiting threads on client eviction (2460)
3696 - Add --ptldebug and --subsystem to lmc (1719)
3697 - update assertion to allow safe interrupt allocation
3698 - set rq_no_resend for cancel requests (2432)
3699 - recalculate ptlrpcd timeout after resend (2494)
3700 - call vfs_rmdir when removing pending directories (2368)
3701 - fix renaming a file to itself (2429)
3702 - lmc creates a default one-stripe lov (2454)
3703 - expand procfs space to handle large clusters (2326)
3704 - increase UML stack to avoid overflow
3705 - update lconf's list of debug and subsystem masks
3706 - fix lfs find --obd (2510)
3707 - /proc tunable for disabling filter read caching (2591)
3708 - stop rpm packages from altering slapd.conf (2301)
3709 - disable nagle in the socknal under 0conf (2578)
3710 - choose mds inode size based on stripe count (2572)
3711 - fix kernel-source rpm problems (2516)
3713 - add --disable-doc to avoid pdf generation (2421)
3714 - update documentation, tests, type-os, comments
3715 - avoid format warnings on ia64
3716 - remove the TOE NAL
3717 - tiny code cleanups by removing unused fields
3719 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3722 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3723 - remove the most blatant lies from BUILDING file (2371)
3724 - change default debug level to reasonable production setting
3725 - reduce client side cache size to reduce cache flush time
3726 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3727 - make TCP zerocopy and pinger support enabled by default (2476)
3728 - sync writes completed after process exits caused crashes (2319)
3729 - maintain correct mount count on the MDS (2356)
3730 - backout 1557, because 2316 wasn't really fixed
3731 - better file I/O statistics gathering in /proc
3732 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3733 - another kernel patch to fix zero-copy TCP function export
3734 - don't take duplicate lock when processing re-sent getattr (2420)
3735 - lctl uses obd_self_export instead of creating new conn (2353)
3736 - MDS/OST recovery case which requires object creation asserted (2425)
3737 - move lfs from /usr/sbin to /usr/bin in packages
3738 - fix race between mds_client_add and mds_client_free (2417)
3739 - use kmalloc instead of slabs in portals (2430)
3740 - don't create duplicate records when a failover MDS is present (2442)
3741 - remove unnecessary mount age check (2332)
3742 - don't remove directory inodes from locks prematurely (2451)
3743 - don't break if MDS service name is the same as hostname (2103)
3744 - fix races in client write RPC generation when cache full (2482)
3746 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3749 - remove now-unused request->rq_obd (278)
3750 - if an allocation fails, print out how much memory we've used (1933)
3751 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3752 - add the "configurable stack size" patch to most series files (1256)
3753 - ability to write large log records, for 100+ OST configs (2306)
3754 - fix NULL deref when filter_prep fails (2314)
3755 - fix operator precedence error in filter_sync
3756 - dynamic allocation of socknal TX descriptors (2315)
3757 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3758 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3759 - ensure that CWARN messages in Portals always get to the syslog
3760 - __init/__exit are not for prototype decls (ldlm_init/exit)
3761 - x86-64 compile warning fixes
3762 - fix gateway LMC keyword conflict (2318)
3763 - fix MDS lock inversions in getattr/reint paths (1844)
3764 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3765 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3766 - DLM race condition prevented some lock evictions (2328)
3767 - ENOMEM detection and retry on socknal sends (2230)
3768 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3769 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3770 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3771 - better error messages when a client is rejected during recovery (1505)
3772 - avoid cancelling locks which were never granted, after failure (2330)
3773 - fix i_sem/journal inversion in mds_client_add (2333)
3774 - fix truncate/getattr lock cycle deadlock (2334)
3775 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3776 - fix two rare exit paths which could leak an l_lock() ref (2321)
3777 - fencepost error in MDS/OST orphan recovery (2226)
3778 - make log record alignment 8 bytes (1988)
3779 - lstripe now fails when requested offset > ost_count (2237)
3780 - ensure that all kernel series have a complete list.h (1607)
3781 - fix crashes in special-file operations (2316)
3782 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3784 - allow configurable automake binary, for testing new versions
3785 - small update to the lfs documentation
3787 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3789 * fix negative export reference count in fsfilt_sync (2312)
3791 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3792 * release candidate 0.9.1
3794 - orphans are moved into the PENDING directory for possible recovery
3795 - replayed opens now open by fid for orphan/rename safety (1042)
3796 - last close of an orphan inode generates a transno (683)
3797 - chdir() and mount() now pin the directory entry (1020)
3798 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3799 - discard very old requests without processing them (1502)
3800 - remove some common, well-understood CERRORs (1505)
3801 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3802 - clear "grant" flags in OST replies until OST grant code lands (1644)
3803 - fix read performance by not clobbering i_blksize on client (1598)
3804 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3805 - make LRU size a /proc tunable, clears locks when reduced (707)
3806 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3807 - server locks take references on exports now (1558)
3808 - build fixes for 2.4.20-rh trees (1663)
3809 - return an error from lov_create if all OSCs are inactive (1751)
3810 - fix import levels when a reconnect happens without a timeout (1597)
3811 - exit early from mds_open if we get a lookup error (1749)
3812 - partial page read at EOF wouldn't wait for disk before sending (1642)
3813 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3814 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3815 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3816 - write/truncate lock inversion (1639)
3817 - Don't auto-load obdclass, portals modules during cleanup (1495)
3818 - fix timestamps from jumping to "now" (1763)
3819 - extra journal assertions (1648)
3820 - add an extra multiunlink test (1771)
3821 - fix read_record/write_record API (1776)
3822 - fix leak of offset_extent, possible incorrect i_size later (1772)
3823 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3824 - Debugging for the kqswnal_get_idle_tx problems (1820)
3825 - Allow recovery to be attempted multiple times (1536)
3826 - Write out MDS last_rcvd file after it is first created (1600)
3827 - Fix tx_descriptor leak in failed transmit situations (1827)
3828 - ext3 journaling fixes for assertion failure after IO error (1871)
3829 - class_export_put() on freed export after completion AST error (1896)
3830 - Fix revalidate looping in VFS (1322)
3831 - Don't access a freed export during MDS_REINT timeout (1521)
3832 - Add open-unlink recovery support on the MDS (1673,1764)
3833 - Return an error if no MDS data was read from last_rcvd (1946)
3834 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3835 - Don't setup a disk device that doesn't match exported UUID (317)
3836 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3837 - avoid committing NULL handle in force close
3838 - local.sh is now a one-stripe LOV configuration
3839 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3840 - fix ext3 htree duplicate directory entry corruption (1516)
3841 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3842 - update to new LOV EA format (2097)
3843 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3845 2003-06-15 Phil Schwan <phil@clusterfs.com>
3848 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3849 - per-import/export recovery handling (958, 931, 959)
3850 - multiple last-rcvd slots, for serving multiple FSes (949)
3851 - connections are again shared between multiple imp/exports (963, 964)
3852 - "umount -f" would hang if any requests needed to be sent (393, 978)
3853 - avoid pinning large req buffer by copying for queued messages (989)
3854 - add "uuid" to "lctl device" command to help upcalls (991)
3855 - "open" RPCs with transnos would confuse recovery counters (1037)
3856 - do proper endian conversion of all wire messages (288, 340, 891)
3857 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3858 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3859 - fix ptlrpc_connection leak in target_handle_connect (1174)
3860 - fix import refcounting bug in OST and MDS cleanup (1134)
3861 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3862 - fix very unlikely obd_types race condition (501)
3863 - remove osc_open hack for echo_client (1187)
3864 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3865 - a failure in read_inode2 leads to deadlock (1139)
3866 - cancel ack-locks as soon as transaction is committed (1072)
3867 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3868 - make sure to commitrw after any preprw to avoid deadlock (1162)
3869 - failing to execute a file in a lustre FS would lock inode (1203)
3870 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3871 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3872 - obd_uuid2tgt would walk off the end of the list (1255)
3873 - on IA64 the osc would give portals incorrect bulk size (1258)
3874 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3875 - fix lock inversion caused by new llite matching code (1282)
3876 - limit the number of dirty pages on a client to 10MB (1286)
3877 - timed out locks were not being corrected cancelled (1289)
3878 - fix O_DIRECT above 4GB on IA-32 (1292)
3879 * major user-visible changes
3880 - fail out/fail over policy now controlled by the upcall (993)
3882 - add OBD_PING to check server availability and failure (954)
3883 - lustre messages are now sent in sending host order (288, 340, 891)
3884 - add eadatalen to MDS getattr reply (340)
3885 - OST read replies may contain second buffer, with per-page status (593)
3887 2003-03-11 Phil Schwan <phil@clusterfs.com>
3890 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3891 - failed open()s could cause deadlock; fixed (867, 869)
3892 - stop cancelling OST locks when files are closed (481)
3893 - overlapping XID spaces caused network corruption (851, 853)
3894 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3895 - setattr_raw added, to do single-RPC, server-side setattrs
3896 - lmc/lconf syntax change for OST UUIDs
3897 - fix crashy race condition between ptlrpc_free_req and osc_close
3898 - don't use request in mdc_enqueue if we hit a timeout (889)
3899 - don't set the inode i_size for regular files from the MDS (896)
3900 - handle out of order completion AST (842)
3901 - don't LBUG if a lock request times out after receiving AST (913)
3902 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3903 - if a bad lock AST arrives, send an error instead of dropping entirely
3904 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3905 - fix leak in bulk IO when only partially completed (899, 900, 926)
3906 - fix O_DIRECT for ia64 (55)
3907 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3908 - C-z after timeout could hang a process forever; fixed (977)
3910 - client-side I/O cache (678, 924, 929, 941, 970)
3912 - READPAGE and SETATTRs which don't take server-side locks get
3915 2003-02-11 Phil Schwan <phil@clusterfs.com>
3918 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3919 - Fix fsfilt_extN_readpage() to read a full page of directory
3920 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3921 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3922 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3923 - Most intent operations were removed, in favour of a new RPC mode
3924 that does a single RPC to the server and bypasses most of the VFS
3925 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3926 - Aggressively cancel local locks on DLM servers
3927 - mds_reint_unlink sends EA to the client if it's the last nlink.
3928 client uses that EA to unlink OST objects.
3929 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3930 - recursive symlinks were fixed (439)
3931 - fixed NULL deref in DEBUG_REQ
3932 - filter_update_lastobjid no longer calls sync, which annoyed extN
3933 - fixed multi-client small-writes to a single file problem (445)
3934 - fixed mtime updates during file writes (607)
3935 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3936 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3937 - fixed Linux OST size reporting problem (444, 656)
3938 - OST now updates object mtime with writes or setattr (607, 619)
3939 - client verifies file size before zeroing page past EOF (445)
3940 - OST now writes last allocated objid to disk with allocation (108)
3941 - LOV on echo now works (409)
3943 - mds_reint_unlink sends a new buffer, with the EA included. this
3944 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3945 set if a regular file was being unlinked, and it was the last link
3946 - use PtlGet from the target for bulk writes (315)
3947 - OST now updates object mtime with writes or setattr (607, 619)
3948 - LDLM now has a grant-time callback to revalidate locked items, if
3950 - Many MDS operations were reorganized to combat race conditions
3952 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3953 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3955 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3958 - Fully reactivate OST imports after reconnection (512, others)
3959 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3960 - More graceful error handling for truncating on dead OST (515)
3961 - Don't error out unless we're actually accessing dead stripes (474)
3962 - Fix garbage sizes when stripes are missing (410)
3963 - LRU counters were broken, causing constant lock purge (433, 432)
3964 - garbage on read from stripes with failed OSTs (441)
3965 - mark OSCs as active before reconnecting during recovery (438)
3966 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3967 - lfind did not preserve OST order in output (443)
3968 - symlinks cause hung clients, incorrect data (439)
3969 - stop dereferencing request after dropping refcount (457)
3970 - don't LASSERT(spin_is_locked) on non-SMP (455)
3971 - fixes for many rename() bugs
3972 - fstat didn't correctly synchronize attributes (399)
3973 - server must handle lock cancellation during blocking AST prep (487)
3974 - bulk descriptors were free()d too soon (511)
3975 - fix paths in lconf, which would load incorrect modules (451, 507)
3976 - fix confusing lconf 'host not found' error message (386)
3977 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3978 - fix race condition in mdc_blocking_ast() for inode access (526)
3979 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3980 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3981 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3982 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3983 - fix lstripe in several ways (kernel side) (527)
3984 - fix request leak in ldlm_cli_enqueue (262)
3985 - incorrect OSC was marked inactive after OST failure
3986 - call mds_fs_cleanup before unmounting filesystem (524)
3987 - fix races between taking ns_lock and ldlm_lock_change_resource
3988 - fix races updating LOV export open file list
3989 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3990 - fix recovery NULL deref in ldlm_cli_cancel_unused
3991 - fix some DLM races by using new hash table for lock handles (419)
3992 - permit the client to specify desired inodes, at replay
3993 - duplicate requests when we queue them for replay reintegration
3994 - fix last_rcvd offset calculation
3995 - sync after each recovered transaction, so we always make progress
3996 - never, not always, ERESTART requests without transnos
3997 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3999 - skip replay if the MDS says that the client is already connected
4000 - don't check for a recovery-enabled export to match lctl's UUID
4001 - don't INC_USE_COUNT for phantom exports
4002 - don't crash when cleaning up phantom exports (567)
4003 - don't double-finish or set replay data for errored mdc_open requests
4004 - abort requests when they time out, so we don't get old replies
4005 - send/receive replies for AST messages again
4006 - if the client says that it doesn't have the lock, cancel it on the
4008 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4009 mark it as destroyed, it will all work out when decref is called
4010 - fix module use counts (22, 581)
4012 - ASTs now expect a reply (server cancels lock on error reply)
4014 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4017 - fix many simultaneous client startup (392)
4018 - fix dentry->d_it clobbering
4019 - credentials weren't being shipped for readdir/getattr operations
4020 - remove invalid assertions triggered during some concurrent MD
4022 - proper Lustre versions added (336, 389)
4023 - fix memory leak for create error case (398)
4024 - fix LOV locking bug that would get cli/srv out of sync
4025 - fix echo client over LOV (409)
4026 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4027 - fix double-O_EXCL intent crash (424)
4028 - avoid sending multiple lock CANCELs (352)
4030 - MDS can do multi-client recovery (modulo bugs in new code)
4032 - many updates, edits, cleanups
4034 2002-11-18 Phil Schwan <phil@clusterfs.com>
4037 - fix null d_it dereference (346)
4038 - fix full OST/dbench hang (333)
4039 - fix permission problem with file removal (286)
4040 - fix removal of OSCs from LOV when they fail
4041 - fix NULL deref during bulk timeout (214)
4042 - fix problems related to multiple filesystems on one MDS (241)
4043 - fixed serious subtle metadata locking bugs
4044 - free locks on clients when inodes are removed due to memory
4046 - fix inode pointer in lock data (285)
4047 - partial support for multiple MDS on a single host (241)
4048 - data locks weren't cancelled at clear_inode time (290, 311)
4049 - intent locks could lead to unbounded lock growth (205)
4050 - added a maximum lock count, an LRU list, and a flusher
4051 - fix multiple rename (365)
4052 - properly abstracted the echo client
4053 - OSC locked 1 byte too many; fixed
4054 - rewrote brw callback code:
4055 - fixed recovery bugs related to LOVs (306)
4056 - fixed too-many-pages-in-one-write crash (191)
4057 - fixed (again) crash in sync_io_timeout (214)
4058 - probably fixed callback-related race (385)
4060 - Add capability to MDS protocol
4061 - LDLM cancellations and callbacks on different portals
4063 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4066 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4067 - defer open object destruction to close time (601981, 138)
4068 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4069 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4070 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4071 - don't call obd_setattr in ll_file_release for destroyed objects
4072 * protocol change to lustre_msg: move |version| and add |flags|
4073 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4074 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4075 * added replay of create, unlink, link and rename operations during
4076 MDS failover; recovery should be much more robust now
4077 * remove failed OSCs from LOVs (only lov_create uses this so far)
4078 * the lustre-HOWTO was brought (more) up to date (582544)
4080 2002-10-23 Phil Schwan <phil@clusterfs.com>
4083 - in-use dentries weren't being reused properly (617851)
4084 - prevent multiple LDLM setup (599178)
4085 - fix LOV size calculations for truncate (617853)
4086 - fix client handling of MDS intent errors (POSIX)
4087 - fix permission bug in lovstripe.c test (624321)
4088 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4089 - truncate past end of file could corrupt data
4090 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4091 - a race in recovery could return ETIMEDOUT to apps (623947)
4092 - building outside the source directory was fixed
4093 * the lustre-HOWTO was brought (more) up to date (582544)
4094 * major progress was made on recovery functionality
4096 2002-10-10 Phil Schwan <phil@clusterfs.com>
4099 - recovery deadlock fix
4100 - rm -rf causes LBUG fix (617817)
4101 - file open by multiple tasks fix (618962)
4102 - directory permissions bugs (602707 and 620007)
4103 - journal_stop fixed with locking (611313)
4104 - O_APPEND failures resolved (618273, perhaps 614459)
4105 - lconf PATH fix (619770)
4106 - IA64 build fix (621450)
4107 - RPC buffer sizes scale with amount of memory
4109 2002-10-01 Phil Schwan <phil@clusterfs.com>
4112 - locks would be cancelled without throwing away data pages,
4113 resulting in inconsistent data (605627)
4114 - inode attributes were not always being refreshed (605627, 612449)
4115 - lconf now continues to cleanup after lctl reports an error
4116 - MDS now enforces user permissions (602707)
4117 - lprocfs cleanup fixed, but not yet enabled (614157)
4118 - fixed infinite server hang, should a client not respond to an AST
4119 - avoid going into recovery if user calls readlink() with a buffer
4120 that's too small (613941)
4121 - AST RPCs no longer require replies (614867) -- this may be changed
4122 - don't crash server if client sends an IOV that's too big (611336)
4123 - fixed lock conversion deadlock (611892)
4124 - fixed the following of symlinks (614622)
4125 * recovery: the server can remove locks from a client that dies, other
4126 clients can make progress
4127 * more extN patch fixes
4128 * compile-time configurable ptlrpc buffer allocations
4130 - collaborative read cache document
4131 - Lustre Lite Performance CDR document-in-progress
4133 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4136 - fix typo in patch-2.4.18
4138 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4141 - clear ptlrpc request each time in handle_incoming_request()
4142 - unlink of files now destroys the object on the OST
4144 2002-09-19 Peter Braam <braam@clusterfs.com>
4146 * add hard link support
4147 * change obdfile creation method
4148 * kernel patch changed
4150 2002-09-19 Peter Braam <braam@clusterfs.com>
4153 - stack overflow bug in extN fixed
4155 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4157 * documentation updates
4158 - add man pages for config tools
4159 - update tests/README to describe testing with new config tools
4160 - finish metadata API descriptions
4161 * bug fixes and cleanups
4162 - statfs workaround for 16TB limit
4163 - LOV stripe allocation improved, can stripe on subset of OSTs
4164 - LOV file size/IO offset was wrong for files > 4GB in size
4165 - object EA data was being dropped, caused files to be unreadable
4166 - memory overflow with non-LOV OST caused memory corruption
4167 - fixed regression tests to work with new config tools, obdfilter
4168 - fixed bug when directory size became larger than 1 block
4169 - fixed bug (for single client case) when PWD was deleted
4170 - invalidate local directory pages when doing intent-based ops
4171 - avoid LDLM oops when lock callback contained bad data
4173 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4175 * documentation updates
4176 * bug fixes and cleanups
4177 - configuration tools
4180 - 64-bit compile warnings
4181 - 64-bit internal statfs data
4183 * test_brw on persistent OST devices
4185 * lprocfs (disabled)
4187 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4189 * documentation updates
4190 * bug fixes and cleanups
4191 * configuration tools
4193 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4196 * many small fixes to 0_5_4
4197 * io/network handling
4198 * thinkos in MDS operations
4200 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4203 * crucial basic fixes to 0.5.3
4204 * IOR, Iozone work over Elan
4207 2002-08-07 Phil Schwan <phil@clusterfs.com>
4208 * version 0_5_3, our first alpha
4209 * we use the new Portals iovs
4210 * documentation updates
4211 * bug fixes and cleanups
4212 * small changes in the DLM wire protocol
4214 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4215 * version 0_5_1 with some initial stability,
4216 * locking on MD and file I/O.
4217 * documentation updates
4218 * several bug fixes since 0.5.0
4219 * small changes in wire protocol
4221 2002-07-18 Phil Schwan <phil@clusterfs.com>
4223 * delivered as Lustre Light Alpha
4224 * fixed a crash after handling invalid MDS requests
4225 * fixed directory pages for architectures with non-4k pages sizes
4227 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4228 * release version v0_4_4
4229 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4231 * Fixes a number of interruption problems with OST operations.
4232 * Update documentation for portals header changes
4233 * Move all wire protocol structs/defines to lustre_idl.h
4234 * Fixes symlink length bug.
4235 * Add tcpdump to repository.
4237 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4238 * release version v0_4_3
4239 * Fixes statfs for inodes on extN.
4240 * Fixes bug in runtests which would delete /etc/hosts.
4241 * Use 64-bit object IDs wherever possible (not into VFS though)
4242 Remove ost_get_info, which is unused by lustre, and out of date.
4244 2002-07-03 Peter Braam <braam@clusterfs.com>
4245 * release version v0_4_2 Fixes a lookup error (type not passed)
4246 * move forward to head of Portals
4247 * move forward to latest Lustre kernel
4249 2002-06-25 Peter Braam <braam@clusterfs.com>
4250 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob