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.
41 Severity : enhancement
43 Description: Build SLES10 patchless client fails
44 Details : The configure was broken by run ./configure with
45 --with-linux-obj=.... argument for patchless client. When the
46 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
47 can't search header adequately. Use absolute path such as
48 -I($LINUX)/include instead.
52 Description: interrupt oig_wait produce painc on resend.
53 Details : brw_redo_request can be used for resend requests from ptlrpcd and
54 private set, and this produce situation when rq_ptlrpcd_data not
55 copyed to new allocated request and triggered LBUG on assert
56 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
57 ptlrpcd set if request is changed and can be safety changed to use
62 Description: LASSERT_{REQ,REP}SWAB macros are buggy
63 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
64 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
65 these macros and replace them with their difinitions instead.
69 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
70 Details : Update which_patch & target file for SLES10 latest kernel.
73 Description: Remove client patches from SLES 10 kernel.
74 Details : This causes SLES 10 clients to behave as patchless clients
75 even on a Lustre-patched (server) kernel.
77 Severity : enhancement
79 Description: self-adjustable client's lru lists
80 Details : use adaptive algorithm for managing client cached locks lru
81 lists according to current server load, other client's work
82 pattern, memory activities, etc. Both, server and client
83 side namespaces provide number of proc tunables for controlling
86 Severity : enhancement
88 Description: light-weight GSS support
89 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
90 and reduce performance overhead.
92 Severity : enhancement
94 Description: Linux keyring support
95 Details : Support using service of Linux keyring for Lustre GSS internal
96 context refresh/cache mechanism.
100 Description: Fix errors in lfs documentation
101 Details : Fixes man pages
105 Description: don't use GFP_* in generic Lustre code.
106 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
110 Description: obdclass is limited by single OBD_ALLOC(idarray)
111 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
115 Description: warnings with build patchless client with vanila 2.6.19 and up
116 Details : change old ctl_table style and replace ctl_table/ctl_table_header
117 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
121 Frequency : Only for SLES
122 Description: sanity_quota fail test_1
123 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
124 file, which makes TSTID[2] inunique.
128 Frequency : Only for SLES9
129 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
130 Details : Update target/ChangeLog/which_patch .
134 Frequency : Only for test_10 in sanity-quota.sh
135 Description: a bug in quota test code
136 Details : checking a test flag in a wrong place
140 Frequency : Only for test_13 in sanity-quota.sh
141 Description: a bug in quota test script
142 Details : shouldn't check overfull allocated quota
146 Description: jbd statistics
147 Details : Port older jbd statistics patch for sles10
149 Severity : enhancement
151 Description: Add printing inode info into message about error in writepage.
156 Description: reply_lock_interpret crash due to race with it and lock cancel.
157 Details : Do not replay locks that are being cancelled. Do not reference
158 locks by their address during replay, just by their handle.
162 Frequency : When flocks are used.
163 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
164 Details : Flock locks might destroy just granted lock if it could be merged
165 with another existing flock, this is done in completion handler,
166 so teach ldlm_cli_enquque_fini that this is a valid case for
172 Description: Oops in read and write path when failing to allocate lock.
173 Details : Check if lock allocation failed and return error back.
177 Description: lstripe command fails for valid OST index
178 Details : The stripe offset is compared to lov->desc.ld_tgt_count
179 instead of lov->desc.ld_active_tgt_count.
183 Description: sanity.sh failed test 103
184 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
189 Description: df doesn't work properly if diskfs blocksize != 4K
190 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
194 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
195 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
199 Description: obd_health_check_timeout too short
200 Details : set obd_health_check_timeout as 1.5x of obd_timeout
204 Description: llapi_file_create() does not allow some changes
205 Details : add llapi_file_open() that allows specifying the mode and
206 open flags, and also returns an open file handle.
210 Description: merge and cleanup kernel patches.
211 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
215 Description: Add journal checksum support.(Kernel part)
216 Details : The journal checksum feature adds two new flags i.e
217 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
218 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
219 indicates that the commit block contains the checksum for
220 the blocks described by the descriptor blocks. Now commit
221 record can be sent to disk without waiting for descriptor
222 blocks to be written to disk. This behavior is controlled
223 using JBD2_FEATURE_ASYNC_COMMIT flag.
227 Description: OSS needs mutliple precreate threads
228 Details : Add ability to start more than one create thread per OSS.
232 Description: Sanity.sh test_65k failed: setstripe should have succeeded
233 Details : execute lfs setstripe on client
237 Description: mds_obd_create error creating tmp object
238 Details : When the user sets quota on root, llog will be affected and can't
239 create files and write files.
242 Frequency : Always on ia64 patchless client, and possibly others.
244 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
245 Details : This allows the patchless client to be loaded on architectures
250 Description: RedHat Update kernel for RHEL5
251 Details : Add the kernel config file for RHEL5.
255 Description: RedHat Update kernel for RHEL5
256 Details : Modify the kernel config file more closer RHEL5.
260 Description: Build failure against Centos5 (RHEL5)
261 Details : Define PAGE_SIZE when it isn't present.
265 Description: test_7 on sanity-quota.sh can't be used on separate machines
266 Details : add facet to handle it
270 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
271 Details : with NFS, the anon dentry's parent was set to itself in
272 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
273 corresponding dentry other than use rec->ur_name.
275 Severity : enhancement
277 Description: lfs setstripe enhancement
278 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
282 Description: testing performance impact of enabling checksumming
283 Details : enable checksum by default, allow --disable-checksum
284 configure option and "-o nochecksum" mount option
288 Description: lustre support for RHEL5
289 Details : Add support for RHEL5.
293 Description: Client eviction due to failover config
294 Details : after a connection loss, the lustre client should attempt to
295 reconnect to the last active server first before trying the
296 other potential connections.
300 Description: when mds and osts use different quota unit(32bit and 64bit),
301 quota will be released repeatly.
302 Details : void sending multiple quota reqs to mds, which will keep the
303 status between the reqs.
307 Description: osts not allocated evenly to files
308 Details : change the condition to increase offset_idx
311 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
313 Description: Data corruption for OSTs that are formatted larger than 2TB
315 Details : When generating the bio request for lustre file writes the
316 sector number would overflow a temporary variable before being
317 used for the IO. The data reads correctly from Lustre (which
318 will overflow in a similar manner) but other file data or
319 filesystem metadata may be corrupted in some cases.
323 Description: Tune the kernel for good SCSI performance.
324 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
325 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
330 Description: rewrite ext2-derived code in obdclass/uuid.c
331 Details : rewrite inherited code (uuid parsing code from ext2 utils)
332 from scratch preserving functionality.
336 Description: "lfs find -obd UUID" prints directories
337 Details : "lfs find -obd UUID" will return all directory names instead
338 of just file names. It is incorrect because the directories
339 do not reside on the OSTs.
343 Description: use i_size_read and i_size_write in 2.6 port
344 Details : replace inode->i_size access with i_size_read/write()
347 Frequency : when removing large files
349 Description: scheduling issue during removal of large Lustre files
350 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
351 It causes scheduling issues when removing large files (17TB in the
355 Frequency : only with liblustre clients on XT3
357 Description: evictions taking too long
358 Details : allow llrd to evict clients directly on OSTs
361 Frequency : only on ppc
363 Description: /proc/fs/lustre/devices broken on ppc
364 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
365 We should make sure the type of 'index' is loff_t and then cast
366 explicitly as needed below. Do not assign an explicitly cast
370 Frequency : only for rhel5
372 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
373 Details : Modify the target file & which_kernel.
375 Severity : enhancement
377 Description: omit set fsid for export NFS
378 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
379 as FS_REQUIRES_DEV for avoid problems with generate fsid.
383 Frequency : Always, for kernels after 2.6.16
384 Description: Fix warning idr_remove called for id=.. which is not
386 Details : Last kernels save old s_dev before kill super and not allow
387 to restore from callback - restore it before call kill_anon_super.
391 Description: buffer overruns could theoretically occur
392 Details : llapi_semantic_traverse() modifies the "path" argument by
393 appending values to the end of the origin string, and a
394 overrun may occur. Adding buffer overrun check in liblustreapi.
398 Description: Fix error on 'ls .' at the top of the Lustre mount.
399 Details : Don't revalidate dentry if it is a root dentry.
403 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
404 Details : Modify vm-tunables-rhel4.patch.
408 Description: Kernel config for 2.6.18-vanilla.
409 Details : Modify targets/2.6-vanilla.target.in.
410 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
411 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
412 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
413 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
417 Description: improve handling recoverable errors
418 Details : if request processig with error which can be recoverable on server
419 request should be resend, otherwise page released from cache and
424 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
425 Details : Modify target file & which_patch.
426 A flaw was found in the IA32 system call emulation provided
427 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
428 value could be stored in the %RAX register, which could trigger an
429 out-of-bounds system call table access. An untrusted local user
430 could exploit this flaw to run code in the kernel
431 (ie a root privilege escalation). (CVE-2007-4573).
435 Description: O_DIRECT bypasses client statistics.
436 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
437 read_ahead_stats were updated. copied the stats section from
438 osc_send_oap_rpc() into async_internal().
442 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
446 Description: change order of libsysio includes
447 Details : '#include sysio.h' should always come before '#include xtio.h'
451 Description: handle "serious error: objid * already exists" more gracefully
452 Details : If LAST_ID value on disk is smaller than the objects existing in
453 the O/0/d* directories, it indicates disk corruption and causes an
454 LBUG(). If the object is 0-length, then we should use the existing
455 object. This will help to avoid a full fsck in most cases.
457 Severity : enhancement
459 Description: adapt the lustre_config script to support the upgrade case
460 Details : Add "-u" option for lustre_config script to support upgrading 1.4
461 server targets to 1.6 in parallel.
465 Description: To avoid grant space > avaible space when the disk is almost
466 full. Without this patch you might see the error "grant XXXX >
467 available" or some LBUG about grant, when the disk is almost
469 Details : In filter_check_grant, for non_grant cache write, we should
470 check the left space by if (*left > ungranted + bytes), instead
471 of (*left > ungranted), because only we are sure the left space
472 is enough for another "bytes", then the ungrant space should be
473 increase. In client, we should update cl_avail_grant only there
474 is OBD_MD_FLGRANT in the reply.
478 Description: Update RHEL 4 kernel to fix local root privilege escalation.
479 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
480 described in CVE-2007-4573. This problem could allow untrusted
481 local users to gain root access.
485 Frequency : when using O_DIRECT and quotas
487 Description: Incorrect file ownership on O_DIRECT output files
488 Details : block usage reported by 'lfs quota' does not take into account
489 files that have been written with O_DIRECT.
491 --------------------------------------------------------------------------------
493 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
495 * Support for kernels:
496 2.6.5-7.283 (SLES 9),
497 2.6.9-55.EL (RHEL 4),
498 2.6.16.46-0.14 (SLES 10),
499 2.6.18.8 vanilla (kernel.org)
500 * Client support for unpatched kernels:
501 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
502 2.6.16 - 2.6.22 vanilla (kernel.org)
503 * Due to recently discovered recovery problems, we do not recommend
504 using patchless RHEL 4 clients with this or any earlier release.
505 * Recommended e2fsprogs version: 1.39.cfs8
506 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
507 * Starting with this release, the ldiskfs backing filesystem required
508 by Lustre is now in its own package, lustre-ldiskfs. This package
509 should be installed. It is versioned separately from Lustre and
510 may be released separately in future.
514 Description: block reactivating mgc import until all deactivates complete
515 Details : Fix race when failing back MDT/MGS to itself (testing)
518 Frequency : at statup only
520 Description: mds_lov_synchronize race leads to various problems
521 Details : simultaneous MDT->OST connections at startup can cause the
522 sync to abort, leaving the OSC in a bad state.
524 Severity : enhancement
526 Description: add optional extra BUILD_VERSION info
527 Details : add a new environment variable (namely LUSTRE_VERS) which allows
528 to override the lustre version.
531 Frequency : 2.6.18 servers only
533 Description: ll_kern_mount() doesn't release the module reference
534 Details : The ldiskfs module reference count never drops down to 0
535 because ll_kern_mount() doesn't release the module reference.
540 Description: server LBUG when using old ost_num_threads parameter
541 Details : Accept the old ost_num_threads parameter but warn that it
542 is deprecated, and fix an off-by-one error that caused an LBUG.
547 Description: Transient SCSI error results in persistent IO issue
548 Details : iobuf->dr_error is not reinitialized to 0 between two
552 Frequency : sometimes when underlying device returns I/O errors
554 Description: OSTs not going read-only during write failures
555 Details : OSTs are not remounted read-only when the journal commit threads
556 get I/O errors because fsfilt_ext3 calls journal_start/stop()
557 instead of the ext3 wrappers.
561 Description: poor connect scaling with increasing client count
562 Details : Don't run filter_grant_sanity_check for more than 100 exports
563 to improve scaling for large numbers of clients.
566 Frequency : SLES10 only
568 Description: sanity-quota.sh quotacheck failed: rc = -22
569 Details : Quotas cannot be enabled on SLES10.
572 Frequency : liblustre clients only
574 Description: getdirentries does not give error when run on compute nodes
575 Details : getdirentries does not fail when the size specified as an argument
576 is too small to contain at least one entry
578 Severity : enhancement
580 Description: Add LNET router traceability for debug purposes
581 Details : If a checksum failure occurs with a router as part of the
582 IO path, the NID of the last router that forwarded the bulk data
583 is printed so it can be identified.
588 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
589 Details : Due to a race condition, liblustre clients were occasionally
592 Severity : enhancement
594 Description: lfs setstripe use optional parameters instead of postional
597 Severity : enhancement
599 Description: Nanosecond timestamp support for ldiskfs
600 Details : The on-disk ldiskfs filesystem has added support for nanosecond
601 resolution timestamps. There is not yet support for this at
602 the Lustre filesystem level.
605 Frequency : during server recovery
607 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
608 Details : request with rq_no_resend flag not awake l_wait_event if they get a
612 Frequency : nfs export on patchless client
614 Description: connectathon hang when test nfs export over patchless client
615 Details : Disconnected dentry cannot be found with lookup, so we do not need
616 to unhash it or make it invalid
619 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
620 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
621 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
626 Description: open req refcounting wrong on reconnect
627 Details : If reconnect happened between getting open reply from server and
628 call to mdc_set_replay_data in ll_file_open, we will schedule
629 replay for unreferenced request that we are about to free.
630 Subsequent close will crash in variety of ways.
631 Check that request is still eligible for replay in
632 mdc_set_replay_data().
637 Description: disable writes to filesystem when reading health_check file
638 Details : the default for reading the health_check proc file has changed
639 to NOT do a journal transaction and write to disk, because this
640 can cause reads of the /proc file to hang and block HA state
641 checking on a healthy but otherwise heavily loaded system. It
642 is possible to return to the previous behaviour during configure
643 with --enable-health-write.
645 Severity : enhancement
647 Description: 64-bit inode version
648 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
649 to the inode. This will be required for version-based recovery.
654 Description: MDS fails to start if a duplicate client export is detected
655 Details : in some rare cases it was possible for a client to connect to
656 an MDS multiple times. Upon recovery the MDS would detect this
657 and fail during startup. Handle this more gracefully.
659 Severity : enhancement
661 Description: Add -o localflock option to simulate old noflock
663 Details : This will achieve local-only flock/fcntl locks
669 Description: log_commit_thread vs filter_destroy race leads to crash
670 Details : Take import reference before releasing llog record semaphore
675 Description: Wrong request locking in request set processing
676 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
677 imp_delayed_list, in this place should be used imp_lock.
680 Frequency : when reconnection
682 Description: Grant Leak when osc reconnect to OST
683 Details : When osc reconnect ost, OST(filter) should check whether it
684 should grant more space to client by comparing fed_grant and
685 cl_avail_grant, and return the granted space to client instead
686 of "new granted" space, because client will call osc_init_grant
687 to update the client grant space info.
690 Frequency : when client reconnect to OST
692 Description: Grant Leak when osc do resend and replay bulk write
693 Details : When osc reconnect to OST, OST(filter)should clear grant info of
694 bulk write request, because the grant info while be sync between
695 OSC and OST when reconnect, and we should ignore the grant info
696 these of resend/replay write req.
701 Description: Grant space more than avaiable left space sometimes.
702 Details : When then OST is about to be full, if two bulk writing from
703 different clients came to OST. Accord the avaliable space of the
704 OST, the first req should be permitted, and the second one
705 should be denied by ENOSPC. But if the seconde arrived before
706 the first one is commited. The OST might wrongly permit second
707 writing, which will cause grant space > avaiable space.
710 Frequency : when client is evicted
712 Description: Grant might be wrongly erased when osc is evicted by OST
713 Details : when the import is evicted by server, it will fork another
714 thread ptlrpc_invalidate_import_thread to invalidate the
715 import, where the grant will be set to 0. While the original
716 thread will update the grant it got when connecting. So if
717 the former happened latter, the grant will be wrongly errased
718 because of this race.
723 Description: Checking Stale with correct fid
724 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
725 to check whether it is stale, because sometimes, we want the
726 enqueue happened anyway, and op_data.fid2 will not be initialized.
728 Severity : enhancement
730 Description: update patchless client
731 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
734 Frequency : only with 2.4 kernel
736 Description: random memory corruption
737 Details : size of struct ll_inode_info is to big for union inode.u and this
738 can be cause of random memory corruption.
743 Description: Memory leak in recovery
744 Details : Lov_mds_md was not free in an error handler in mds_create_object.
745 It should also check obd_fail before fsfilt_start, otherwise if
746 fsfilt_start return -EROFS,(failover mds during mds recovery).
747 then the req will return with repmsg->transno = 0 and rc = EROFS.
748 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
749 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
750 be freed no matter whether fsfilt_commit success or not.
753 Frequency : only with huge count clients
755 Description: Prevents from taking the superblock lock in llap_from_page for
757 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
758 for llap_from_page call in ll_removepage prevents from taking the
759 superblock lock for a soon died page.
764 Description: Not check open intent error before release open handle
765 Details : in some rare cases, the open intent error is not checked before
766 release open handle, which may cause
767 ASSERTION(open_req->rq_transno != 0), because it tries to release
768 the failed open handle.
773 Description: Set cat log bitmap only after create log success.
774 Details : in some rare cases, the cat log bitmap is set too early. and it
775 should be set only after create log success.
779 Description: Accessing a block bevice can re-enable I/O when Lustre is
780 tearing down a device.
781 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
786 Description: service threads may hog cpus when there are a lot of requests
788 Details : Insert cond_resched to give other threads a chance to use some of
794 Description: the cat log was not initialized in recovery
795 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
796 unlink log on mds will not be initialized until mds post
797 recovery. And also in mds post recovery, the unlink log will
798 initialization will be done asynchronausly, so there will be race
799 between add unlink log and unlink log initialization.
803 Description: brw_stats were being printed incorrectly
804 Details : brw_stats were being printed as log2 but all of them were not
805 recorded as log2. Also remove some code duplication arising from
806 filter_tally_{read,write}.
810 Frequency : rare, only in recovery.
811 Description: ASSERTION(req->rq_type != LI_POISON) failed
812 Details : imp_lock should be held while iterating over imp_sending_list for
813 prevent destroy request after get timeout in ptlrpc_queue_wait.
817 Description: replay-single.sh test 52 fails
818 Details : A lock's skiplist need to be cleanup when it being unlinked
819 from its resource list.
823 Description: Short directio read returns full requested size rather than
825 Details : Direct I/O operations should return actual amount of bytes
826 transferred rather than requested size.
828 Severity : enhancement
830 Description: metadata RPC reduction (e.g. for rm performance)
831 Details : decrease the amount of synchronous RPC between clients and servers
832 by canceling conflicing lock before the operation on the client side
833 and packing thier handles into the main operation RPC to server.
835 Severity : enhancement
837 Description: Async OSC create to avoid the blocking unnecessarily.
838 Details : If a OST has no remain object, system will block on the creating
839 when need to create a new object on this OST. Now, ways use
840 pre-created objects when available, instead of blocking on an
841 empty osc while others are not empty. If we must block, we block
842 for the shortest possible period of time.
844 --------------------------------------------------------------------------------
846 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
851 Frequency : on some architectures
853 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
854 Details : Uninitialized flags sometimes cause configuration commands to
858 Frequency : patchless clients only
860 Description: missing __iget() symbol export
861 Details : The __iget() symbol export is missing. To avoid the need for
862 this on patchless clients the deathrow inode reaper is turned
863 off, and we depend on the VM to clean up old inodes. This
864 dependency was during via the fix for bug 12181.
866 --------------------------------------------------------------------------------
868 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
870 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
871 INTEROPERATE with older versions automatically. In many cases a
872 special upgrade step is needed. Please read the
873 user documentation before upgrading any part of a live system.
874 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
875 version will not interoperate with 1.6 betas before beta5 (1.5.95).
876 * WARNING: Lustre configuration and startup changes are required with
877 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
883 Severity : enhancement
885 Description: MountConf
886 Details : Lustre configuration is now managed via mkfs and mount
887 commands instead of lmc and lconf. New obd types (MGS, MGC)
888 are added for dynamic configuration management. See
889 https://mail.clusterfs.com/wikis/lustre/MountConf for
892 Severity : enhancement
894 Description: dynamic OST addition
895 Details : OSTs can now be added to a live filesystem
897 Severity : enhancement
899 Description: startup order invariance
900 Details : MDTs and OSTs can be started in any order. Clients only
901 require the MDT to complete startup.
903 Severity : enhancement
905 Description: parallel, asynchronous orphan cleanup
906 Details : orphan cleanup is now performed in separate threads for each
907 OST, allowing parallel non-blocking operation.
909 Severity : enhancement
911 Description: optimized stripe assignment
912 Details : stripe assignments are now made based on ost space available,
913 ost previous usage, and OSS previous usage, in order to try
914 to optimize storage space and networking resources.
916 Severity : enhancement
918 Description: Permanently set tunables
919 Details : All writable /proc/fs/lustre tunables can now be permanently
920 set on a per-server basis, at mkfs time or on a live
923 Severity : enhancement
925 Description: Lustre message v2
926 Details : Add lustre message format v2.
928 Severity : enhancement
930 Description: client OST exclusion list
931 Details : Clients can be started with a list of OSTs that should be
932 declared "inactive" for known non-responsive OSTs.
936 Description: SPEC SFS validation failure on NFS v2 over lustre.
937 Details : Changes the blocksize for regular files to be 2x RPC size,
938 and not depend on stripe size.
940 Severity : enhancement
942 Description: Multiple MD RPCs in flight.
943 Details : Further unserialise some read-only MDS RPCs - learn about intents.
944 To avoid overly-overloading MDS, introduce a limit on number of
945 MDS RPCs in flight for a single client and add /proc controls
946 to adjust this limit.
948 Severity : enhancement
950 Description: client read/write statistics
951 Details : Add client read/write call usage stats for performance
952 analysis of user processes.
953 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
954 file access. extents_stats shows chunk size distribution.
955 extents_stats_per_process show chunk size distribution per
958 Severity : enhancement
960 Description: mds statistics
961 Details : Add detailed mds operations statistics in
962 /proc/fs/lustre/mds/*/stats.
966 Description: Failure of copying files with lustre special EAs.
967 Details : Client side always return success for setxattr call for lustre
968 special xattr (currently only "trusted.lov").
973 Description: Refcount LNET uuids
974 Details : The global LNET uuid list grew linearly with every startup;
975 refcount repeated list entries instead of always adding to
978 Severity : enhancement
980 Description: Dynamic service threads
981 Details : Within a small range, start extra service threads
982 automatically when the request queue builds up.
985 Frequency : mixed-endian client/server environments
987 Description: mixed-endian crashes
988 Details : The new msg_v2 system had some failures in mixed-endian
991 Severity : enhancement
993 Description: Easy OST removal
994 Details : OSTs can be permanently deactivated with e.g. 'lctl
995 conf_param lustre-OST0001.osc.active=0'
997 Severity : enhancement
999 Description: MGS proc entries
1000 Details : Added basic proc entries for the MGS showing what filesystems
1003 Severity : enhancement
1005 Description: provide MGS failover
1006 Details : Added config lock reacquisition after MGS server failover.
1008 Severity : enhancement
1010 Description: add Linux 2.4 support
1011 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1015 Description: a large application tries to do I/O to the same resource and dies
1016 in the middle of it.
1017 Details : Check the req->rq_arrival time after the call to
1018 ost_brw_lock_get(), but before we do anything about
1019 processing it & sending the BULK transfer request. This
1020 should help move old stale pending locks off the queue as
1021 quickly as obd_timeout.
1024 Frequency : when an incorrect nid is specified during startup
1026 Description: ptlrpc connect to non-existant node causes kernel crash
1027 Details : LNET can't be re-entered from an event callback, which
1028 happened when we expire a message after the export has been
1029 cleaned up. Instead, hand the zombie cleanup off to another
1032 Severity : enhancement
1034 Description: plain/inodebits lock performance improvement
1035 Details : Grouping plain/inodebits in granted list by their request modes
1036 and bits policy, thus improving the performance of search through
1040 Frequency : only if OST filesystem is corrupted
1042 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1043 Details : for a short time RPCs with bulk IO are in the replay list,
1044 but replay of bulk IOs is unimplemented. If the OST filesystem
1045 is corrupted due to disk cache incoherency and then replay is
1046 started it is possible to trip an assertion. Avoid putting
1047 committed RPCs into the replay list at all to avoid this issue.
1050 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1052 Description: System hang on startup
1053 Details : This bug allowed the liblustre (e.g. catamount) client to
1054 return to the app before handling all startup RPCs. This
1055 could leave the node unresponsive to lustre network traffic
1056 and manifested as a server ptllnd timeout.
1058 Severity : enhancement
1060 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1061 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1062 boolean to control whether to print peer debug info when a
1063 client's RPC times out.
1066 Frequency : only for kernels with patches from Lustre below 1.4.3
1068 Description: Remove old rdonly API
1069 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1072 Frequency : only for devices with external journals
1074 Description: Set external device read-only also
1075 Details : During a commanded failover stop, we set the disk device
1076 read-only while the server shuts down. We now also set any
1077 external journal device read-only at the same time.
1080 Frequency : when upgrading from 1.4 while trying to change parameters
1082 Description: The wrong (new) MDC name was used when setting parameters for
1083 upgraded MDT's. Also allows changing of OSC (and MDC)
1084 parameters if --writeconf is specified at tunefs upgrade time.
1087 Frequency : when setting specific ost indicies
1089 Description: QOS code breaks on skipped indicies
1090 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1091 created with --index need not be sequential.
1093 Severity : enhancement
1095 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1096 Details : The uninit_groups feature works in conjunction with the kernel
1097 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1098 up the pass1 processing of e2fsck. This is a read-only feature
1099 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1100 mount filesystems that have had this feature enabled.
1102 Severity : enhancement
1104 Description: Improve multi-block allocation algorithm to avoid fragmentation
1105 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1106 allocation locality and avoid filesystem fragmentation.
1108 ------------------------------------------------------------------------------
1110 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1112 * Support for kernels:
1113 2.4.21-47.0.1.EL (RHEL 3)
1114 2.6.5-7.283 (SLES 9)
1115 2.6.9-42.0.10.EL (RHEL 4)
1116 2.6.12.6 vanilla (kernel.org)
1117 2.6.16.27-0.9 (SLES 10)
1118 * Recommended e2fsprogs version: 1.39.cfs5
1120 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1124 Frequency : occasional, depends on client load and configuration
1125 Bugzilla : 12181, 12203
1126 Description: data loss for recently-modified files
1128 Details : In some cases it is possible that recently written or created
1129 files may not be written to disk in a timely manner (this should
1130 normally be within 30s unless client IO load is very high).
1131 The problem appears as zero-length files or files that are a
1132 multiple of 1MB in size after a client crash or client eviction
1133 that are missing data at the end of the file.
1135 This problem is more likely to be hit on clients where files are
1136 repeatedly created and unlinked in the same directory, clients
1137 have a large amount of RAM, have many CPUs, the filesystem has
1138 many OSTs, the clients are rebooted frequently, and/or the files
1139 are not accessed by other nodes after being written.
1141 The presence of the problem can be detected by looking at
1142 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1143 smaller than the second (nr_unused) then dirty files will not
1144 be flushed automatically to disk. "sync; sleep 10" should be
1145 run several times on the node before unmounting it to update
1146 Lustre (this is also safe to run on nodes without this problem).
1148 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1149 that can cause this same problem, so customers using that kernel
1150 also need to update the kernel in addition to Lustre. In order
1151 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1153 It is normal that files written just before a client crash (less
1154 than 30s) may not yet have been flushed to disk, even for local
1158 Frequency : frequent on thin XT3 nodes
1160 Description: UUID collision on thin XT3 Linux nodes
1161 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1162 randomly, since we relied on an insufficiently-seeded PRNG.
1167 Description: OSS hangs after "All ost request buffers busy"
1168 Details : A deadlock between quota and journal operations caused OSS
1169 hangs after printing "All ost request buffers busy."
1172 Frequency : always on liblustre builds
1174 Description: Cleanup compiler warnings on liblustre
1177 Frequency : always on liblustre builds on XT3
1179 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1183 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1184 > 32000 subdirectories
1185 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1186 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1187 subdirectory count crosses 32000. This will aid e2fsck to
1188 correctly handle more than 32000 subdirectories.
1191 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1193 Description: System hang on startup
1194 Details : This bug allowed the liblustre (e.g. catamount) client to
1195 return to the app before handling all startup RPCs. This
1196 could leave the node unresponsive to lustre network traffic
1197 and manifested as a server ptllnd timeout.
1199 Severity : enhancement
1201 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1202 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1203 boolean to control whether to print peer debug info when a
1204 client's RPC times out.
1209 Description: make O_SYNC working on 2.6 kernels
1210 Details : 2.6 kernels use different method for mark pages for write,
1211 so need add a code to lustre for O_SYNC work.
1216 Description: Failure to close file and release space on NFS
1217 Details : Put inode details into lock acquired in ll_intent_file_open.
1218 Use mdc_intent_lock in ll_intent_open to properly
1219 detect all kind of errors unhandled by mdc_enqueue.
1224 Description: proc file read during shutdown sometimes raced obd removal,
1226 Details : Add lock to prevent obd access after proc file removal.
1229 Frequency : Only for files larger than 4GB on 32-bit clients.
1231 Description: improperly doing page alignment of locks
1232 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1233 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1238 Description: RPCs being resent when they shouldn't be
1239 Details : Some RPCs that should not be resent are being resent. This
1240 can cause inconsistencies in the RPC state machine. Do not
1241 resend such requests.
1244 Frequency : rare, only with NFS export
1246 Description: Crash on NFS re-export node
1247 Details : under very unusual load conditions an assertion is hit in
1248 ll_intent_file_open()
1251 Frequency : only if OST filesystem is corrupted
1253 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1254 Details : for a short time RPCs with bulk IO are in the replay list,
1255 but replay of bulk IOs is unimplemented. If the OST filesystem
1256 is corrupted due to disk cache incoherency and then replay is
1257 started it is possible to trip an assertion. Avoid putting
1258 committed RPCs into the replay list at all to avoid this issue.
1263 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1264 Details : Large single O_DIRECT read and write calls can fail to allocate
1265 a sufficiently large buffer to process the request. In case of
1266 allocation failure the allocation is retried with a smaller
1267 buffer and broken into smaller requests.
1269 Severity : enhancement
1271 Description: Add -o localflock option to simulate old noflock behaviour.
1272 Details : This will achieve local-only flock/fcntl locks coherentness.
1277 Description: versioning check is incomplete
1278 Details : Checking the version difference of client vs. server, report
1279 error if the gap is too big.
1284 Description: add support PG_writeback bit
1285 Details : add support for PG_writeback bit for Lustre, for more carefull
1286 work with page cache in 2.6 kernel. This also fix some deadlocks
1287 and remove hack for work O_SYNC with 2.6 kernel.
1289 Severity : enhancement
1291 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1292 Details : The uninit_groups feature works in conjunction with the kernel
1293 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1294 up the pass1 processing of e2fsck. This is a read-only feature
1295 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1296 mount filesystems that have had this feature enabled.
1298 Severity : enhancement
1300 Description: Improve multi-block allocation algorithm to avoid fragmentation
1301 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1302 allocation locality and avoid filesystem fragmentation.
1304 ------------------------------------------------------------------------------
1306 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1308 * Support for kernels:
1309 2.6.9-42.0.3.EL (RHEL 4)
1310 2.6.5-7.276 (SLES 9)
1311 2.4.21-47.0.1.EL (RHEL 3)
1312 2.6.12.6 vanilla (kernel.org)
1313 2.6.16.21-0.8 (SLES10)
1314 * Recommended e2fsprogs version: 1.39.cfs2-0
1316 * The backwards-compatible /proc/sys/portals symlink has been removed
1317 in this release. Before upgrading, please ensure that you change
1318 any configuration scripts or /etc/sysctl.conf files that access
1319 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1320 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1321 in advance of the upgrade on any system running Lustre 1.4.6 or
1322 newer, since /proc/sys/lnet was added in that version.
1323 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1327 Frequency : only when quota is used
1329 Description: avoid scanning export list for quota master
1330 Details : Change the algorithms to avoid scanning export list in order
1331 to improve the efficiency.
1334 Frequency : MDS failover only, very rarely
1336 Description: "went back in time" messages on mds failover
1337 Details : The greatest transno may be lost when the current operation
1338 finishes with an error (transno==0) and the client's last_rcvd
1339 record is over-written. Save the greatest transno in the
1340 mds_last_transno for this case.
1343 Frequency : always for specific kernels and striping counts
1345 Description: client may get "Matching packet too big" without ACL support
1346 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1347 when trying to access files in certain configurations. The
1348 clients should in fact be denied when mounting because they do
1349 not understand ACLs.
1352 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1354 Description: many clients connecting with IO in progress causes connect timeouts
1355 Details : Avoid synchronous journal commits to avoid delays caused by many
1356 clients connecting/disconnecting when bulk IO is in progress.
1357 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1358 OST_IO_PORTAL to avoid delays behind potentially many pending
1362 Frequency : occasionally with multiple writers to a single file
1364 Description: shared writes to file may result in wrong size reported by stat()
1365 Details : Allow growing of kms when extent lock is cancelled
1368 Frequency : always with random mmap IO to multi-striped file
1370 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1371 Details : Only if the hole is at the end of OST object so that kms is too
1372 small. Fix is to increase kms accordingly in ll_nopage.
1375 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1377 Description: writes to a missing object would leak memory on the OST
1378 Details : If there is an inconsistency between the MDS and OST filesystems,
1379 such that the MDS references an object that doesn't exist, writes
1380 to that object will leak memory due to incorrect cleanup in the
1381 error handling path, eventually running out of memory on the OST.
1386 Description: Creating too long symlink causes lustre errors
1387 Details : Check symlink and name lengths before sending requests to MDS.
1390 Frequency : only if flock is enabled (not on by default)
1392 Description: posix locks not released on fd closure on 2.6.9+
1393 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1394 kernels, this caused such locks not to be released on fd close and
1395 then assertions on fs unmount about still used locks.
1398 Frequency : MDS failover only, very rarely
1400 Description: clients may get ASSERTION(granted_lock != NULL)
1401 Details : When request was taking a long time, and a client was resending
1402 a getattr by name lock request. The were multiple lock requests
1403 with the same client lock handle and
1404 mds_getattr_name->fixup_handle_for_resent_request found one of the
1405 lock handles but later failed with ASSERTION(granted_lock != NULL).
1410 Description: handle->h_buffer_credits > 0, assertion failure
1411 Details : h_buffer_credits is zero after truncate, causing assertion
1412 failure. This patch extends the transaction or creates a new
1416 Frequency : NFS re-export or patchless client
1417 Bugzilla : 11179, 10796
1418 Description: Crash on NFS re-export node (__d_move)
1419 Details : We do not want to hash the dentry if we don't have a lock.
1420 But if this dentry is later used in d_move, we'd hit uninitialised
1421 list head d_hash, so we just do this to init d_hash field but
1422 leave dentry unhashed.
1425 Frequency : NFS re-export or patchless client
1427 Description: NFS exports has problem with symbolic link
1428 Details : lustre client didn't properly install dentry when re-exported
1429 to NFS or running patchless client.
1432 Frequency : NFS re-export or patchless client
1434 Description: Various nfs/patchless fixes.
1435 Details : fixes reuse disconected alias for lookup process - this fixes
1436 warning "find_exported_dentry: npd != pd",
1437 fix permission error with open files at nfs.
1438 fix apply umask when do revalidate.
1441 Frequency : occasional
1443 Description: Crash on NFS re-export node
1444 Details : calling clear_page() on the wrong pointer triggered oops in
1445 generic_mapping_read().
1448 Frequency : rarely, using O_DIRECT IO
1450 Description: unaligned directio crashes client with LASSERT
1451 Details : check for unaligned buffers before trying any requests.
1454 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1456 Description: stale data returned from RAID cache
1457 Details : If only a small amount of IO is done to the RAID device before
1458 reading it again it is possible to get stale data from the RAID
1459 cache instead of reading it from disk.
1462 Frequency : always for sles10 kernel
1464 Description: sles10 support
1465 Details : ll_follow_link: compile fixes and using of nd_set_link
1466 under newer kernels.
1469 Frequency : depends on arch, kernel and compiler version, always on sles10
1472 Description: recursive or deep enough symlinks cause stack overflow
1473 Details : getting rid of large stack-allocated variable in
1477 Frequency : depends on hardware
1479 Description: lustre write performance loss in the SLES10 kernel
1480 Details : the performance loss is caused by using of write barriers in the
1481 ext3 code. The SLES10 kernel turns barrier support on by
1482 default. The fix is to undo that change for ldiskfs.
1484 ------------------------------------------------------------------------------
1486 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1488 * Support for kernels:
1489 2.6.9-42.0.3EL (RHEL 4)
1490 2.6.5-7.276 (SLES 9)
1491 2.4.21-47.0.1.EL (RHEL 3)
1492 2.6.12.6 vanilla (kernel.org)
1496 Frequency : quota enabled and large files being deleted
1498 Description: releasing more than 4GB of quota at once hangs OST
1499 Details : If a user deletes more than 4GB of files on a single OST it
1500 will cause the OST to spin in an infinite loop. Release
1501 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1506 Description: statfs data retrieved from /proc may be stale or zero
1507 Details : When reading per-device statfs data from /proc, in the
1508 {kbytes,files}_{total,free,avail} files, it may appear
1509 as zero or be out of date.
1512 Frequency : systems with MD RAID1 external journal devices
1514 Description: lconf's call to blkid is confused by RAID1 journal devices
1515 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1516 of returning all block devices that match the journal UUID.
1519 Frequency : always, for aggregate stripe size over 4GB
1521 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1522 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1523 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1526 Frequency : always if "lfs find" used on a local file/directory
1528 Description: "lfs find" segfaults if used on a local file/directory
1529 Details : The case where a directory component was not specified wasn't
1530 handled correctly. Handle this properly.
1533 Frequency : always on ppc64
1535 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1536 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1537 perform endianness conversion.
1540 Frequency : rarely (truncate to non-zero file size after write under load)
1541 Bugzilla : 10730, 10687
1542 Description: Files padded with zeros to next 4K multiple
1543 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1544 it is possible that files that are truncated to a non-zero size
1545 immediately after being written are filled with zero bytes beyond
1546 the truncated size. No file data is lost.
1548 Severity : enhancement
1550 Description: Allow recovery/failover for liblustre clients.
1551 Details : liblustre clients were unaware of failover configurations until
1554 Severity : enhancement
1556 Description: user file locks should fail when not mounting with flock option
1557 Details : Set up an error-returning stub in ll_file_operations.lock field
1558 to prevent incorrect behaviour when client is mounted without
1559 flock option. Also, set up properly f_op->flock field for
1563 Frequency : always on ia64
1565 Description: "lfs df" loops on printing out MDS statfs information
1566 Details : The obd_ioctl_data was not initialized and in some systems
1567 this caused a failure during the ioctl that did not return
1568 an error. Initialize the struct and return an error on failure.
1571 Frequency : SLES 9 only
1573 Description: Error of copying files with lustre special EAs as root
1574 Details : Client side always return success for setxattr call for lustre
1575 special xattr (currently only "trusted.lov").
1578 Frequency : rarely on clusters with both ia64+i386 clients
1580 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1581 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1582 clients are doing concurrent writes to the same file it is
1583 possible that the ia64 clients may LASSERT because the OST
1584 extent locks are not PAGE_SIZE aligned. Ensure that grown
1585 locks are always aligned on the request boundary.
1588 Frequency : specific use, occasional
1590 Description: Overwriting in use executable truncates on-disk binary image
1591 Details : If one node attempts to overwrite an executable in use by
1592 another node, we now correctly return ETXTBSY instead of
1593 truncating the file.
1595 Severity : enhancement
1597 Description: Async OSC create to avoid the blocking unnecessarily.
1598 Details : If a OST has no remain object, system will block on the creating
1599 when need to create a new object on this OST. Now, ways use
1600 pre-created objects when available, instead of blocking on an
1601 empty osc while others are not empty. If we must block, we block
1602 for the shortest possible period of time.
1607 Description: chmod on Lustre root is propagated to other clients
1608 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1609 invalid by the follow_mount time.
1612 Frequency : liblustre clients only
1614 Description: Race in 'instant cancel' lock handling could lead to such locks
1615 never to be granted in case of SMP MDS
1616 Details : Do not destroy not yet granted but cbpending locks in
1620 Frequency : replay/resend of open
1622 Description: non null lock assetion failure in mds_intent_policy
1623 Details : Trying to replay/resend lockless open requests resulted in
1624 mds_open() returning 0 with no lock. Now it sets a flag if
1625 a lock is going to be returned.
1627 Severity : enhancement
1629 Description: Checksum enhancements
1630 Details : New checksum enhancements allow for resending RPCs that failed
1633 Severity : enhancement
1635 Description: Tunables on number of dirty pages in cacche
1636 Details : Allow to set limit on number of dirty pages cached.
1641 Description: client crash on unmount - lock still has references
1642 Details : In some error handling cases it was possible to leak a lock
1643 reference on a client while accessing a file. This was not
1644 harmful to the client during operation, but would cause the
1645 client to crash when the filesystem is unmounted.
1648 Frequency : specific case, rare
1650 Description: ETXTBSY on mds though file not in use by client
1651 Details : ETXTBSY is no longer incorrectly returned when attempting to
1652 chmod or chown a directory that the user previously tried to
1653 execute or a currently-executing binary.
1656 Frequency : extremely rare except on liblustre-based clients
1658 Description: Lustre space not freed when files are deleted
1659 Details : Clean up open-unlinked files after client eviction. Previously
1660 the unlink was skipped and the files remained as orphans.
1665 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1666 Details : In some cases it was possible to send a blocking callback to
1667 a client doing a glimpse, even though that client didn't get
1668 a lock granted. When the glimpse lock is cancelled on the OST
1669 the freed lock is left on the waiting list and corrupted the list.
1672 Frequency : all core dumps
1674 Description: Broke core dumps to lustre
1675 Details : Negative dentry may be unhashed if parent does not have UPDATE
1676 lock, but some callers, e.g. do_coredump, expect dentry to be
1677 hashed after successful create, hash it in ll_create_it.
1679 ------------------------------------------------------------------------------
1681 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1683 * Support for kernels:
1684 2.6.9-42.0.2.EL (RHEL 4)
1685 2.6.5-7.276 (SLES 9)
1686 2.4.21-40.EL (RHEL 3)
1687 2.6.12.6 vanilla (kernel.org)
1691 Frequency : always on RHEL 3
1693 Description: Number of open files grows over time
1694 Details : The number of open files grows over time, whether or not
1695 Lustre is started. This was due to a filp leak introduced
1696 by one of our kernel patches.
1698 ------------------------------------------------------------------------------
1700 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1702 * Support for kernels:
1703 2.6.9-42.EL (RHEL 4)
1704 2.6.5-7.276 (SLES 9)
1705 2.4.21-40.EL (RHEL 3)
1706 2.6.12.6 vanilla (kernel.org)
1711 Bugzilla : 5719, 9635, 9792, 9684
1712 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1713 Details : If a server is under heavy load and cannot reply to new
1714 connection requests before the client resends the (re)connect,
1715 the connection handling code can behave badly if two service
1716 threads are concurrently handing separate (re)connections from
1717 the same client. Add better locking to the connection handling
1718 code, and ensure that only a single connection will be processed
1719 for a given client UUID, even if the lock is dropped.
1721 Severity : enhancement
1723 Description: add TCP zero-copy support to kernel
1724 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1725 sends if the hardware supports scatter-gather and checksumming.
1726 This allows socklnd to do client-write and server-read more
1727 efficiently and reduce CPU utilization from skbuf copying.
1730 Frequency : only if NFS exporting from client
1732 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1733 Details : If setting attributes on a file created under NFS that had
1734 never been opened it would be possible to oops the client
1735 if the file had no objects.
1738 Frequency : always for liblustre
1740 Description: liblustre client does MDS+OSTs setattr RPC for each write
1741 Details : When doing a write from a liblustre client, the client
1742 incorrectly issued an RPC to the MDS and each OST the file was
1743 striped over in order to update the timestamps. When writing
1744 with small chunks and many clients this could overwhelm the MDS
1745 with RPCs. In all cases it would slow down the write because
1746 these RPCs are unnecessary.
1748 Severity : enhancement
1750 Description: allow number of MDS service threads to be changed at module load
1751 Details : It is now possible to change the number of MDS service threads
1752 running. Adding "options mds mds_num_threads={N}" to the MDS's
1753 /etc/modprobe.conf will set the number of threads for the next
1754 time Lustre is restarted (assuming the "mds" module is also
1755 reloaded at that time). The default number of threads will
1756 stay the same, 32 for most systems.
1761 Description: OST crash if filesystem is unformatted or corrupt
1762 Details : If an OST is started on a device that has never been formatted
1763 or if the filesystem is corrupt and cannot even mount then the
1764 error handling cleanup routines would dereference a NULL pointer.
1769 Description: NULL pointer deref in llap_from_page.
1770 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1771 incorrect) mapping if the page was truncated/reclaimed while it was
1772 searched for. Check for this condition and skip such pages when
1773 doing readahead. Introduce extra check to llap_from_page() to
1774 verify page->mapping->host is non-NULL (so page is not anonymous).
1777 Frequency : Sometimes when using sys_sendfile
1779 Description: "page not covered by a lock" warnings from ll_readpage
1780 Details : sendfile called ll_readpage without right page locks present.
1781 Now we introduced ll_file_sendfile that does necessary locking
1782 around call to generic_file_sendfile() much like we do in
1786 Frequency : with certain MDS communication failures at client mount time
1788 Description: NULL pointer deref after failed client mount
1789 Details : a client connection request may delayed by the network layer
1790 and not be sent until after the PTLRPC layer has timed out the
1791 request. If the client fails the mount immediately it will try
1792 to clean up before the network times out the request. Add a
1793 reference from the request import to the obd device and delay
1794 the cleanup until the network drops the request.
1797 Frequency : occasionally during client (re)connect
1799 Description: assertion failure during client (re)connect
1800 Details : processing a client connection request may be delayed by the
1801 client or server longer than the client connect timeout. This
1802 causes the client to resend the connection request. If the
1803 original connection request is replied in this interval, the
1804 client may trip an assertion failure in ptlrpc_connect_interpret()
1805 which thought it would be the only running connect process.
1808 Frequency : only with obd_echo servers and clients that are rebooted
1810 Description: kernel BUG accessing uninitialized data structure
1811 Details : When running an obd_echo server it did not start the ping_evictor
1812 thread, and when a client was evicted an uninitialized data
1813 structure was accessed. Start the ping_evictor in the RPC
1814 service startup instead of the OBD startup.
1816 Severity : enhancement
1817 Bugzilla : 10193 (patchless)
1818 Description: Remove dependency on various unexported kernel interfaces.
1819 Details : No longer need reparent_to_init, exit_mm, exit_files,
1820 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1823 Frequency : rare (only users of deprecated and unsupported LDAP config)
1825 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1826 Details : LDAP apparently contains 'lustreName' attributes instead of
1827 'name'. A simple remapping of the name is sufficient.
1830 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1832 Description: waiting_locks_callback trips kernel BUG if client is evicted
1833 Details : Running with the dump_on_timeout debug flag turned on makes
1834 it possible that the waiting_locks_callback() can try to dump
1835 the Lustre kernel debug logs from an interrupt handler. Defer
1836 this log dumping to the expired_lock_main() thread.
1838 Severity : enhancement
1840 Description: Support NFS exporting on 2.6 kernels.
1841 Details : Implement non-rawops metadata methods for NFS server to use without
1842 changing NFS server code.
1845 Frequency : very rare (synthetic metadata workload only)
1847 Description: two racing renames might cause an MDS thread to deadlock
1848 Details : Running the "racer" program may cause one MDS thread to rename
1849 a file from being the source of a rename to being the target of
1850 a rename at exactly the same time that another thread is doing
1851 so, and the second thread has already enqueued these locks after
1852 doing a lookup of the target and is trying to relock them in
1853 order. Ensure that we don't try to re-lock the same resource.
1856 Frequency : only very large systems with liblustre clients
1858 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1859 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1860 all OSTs in parallel. This allows the eviction of stale liblustre
1861 clients to proceed much faster than if they were done in series,
1862 and also offers similar improvements for other set_info RPCs.
1867 Description: excessive CPU usage during initial read phase on client
1868 Details : During the initial read phase on a client, it would agressively
1869 retry readahead on the file, consuming too much CPU and impacting
1870 performance (since 1.4.5.8). Improve the readahead algorithm
1871 to avoid this, and also improve some other common cases (read
1872 of small files in particular, where "small" is files smaller than
1873 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1878 Description: MDS crash when receiving packet with unknown intent.
1879 Details : Do not LBUG in unknown intent case, just return -EFAULT
1881 Severity : enhancement
1882 Bugzilla : 9293, 9385
1883 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1884 Details : Do not serialize getattr (non-intent version) and statfs.
1887 Frequency : occasional, when OST network is overloaded/intermittent
1889 Description: client evicted by OST after bulk IO timeout
1890 Details : If a client sends a bulk IO request (read or write) the OST
1891 may evict the client if it is unresposive to its data GET/PUT
1892 request. This is incorrect if the network is overloaded (takes
1893 too long to transfer the RPC data) or dropped the OST GET/PUT
1894 request. There is no need to evict the client at all, since
1895 the pinger and/or lock callbacks will handle this, and the
1896 client can restart the bulk request.
1899 Frequency : Always when mmapping file with no objects
1901 Description: client crashes when mmapping file with no objects
1902 Details : Check that we actually have objects in a file before doing any
1903 operations on objects in ll_vm_open, ll_vm_close and
1909 Description: Request leak when working with deleted CWD
1910 Details : Introduce advanced request refcount tracking for requests
1911 referenced from lustre intent.
1913 Severity : Enhancement
1915 Description: Cache open file handles on client.
1916 Details : MDS now will return special lock along with openhandle, if
1917 requested and client is allowed to hold openhandle, even if unused,
1918 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1919 closing files for every read/write openration.
1921 Severity : Enhancement
1923 Description: Cache open negative dentries on client when possible.
1924 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1925 negative dentries on lock revocation.
1930 Description: Remounting a client read-only wasn't possible with a zconf mount
1931 Details : It wasn't possible to remount a client read-only with llmount.
1933 Severity : enhancement
1934 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1935 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1936 MPICH is included to add Lustre-specific ADIO interfaces.
1937 This is based closely on the UFS ADIO layer and only differs
1938 in file creation, in order to allow the OST striping to be set.
1939 This is user-contributed code and not supported by CFS.
1944 Description: extended inode attributes (immutable, append-only) work improperly
1945 when 2.4 and 2.6 kernels are used on client/server or vice versa
1946 Details : Introduce kernel-independent values for these flags.
1948 Severity : enhancement
1951 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1952 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1953 granularity. Now, Fractional MB granularity is be supported,
1954 this is very useful for low memory system.
1956 Severity : enhancement
1958 Description: Getattr by fid
1959 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1960 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1961 have locks and avoids getattr rpc requests for it on every stat.
1964 Frequency : Always, for filesystems larger than 2TB
1966 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1967 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1968 filesystems larger than 2TB to be mounted with mballoc enabled.
1971 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1973 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1974 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1975 filesystems larger than 2TB is mounted then the kernel will
1976 silently corrupt the start of the filesystem. CONFIG_LBD is
1977 enabled for all CFS-supported kernels, but the possibility of
1978 this happening with a modified kernel config exists.
1980 Severity : enhancement
1982 Description: add client O_DIRECT support for 2.6 kernels
1983 Details : It is now possible to do O_DIRECT reads and writes to files
1984 in the Lustre client mountpoint on 2.6 kernel clients.
1986 Severity : enhancement
1988 Description: parallel glimpse, setattr, statfs, punch, destroy requests
1989 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
1990 parallel, not waiting for response from every OST before sending
1991 a rpc to the next OST.
1996 Description: setattr vs write race when updating file timestamps
1997 Details : Client processes that update a file timestamp into the past
1998 right after writing to the file (e.g. tar) it is possible that
1999 the updated file modification time can be reset to the current
2000 time due to a race between processing the setattr and write RPC.
2002 Severity : enhancement
2004 Description: Bring 'lfs find' closer in line with regular Linux find.
2005 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2006 -print0 options and obtains all the needed info through the lustre
2009 Severity : enhancement
2011 Description: support up to 1024 configured devices on one node
2012 Details : change obd_dev array from statically allocated to dynamically
2013 allocated structs as they are first used to reduce memory usage
2018 Description: Flush dirty partially truncated pages during truncate
2019 Details : Immediatelly flush partially truncated pages in filter_setattr,
2020 this way we completely avoid having any pages in page cache on OST
2021 and can retire ugly workarounds during writes to flush such pages.
2026 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2027 Details : protect inode from truncation within vfs_unlink() context
2028 just take a reference before calling vfs_unlink() and release it
2029 when parent's i_sem is free.
2034 Description: last_id value checked outside lock on OST caused LASSERT failure
2035 Details : If there were multiple MDS->OST object precreate requests in
2036 flight, it was possible that the OST's last object id was checked
2037 outside a lock and incorrectly tripped an assertion. Move checks
2038 inside locks, and discard old precreate requests.
2041 Frequency : always, if extents are used on OSTs
2043 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2044 Details : OSTs using the extents format would not zero the high 16 bits of
2045 the index physical block number. This is not a problem for any
2046 OST filesystems smaller than 16TB, and no kernels support ext3
2047 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2048 new/modified files) and can be fixed for existing filesystems
2049 with e2fsprogs-1.39-cfs1.
2054 Description: import connection selection may be incorrect if timer wraps
2055 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2056 import connections to be ignored if the 32-bit jiffies counter
2057 wraps. Use a 64-bit jiffies counter.
2060 Frequency : very large clusters immediately after boot
2062 Description: LNET request buffers exhausted under heavy short-term load
2063 Details : If a large number of client requests are generated on a service
2064 that has previously never seen so many requests it is possible
2065 that the request buffer growth cannot keep up with the spike in
2066 demand. Instead of dropping incoming requests, they are held in
2067 the LND until the RPC service can accept more requests.
2070 Frequency : Sometimes during replay
2072 Description: Assertion failure in ll_local_open after replay.
2073 Details : If replay happened on an open request reply before we were able
2074 to set replay handler, reply will become not swabbed tripping the
2075 assertion in ll_local_open. Now we set the handler right after
2076 recognising of open request
2079 Frequency : very rare
2081 Description: kernel reports "badness in vsnprintf"
2082 Details : Reading from the "recovery_status" /proc file in small chunks
2083 may cause a negative length in lprocfs_obd_rd_recovery_status()
2084 call to vsnprintf() (which is otherwise harmless). Exit early
2085 if there is no more space in the output buffer.
2087 Severity : enhancement
2089 Description: clear OBD RPC statistics by writing to them
2090 Details : It is now possible to clear the OBD RPC statistics by writing
2091 to the "stats" file.
2096 Description: Client mtime is not the same on different clients after utimes
2097 Details : In some cases, the client was using the utimes() syscall on
2098 a file cached on another node. The clients now validate the
2099 ctime from the MDS + OSTs to determine which one is right.
2104 Description: Inability to activate failout mode
2105 Details : lconf script incorrectly assumed that in pythong string's numeric
2106 value is used in comparisons.
2109 Frequency : always with multiple stripes per file
2111 Description: Inefficient object allocation for mutli-stripe files
2112 Details : When selecting which OSTs to stripe files over, for files with
2113 a stripe count that divides evenly into the number of OSTs,
2114 the MDS is always picking the same starting OST for each file.
2115 Return the OST selection heuristic to the original design.
2120 Description: mount failures may take full timeout to return an error
2121 Details : Under some heavy load conditions it is possible that a
2122 failed mount can wait for the full obd_timeout interval,
2123 possibly several minutes, before reporting an error.
2124 Instead return an error as soon as the status is known.
2126 Frequency : quota enabled and large files being deleted
2128 Description: releasing more than 4GB of quota at once hangs OST
2129 Details : If a user deletes more than 4GB of files on a single OST it
2130 will cause the OST to spin in an infinite loop. Release
2131 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2136 Description: statfs data retrieved from /proc may be stale or zero
2137 Details : When reading per-device statfs data from /proc, in the
2138 {kbytes,files}_{total,free,avail} files, it may appear
2139 as zero or be out of date.
2142 Frequency : systems with MD RAID1 external journal devices
2144 Description: lconf's call to blkid is confused by RAID1 journal devices
2145 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2146 of returning all block devices that match the journal UUID.
2149 Frequency : always, for aggregate stripe size over 4GB
2151 Description: assertion fails when trying to use 4GB stripe size
2152 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2153 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2156 Frequency : always on ppc64
2158 Description: the first write on an ext3 filesystem with mballoc got stuck
2159 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2160 perform endianness conversion.
2162 ------------------------------------------------------------------------------
2164 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2166 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2167 INTEROPERATE with older versions automatically. Please read the
2168 user documentation before upgrading any part of a live system.
2169 * WARNING: Lustre networking configuration changes are required with
2170 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2173 * Support for kernels:
2174 2.6.9-22.0.2.EL (RHEL 4)
2175 2.6.5-7.244 (SLES 9)
2176 2.6.12.6 vanilla (kernel.org)
2179 Severity : enhancement
2180 Bugzilla : 7981/8208
2181 Description: Introduced Lustre Networking (LNET)
2182 Details : LNET is new networking infrastructure for Lustre, it includes
2183 a reorganized network configuration mode (see the user
2184 documentation for full details) as well as support for routing
2185 between different network fabrics. Lustre Networking Devices
2186 (LNDs) for the supported network fabrics have also been
2187 created for this new infrastructure.
2189 Severity : enhancement
2190 Description: Introduced Access control lists
2191 Details : clients can set ACLs on files and directories in order to have
2192 more fine-grained permissions than the standard Unix UGO+RWX.
2193 The MDS must be started with the "-o acl" mount option.
2195 Severity : enhancement
2196 Description: Introduced filesystem quotas
2197 Details : Administrators may now establish per-user quotas on the
2200 Severity : enhancement
2202 Description: Configuration change for the XT3
2203 The PTLLND is now used to run Lustre over Portals on the XT3
2204 The configure option(s) --with-cray-portals are no longer used.
2205 Rather --with-portals=<path-to-portals-includes> is used to
2206 enable building on the XT3. In addition to enable XT3 specific
2207 features the option --enable-cray-xt3 must be used.
2212 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2213 Details : A race between allocating a new llcd and re-getting the llcd_lock
2214 allowed another thread to grab newly-allocated llcd.
2216 Severity : enhancement
2218 Description: 2.6 OST async journal commit and locking fix to improve performance
2219 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2220 2.6 kernels are now async as they already were in 2.4 kernels so
2221 that they can commit concurrently with the network bulk transfer.
2222 For block-allocated files the filter allocation semaphore is held
2223 to avoid filesystem fragmentation during allocation. BKL lock
2224 removed for 2.6 xattr operations where it is no longer needed.
2229 Description: lconf incorrectly determined whether two IP networks could talk
2230 Details : In some more complicated routing and multiple-network
2231 configurations, lconf will avoid trying to make a network
2232 connection to a disjoint part of the IP space. It was doing the
2233 math incorrectly for one set of cases.
2238 Description: Fix for potential infinite loop processing records in an llog.
2239 Details : If an llog record is corrupted/zeroed, it is possible to loop
2240 forever in llog_process(). Validate the llog record length
2241 and skip the remainder of the block on error.
2244 Frequency : occasional (liblustre only)
2246 Description: liblustre could not open files whose last component is a symlink
2247 Details : sysio_path_walk() would incorrectly pass the open intent to
2248 intermediate path components.
2251 Frequency : rare (liblustre only with non-standard tuning)
2252 Bugzilla : 7201 (7350)
2253 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2254 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2255 after it has been released, only do this for referenced locks
2257 Severity : enhancement
2259 Description: specifying an (invalid) directory default stripe_size of -1
2260 would reset the directory default striping
2261 Details : stripe_size -1 was used internally to signal directory stripe
2262 removal, now use "all default" to signal dir stripe removal
2263 as a directory striping of "all default" is not useful
2266 Frequency : common for large clusters running liblustre clients
2268 Description: doing an ls when liblustre clients are running is slow
2269 Details : sending a glimpse AST to a liblustre client waits for every AST
2270 to time out, as liblustre clients will not respond. Since they
2271 cannot cache data we refresh the OST lock LVB from disk instead.
2273 Severity : enhancement
2275 Description: doing an ls at the same time as file IO can be slow
2276 Details : enqueue and other "small" requests can be blocked behind many
2277 large IO requests. Create a new OST IO portal for non-IO
2278 requests so they can be processed faster.
2281 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2283 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2284 created file with mode 000
2285 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2286 allow an ftruncate() as a non-root user to a file with mode 000.
2287 Lustre now respects this flag to disable mode checks when
2288 truncating a file owned by the user
2291 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2294 Description: Revoking locks from clients that went dead or catatonic might take
2296 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2297 cancellation of such locks instant on servers without waiting for
2298 any reply from clients. Clients drops these locks when cancel
2299 notification from server is received without replying.
2302 Frequency : liblustre-only, when liblustre client dies or becomes busy
2304 Description: Doing ls on Linux clients can take a long time with active
2306 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2307 granting such locks to it in the first place if possible. Locks
2308 are taken by proxy on the OST during the read or write and
2309 dropped immediately afterward. Add connect flags handling, do
2310 not grant locks to liblustre clients for glimpse ASTs.
2312 Severity : enhancement
2314 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2315 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2316 sequential reads and behaves badly with applications doing random
2317 reads. Improve it by reading ahead at least read region, and
2318 avoiding excessive large RPC for small reads.
2320 Severity : enhancement
2322 Description: Creating more than 1000 files for a single job may cause a load
2323 imbalance on the OSTs if there are also a large number of OSTs.
2324 Details : qos_prep_create() uses an OST index reseed value that is an
2325 even multiple of the number of available OSTs so that if the
2326 reseed happens in the middle of the object allocation it will
2327 still utilize the OSTs as uniformly as possible.
2332 Description: OST or MDS may oops in ping_evictor_main()
2333 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2334 but doesn't restart at beginning of obd_exports_timed list
2337 Severity : enhancement
2339 Description: improve by-nid export eviction on the MDS and OST
2340 Details : allow multiple exports with the same NID to be evicted at one
2341 time without re-searching the exports list.
2344 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2346 Description: MDS may oops in groups_free()
2347 Details : in rare race conditions a newly allocated group_info struct is
2348 freed again, and this can be NULL. The 2.4 compatibility code
2349 for groups_free() checked for a NULL pointer, but 2.6 did not.
2352 Frequency : common for liblustre clients doing little filesystem IO
2353 Bugzilla : 9352, 7313
2354 Description: server may evict liblustre clients accessing contended locks
2355 Details : if a client is granted a lock or receives a completion AST
2356 with a blocking AST already set it would not reply to the AST
2357 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2360 Frequency : lfs setstripe, only systems with more than 160 OSTs
2362 Description: unable to set striping with a starting offset beyond OST 160
2363 Details : llapi_create_file() incorrectly limited the starting stripe
2364 index to the maximum single-file stripe count.
2367 Frequency : LDAP users only
2369 Description: lconf did not handle in-kernel recovery with LDAP properly
2370 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2372 Severity : enhancement
2374 Description: bind OST threads to NUMA nodes to improve performance
2375 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2376 node and do their allocations there to localize memory access
2378 Severity : enhancement
2380 Description: llmount can determine client NID directly from Myrinet (GM)
2381 Details : the client NID code from gmnalnid was moved directly into
2382 llmount, removing the need to use this or specifying the
2383 client NID explicitly when mounting GM clients with zeroconf
2386 Frequency : if client is started with down MDS
2388 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2389 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2390 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2391 ptlrpc_import_delayed_req() because we didn't check if the
2392 request was interrupted, and we also didn't break out of the
2393 event loop if there was no timeout
2398 Description: data loss during non-page-aligned writes to a single file from
2399 both multiple nodes and multiple threads on one node at same time
2400 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2401 inconsistency led to false short-reads, that were cached and later
2402 used by ->prepare_write() to fill in partially written page,
2403 leading to data loss.
2406 Frequency : always, if lconf --abort_recovery used
2408 Description: lconf --abort_recovery fails with 'Operation not supported'
2409 Details : lconf was attempting to abort recovery on the MDT device and not
2412 Severity : enhancement
2414 Description: remove cleanup logs
2415 Details : replace lconf-generated cleanup logs with lustre internal
2416 cleanup routines. Eliminates the need for client-cleanup and
2419 Severity : enhancement
2421 Description: add support for EAs (user and system) on lustre filesystems
2422 Details : it is now possible to store extended attributes in the Lustre
2423 client filesystem, and with the user_xattr mount option it
2424 is possible to allow users to store EAs on their files also
2426 Severity : enhancement
2428 Description: Add possibility (config option) to show minimal available OST free
2430 Details : When compiled with --enable-mindf configure option, statfs(2)
2431 (and so, df) will return least minimal free space available from
2432 all OSTs as amount of free space on FS, instead of summary of
2433 free spaces of all OSTs.
2435 Severity : enhancement
2437 Description: do not expand extent locks acquired on OST-side
2438 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2439 by server: they are not cached anyway.
2442 Frequency : when mmap is used/binaries executed from Lustre
2444 Description: Unmmap pages before throwing them away from read cache.
2445 Details : llap_shrink cache now attempts to unmap pages before discarding
2446 them (if unmapping failed - do not discard). SLES9 kernel has
2447 extra checks that trigger if this unmapping is not done first.
2452 Description: lconf didn't resolve symlinks before checking to see whether a
2453 given mountpoint was already in use
2456 Frequency : when migrating failover services
2457 Bugzilla : 6395, 9514
2458 Description: When migrating a subset of services from a node (e.g. failback
2459 from a failover service node) the remaining services would
2460 time out and evict clients.
2461 Details : lconf --force (implied by --failover) sets the global obd_timeout
2462 to 5 seconds in order to quickly disconnect, but this caused
2463 other RPCs to time out too quickly. Do not change the global
2464 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2466 Severity : enhancement
2467 Frequency : if MDS is started with down OST
2468 Bugzilla : 9439,5706
2469 Description: Allow startup/shutdown of an MDS without depending on the
2470 availability of the OSTs.
2471 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2472 Add appropriate locking and lov-osc refcounts for safe
2473 cleaning. Add osc abort_inflight calls in case the
2474 synchronize never started.
2477 Frequency : occasional (Cray XT3 only)
2479 Description: root not authorized to access files in CRAY_PORTALS environment
2480 Details : The client process capabilities were not honoured on the MDS in
2481 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2482 been accessed by an authorized user then root was able to access
2483 the file on the local client also. The root user capabilities
2484 are now allowed on the MDS, as this environment has secure UID.
2487 Frequency : occasional
2489 Description: ldiskfs "too long searching" message happens too often
2490 Details : A debugging message (otherwise harmless) prints too often on
2491 the OST console. This has been reduced to only happen when
2492 there are fragmentation problems on the filesystem.
2497 Description: Division by zero in statfs when all OSCs are inactive
2498 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2499 lov_statfs divides by value returned by lov_get_stripecnt().
2503 Bugzilla : 9489, 3273
2504 Description: First write from each client to each OST was only 4kB in size,
2505 to initialize client writeback cache, which caused sub-optimal
2506 RPCs and poor layout on disk for the first writen file.
2507 Details : Clients now request an initial cache grant at (re)connect time
2508 and so that they can start streaming writes to the cache right
2509 away and always do full-sized RPCs if there is enough data.
2510 If the OST is rebooted the client also re-establishes its grant
2511 so that client cached writes will be honoured under the grant.
2516 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2517 Details : Now I/O RPCs go to different portal number and (presumably) fast
2518 lock requests (and glimses) and other RPCs get their own service
2519 threads pool that should be able to service those RPCs
2522 Severity : enhancement
2524 Description: Ability to exchange lustre version between client and servers and
2525 issue warnings at client side if client is too old. Also for
2526 liblustre clients there is ability to refuse connection of too old
2528 Details : New 'version' field is added to connect data structure that is
2529 filled with version info. That info is later checked by server and
2533 Frequency : rare, liblustre only.
2534 Bugzilla : 9296, 9581
2535 Description: Two simultaneous writes from liblustre at offset within same page
2536 might proceed at the same time overwriting eachother with stale
2538 Details : I/O lock withing llu_file_prwv was released too early, before data
2539 actually was hitting the wire. Extended lock-holding time until
2540 server acknowledges receiving data.
2543 Frequency : extremely rare. Never observed in practice.
2545 Description: avoid generating lustre_handle cookie of 0.
2546 Details : class_handle_hash() generates handle cookies by incrementing
2547 global counter, and can hit 0 occasionaly (this is unlikely, but
2548 not impossible, because initial value of cookie counter is
2549 selected randonly). Value of 0 is used as a sentinel meaning
2550 "unassigned handle" --- avoid it. Also coalesce two critical
2551 sections in this function into one.
2553 Severity : enhancement
2555 Description: allow liblustre clients to delegate truncate locking to OST
2556 Details : To avoid overhead of locking, liblustre client instructs OST to
2557 take extent lock in ost_punch() on client's behalf. New connection
2558 flag is added to handle backward compatibility.
2560 Severity : enhancement
2561 Bugzilla : 4928, 7341, 9758
2562 Description: allow number of OST service threads to be specified
2563 Details : a module parameter allows the number of OST service threads
2564 to be specified via "options ost ost_num_threads={N}" in the
2565 OSS's /etc/modules.conf or /etc/modprobe.conf.
2569 Bugzilla : 6146, 9635, 9895
2570 Description: servers crash with bad pointer in target_handle_connect()
2571 Details : In rare cases when a client is reconnecting it was possible that
2572 the connection request was the last reference for that export.
2573 We would temporarily drop the export reference and get a new
2574 one, but this may have been the last reference and the export
2575 was just destroyed. Get new reference before dropping old one.
2577 Severity : enhancement
2578 Frequency : if client is started with failover MDS
2580 Description: Allow multiple MDS hostnames in the mount command
2581 Details : Try to read the configuration from all specified MDS
2582 hostnames during a client mount in case the "primary"
2585 Severity : enhancement
2587 Description: Stop sending data to evicted clients as soon as possible.
2588 Details : Check if the client we are about to send or are sending data to
2589 was evicted already. (Check is done every second of waiting,
2590 for which l_wait_event interface was extended to allow checking
2591 of exit condition at specified intervals).
2594 Frequency : rare, normally only when NFS exporting is done from client
2596 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2598 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2601 Frequency : always, if srand() is called before liblustre initialization
2603 Description: Liblustre uses system PRNG disturbing its usage by user application
2604 Details : Introduce internal to lustre fast and high-quality PRNG for
2605 lustre usage and make liblustre and some other places in generic
2606 lustre code to use it.
2608 Severity : enhancement
2609 Bugzilla : 9477, 9557, 9870
2610 Description: Verify that the MDS configuration logs are updated when xml is
2611 Details : Check if the .xml configuration logs are newer than the config
2612 logs stored on the MDS and report an error if this is the case.
2613 Request --write-conf, or allow starting with --old_conf.
2615 Severity : enhancement
2617 Description: Handle symlinks in the path when checking if Lustre is mounted.
2618 Details : Resolve intermediate symlinks when checking if a client has
2619 mounted a filesystem to avoid duplicate client mounts.
2624 Description: lconf can hit an error exception but still return success.
2625 Details : The lconf command catches the Command error exception at the top
2626 level script context and will exit with the associated exit
2627 status, but doesn't ensure that this exit status is non-zero.
2632 Description: failure of ptlrpc thread startup can cause oops
2633 Details : Starting a ptlrpc service thread can fail if there are a large
2634 number of threads or the server memory is very fragmented.
2635 Handle this without oopsing.
2638 Frequency : always, only if liblustre and non-default acceptor port was used
2640 Description: liblustre cannot connect to servers with non-default acceptor port
2641 Details : tcpnal_set_default_params() was not called and was therefore
2642 ignoring the environment varaible TCPNAL_PORT, as well as other
2643 TCPNAL_ environment variables
2648 Description: two objects could be created on the same OST for a single file
2649 Details : If an OST is down, in some cases it was possible to create two
2650 objects on a single OST for a single file. No problems other
2651 than potential performance impact and spurious error messages.
2655 Bugzilla : 5681, 9562
2656 Description: Client may oops in ll_unhash_aliases
2657 Details : Client dcache may become inconsistent in race condition.
2658 In some cases "getcwd" can fail if the current directory is
2664 Description: Inode refcounting problems in NFS export code
2665 Details : link_raw functions used to call d_instantiate without obtaining
2666 extra inode reference first.
2670 Bugzilla : 9942, 9903
2671 Description: Referencing freed requests leading to crash, memleaks with NFS.
2672 Details : We used to require that call to ll_revalidate_it was always
2673 followed by ll_lookup_it. Also with revalidate_special() it is
2674 possible to call ll_revalidate_it() twice for the same dentry
2675 even if first occurence returned success. This fix changes semantic
2676 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2677 reference on a request referred from the intent.
2678 ll_intent_release() then releases such a request.
2681 Frequency : rare, normally benchmark loads only
2683 Description: unlinked inodes were kept in memory on the client
2684 Details : If a client is repeatedly creating and unlinking files it
2685 can accumulate a lot of stale inodes in the inode slab cache.
2686 If there is no other client load running this can cause the
2687 client node to run out of memory. Instead flush old inodes
2688 from client cache that have the same inode number as a new inode.
2691 Frequency : SLES9 2.6.5 kernel and long filenames only
2692 Bugzilla : 9969, 10379
2693 Description: utime reports stale NFS file handle
2694 Details : SLES9 uses out-of-dentry names in some cases, which confused
2695 the lustre dentry revalidation. Change it to always use the
2699 Frequency : rare, unless heavy write-truncate concurrency is continuous
2700 Bugzilla : 4180, 6984, 7171, 9963, 9331
2701 Description: OST becomes very slow and/or deadlocked during object unlink
2702 Details : filter_destroy() was holding onto the parent directory lock
2703 while truncating+unlinking objects. For very large objects this
2704 may block other threads for a long time and slow overall OST
2705 responsiveness. It may also be possible to get a lock ordering
2706 deadlock in this case, or run out of journal credits because of
2707 the combined truncate+unlink. Solution is to do object truncate
2708 first in one transaction without parent lock, and then do the
2709 final unlink in a new transaction with the parent lock. This
2710 reduces the lock hold time dramatically.
2713 Frequency : rare, 2.4 kernels only
2715 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2716 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2717 before unmounting their filesystem in order to avoid deadlock.
2718 The kernel_locked() function in 2.4 kernels only checks whether
2719 the kernel lock is held, not whether it is this process that is
2720 holding it as 2.6 kernels do.
2725 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2726 Details : The client ptlrpc code may be trying to reconnect to a down
2727 server before a previous connection attempt has timed out.
2728 Increase the reconnect interval to be longer than the connection
2729 timeout interval to avoid sending duplicate connections to
2733 Frequency : echo_client brw_test command
2735 Description: fix echo_client to work with OST preallocated code
2736 Details : OST preallocation code (5137) didn't take echo_client IO path
2737 into account: echo_client calls filter methods outside of any
2738 OST thread and, hence, there is no per-thread preallocated
2739 pages and buffers to use. Solution: hijack pga pages for IO. As
2740 a byproduct, this avoids unnecessary data copying.
2744 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2745 Description: Client can oops in mdc_commit_close() after open replay
2746 Details : It was possible for the MDS to return an open request with no
2747 transaction number in mds_finish_transno() if the client was
2748 evicted, but without actually returning an error. Clients
2749 would later try to replay that open and may trip an assertion
2750 Simplify the client close codepath, and always return an error
2751 from the MDS in case the open is not successful.
2754 Frequency : rare, 2.6 OSTs only
2756 Description: OST may deadlock under high load on fragmented files
2757 Details : If there was a heavy load and highly-fragmented OST filesystems
2758 it was possible to have all the OST threads deadlock waiting on
2759 allocation of biovecs, because the biovecs were not released
2760 until the entire RPC IO was completed. Instead, release biovecs
2761 as soon as they are complete to ensure forward IO progress.
2763 Severity : enhancement
2765 Description: Support for specifying external journal device at mount
2766 Details : If an OST or MDS device is formatted with an external journal
2767 device, this device major/minor is stored in the ext3 superblock
2768 and may not be valid for failover. Allow detecting and
2769 specifying the external journal at mount time.
2774 Description: Mounting an MDS with pending unlinked files may cause oops
2775 Details : target_finish_recovery() calls mds_postrecov() which returned
2776 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2777 considers this an error and immediately begins cleaning up the
2778 lov, just after starting the mds_lov process
2780 Severity : enhancement
2782 Description: Implement 'lfs df' to report actual free space on per-OST basis
2783 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2784 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2785 usage of objects; '-h' to report in human readable format.
2787 ------------------------------------------------------------------------------
2789 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2796 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2797 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2798 at startup during orphan unlinking. Instead, skip buddy update
2799 before setup, as it will be regenerated after recovery is complete.
2804 Description: If an OST is inactive, its locks might reference stale inodes.
2805 Details : lov_change_cbdata() must iterate over all namespaces, even if
2806 they are inactive to clear inode references from the lock.
2808 Severity : enhancement
2809 Frequency : occasional, if non-standard max_dirty_mb used
2811 Description: Client will block write RPCs if not enough grant
2812 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2813 then the client will block writes while waiting for another RPC
2814 to complete instead of consuming its dirty limit. With change
2815 we get improved performance when max_dirty_mb is small.
2817 Severity : enhancement
2818 Bugzilla : 3389, 6253
2819 Description: Add support for supplementary groups on the MDS.
2820 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2821 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2822 lookups for user supplementary groups into a cache.
2826 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2827 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2828 for liblustre O_EXCL is enforced by MDS.
2832 Description: readdir never returns NULL in liblustre.
2833 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2834 dentry in struct dirent.
2838 Frequency : liblustre only, depends on application IO pattern
2839 Description: liblustre clients evicted if not contacting servers
2840 Details : Don't put liblustre clients into the ping_evictor list, so
2841 they will not be evicted by the pinger ever.
2843 Severity : enhancement
2845 Description: Add ability to evict clients by NID from MDS.
2846 Details : By echoing "nid:$NID" string into
2847 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2848 $NID would be instantly evicted from this MDS and from all active
2849 OSTs connected to it.
2853 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2854 Details : lookup_it_finish() used to query file size from OSTs that was not
2859 Description: service threads change working directory to that of init
2860 Details : Starting lustre service threads may pin the working directory
2861 of the parent thread, making that filesystem busy. Threads
2862 now change to the working directory of init to avoid this.
2866 Frequency : during shutdown only
2867 Description: shutdown with a failed MDS or OST can cause unmount to hang
2868 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2873 Frequency : 2.6 only
2874 Description: chmod/chown may include an extra supplementary group
2875 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2876 supplementary group and if none is specified this is used.
2879 Bugzilla : 5479 (6816)
2881 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2882 Details : If lookup is in progress on a file that is unlinked we might try
2883 to revalidate the inode and fail in revalidate after lookup is
2884 complete and ll_file_open() enqueues the open again but
2885 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2888 Frequency : always, if lconf --abort_recovery used
2890 Description: lconf --abort_recovery fails with 'Operation not supported'
2891 Details : lconf was attempting to abort recovery on the MDT device and not
2894 ------------------------------------------------------------------------------
2896 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2901 Frequency : rare (only unsupported configurations with a node running as an
2903 Bugzilla : 6514, 5137
2904 Description: Mounting a Lustre file system on a node running as an OST could
2906 Details : OSTs now preallocates memory needed to write out data at
2907 startup, instead of when needed, to avoid having to
2908 allocate memory in possibly low memory situations.
2909 Specifically, if the file system is mounted on on OST,
2910 memory pressure could force it to try to write out data,
2911 which it needed to allocate memory to do. Due to the low
2912 memory, it would be unable to do so and the node would
2913 become unresponsive.
2915 Severity : enhancement
2917 Description: Addition of lconf --service command line option
2918 Details : lconf now accepts a '--service <arg>' option, which is
2919 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2921 Severity : enhancement
2923 Description: Failover mode is now the default for OSTs.
2924 Details : By default, OSTs will now run in failover mode. To return to
2925 the old behaviour, add '--failout' to the lmc line for OSTs.
2927 Severity : enhancement
2929 Description: Health checks are now provided for MDS and OSTs
2930 Details : Additional detailed health check information on MSD and OSTs
2931 is now provided through the procfs health_check value.
2934 Frequency : occasional, depends on IO load
2936 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2937 after numerous create/delete cycles
2938 Details : The ext3 inode allocation policy would not allocate new inodes
2939 very well on the OSTs because there are no new directories
2940 being created. Instead we look for groups with free space if
2941 the parent directories are nearly full.
2946 Description: Network or server problems during mount may cause partially
2947 mounted clients instead of returning an error.
2948 Details : The config llog parsing code may overwrite the error return
2949 code during mount error handling, returning success instead
2955 Description: MDS can fail to allocate large reply buffers
2956 Details : After long uptimes the MDS can fail to allocate large reply
2957 buffers (e.g. zconf client mount config records) due to memory
2958 fragmentation or consumption by the buffer cache. Preallocate
2959 some large reply buffers so that these replies can be sent even
2960 under memory pressure.
2964 Frequency : rare (liblustre)
2965 Description: fsx running with liblustre complained that using truncate() to
2966 extend the file doesn't work. This patch corrects that issue.
2967 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2968 ATTR_SIZE and lsm use in llu_setattr_raw.
2972 Frequency : rare, only 2.6 kernels
2973 Description: Unusual file access patterns on the MDS may result in inode
2974 data being lost in very rare circumstances.
2975 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2976 code in the 2.6 kernel caused the inode and/or EA data not to
2977 be read from disk, causing single-file corruption.
2981 Frequency : rare, only 2.6 filesystems using extents
2982 Description: Heavy concurrent write and delete load may cause data corruption.
2983 Details : It was possible under high-load situations to have an extent
2984 metadata block in the block device cache from a just-unlinked
2985 file overwrite a newly-allocated data block. We now unmap any
2986 metadata buffers that alias just-allocated data blocks.
2990 Frequency : filesystems with default stripe_count larger than 77
2991 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
2992 Details : lconf specifies an inode size of 4096 bytes when the default
2993 stripe_count is larger than 77. This conflicts with the default
2994 inode density of 1 per 4096 bytes. Allocate smaller inodes in
2995 this case to avoid pinning too much memory for large EAs.
2997 ------------------------------------------------------------------------------
2999 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3004 Frequency : rare (extremely heavy IO load with hundreds of clients)
3006 Description: Client is evicted, gets IO error writing to file
3007 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3008 caused clients to be evicted for AST timeouts. The fixes in
3009 bug 5192 mean we no longer need to have such short AST timeouts
3010 so ldlm_timeout has been increased.
3013 Frequency : occasional during --force or --failover shutdown under load
3014 Bugzilla : 5949, 4834
3015 Description: Server oops/LBUG if stopped with --force or --failover under load
3016 Details : a collection of import/export refcount and cleanup ordering
3017 issues fixed for safer force cleanup
3020 Frequency : only filesystems larger than 120 OSTs
3021 Bugzilla : 5990, 6223
3022 Description: lfs getstripe would oops on a very large filesystem
3023 Details : lov_getconfig used kfree on vmalloc'd memory
3026 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3027 Bugzilla : 6242, 6243
3028 Description: reading from files that had been truncated to a non-zero size
3029 but never opened returned no data
3030 Details : ll_file_read() reads zeros from no-object files to EOF
3035 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3036 Details : There's little harm in aborting MDS/OSS recovery and letting it
3037 try again, so I removed the LASSERT and return an error instead.
3039 Severity : enhancement
3041 Description: New debugging infrastructure for tracking down data corruption
3042 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3043 (b) cover more of the client-side code path, and (c) try to narrow
3044 down where problems occurred
3048 Bugzilla : 3819, 4364, 4397, 6313
3049 Description: Racing close and eviction MDS could cause assertion in mds_close
3050 Details : It was possible to get multiple mfd references during close and
3051 client eviction, leading to one thread referencing a freed mfd.
3053 Severity: : enhancement
3054 Bugzilla : 3262, 6359
3055 Description: Attempts to reconnect to servers are now more aggressive.
3056 Details : This builds on the enhanced upcall-less recovery that was added
3057 in 1.4.2. When trying to reconnect to servers, clients will
3058 now try each server in the failover group every 10 seconds. By
3059 default, clients would previously try one server every 25 seconds.
3064 Description: After recovery, certain operations trigger a failed
3065 assertion on a client.
3066 Details : Failing over an mds, using lconf -d --failover, while a
3067 client was doing a readdir() call would cause the client to
3068 LBUG after recovery completed and the readdir() was resent.
3070 Severity : enhancement
3072 Description: Default groups are now added by lconf
3073 Details : You can now run lconf --group <servicename> without having to
3074 manually add groups with lmc.
3077 Frequency : occasional
3079 Description: Nodes with an elan id of 0 trigger a failed assertion
3082 Frequency : always when accessing e.g. tty/console device nodes
3084 Description: tty and some other devices nodes cannot be used on lustre
3085 Details : file's private_data field is used by device data and lustre
3086 values in there got lost. New field was added to struct file to
3087 store fs-specific private data.
3090 Frequency : when exporting Lustre via NFS
3092 Description: NFSD failed occasionally when looking up a path component
3093 Details : NFSD is looking up ".." which was broken in ext3 directories
3094 that had grown large enough to become hashed.
3097 Frequency : Clusters with multiple interfaces not on the same subnet
3099 Description: Nodes will repeatedly try to reconnect to an interface which it
3100 cannot reach and report an error to the log.
3101 Details : Extra peer list entries will be created by lconf with some peers
3102 unreachable. lconf now validates the peer before adding it.
3105 Frequency : Only if a default stripe is set on the filesystem root.
3107 Description: Setting a default stripe on the filesystem root prevented the
3108 filesystem from being remounted.
3109 Details : The client was sending extra request flags in the root getattr
3110 request and did not allocate a reply buffer for the dir EA.
3113 Frequency : occasional, higher if lots of files are accessed by one client
3114 Bugzilla : 6159, 6097
3115 Description: Client trips assertion regarding lsm mismatch/magic
3116 Details : While revalidating inodes the VFS looks up inodes with ifind()
3117 and in rare cases can find an inode that is being freed.
3118 The ll_test_inode() code will free the lsm during ifind()
3119 when it finds an existing inode and then the VFS later attaches
3120 this free lsm to a new inode.
3124 Bugzilla : 6422, 7030
3125 Description: MDS deadlock between mkdir and client eviction
3126 Details : Creating a new file via mkdir or mknod (starting a transaction
3127 and getting the ns lock) can deadlock with client eviction
3128 (gets ns lock and trying to finish a synchronous transaction).
3131 Frequency : occasional
3132 Description: While starting a server, the fsfilt_ext3 module could not be
3134 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3135 kernels. Previously, lconf would still use the ext3 name
3136 when trying to load modules. Now, it will correctly use
3137 ext3 on 2.4 and ldiskfs on 2.6.
3139 Severity : enhancement
3140 Description: The default stripe count has been changed to 1
3141 Details : The interpretation of the default stripe count (0, to lfs
3142 or lmc) has been changed to mean striping across a single
3143 OST, rather than all available. For general usage we have
3144 found a stripe count of 1 or 2 works best.
3146 Severity : enhancement
3147 Description: Add support for compiling against Cray portals.
3148 Details : Conditional compiling for some areas that are different
3152 Frequency : occasional
3153 Bugzilla : 6409, 6834
3154 Description: Creating files with an explicit stripe count may lead to
3155 a failed assertion on the MDS
3156 Details : If some OSTs are full or unavailable, creating files may
3157 trigger a failed assertion on the MDS. Now, Lustre will
3158 try to use other servers or return an error to the
3162 Frequency : occasional
3164 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3165 caused lock timeout during readahead (since 1.4.2).
3166 Details : Processes doing readahead might match a lock that hasn't been
3167 granted yet if there are overlapping and conflicting lock
3168 requests. The readahead process waits on ungranted lock
3169 (original lock is CBPENDING), while OST waits for that process
3170 to cancel CBPENDING read lock and eventually evicts client.
3172 Severity : enhancement
3174 Description: Initial enabling of flock support for clients
3175 Details : Implements fcntl advisory locking and file status functions.
3176 This feature is provided as an optional mount flag (default
3177 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3178 locking are implemented yet, and those that are are not guaranteed
3179 to be completely correct in production environments.
3180 mount -t lustre -o [flock|noflock] ...
3183 Frequency : occasional
3185 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3186 assertion in the ext3 JBD (journaling) layer.
3187 Details : The b_committed_data struct is protected by the big kernel lock
3188 in 2.4 kernels, serializing journal_commit_transaction() and
3189 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3190 access to this struct. In 2.6 kernels there is finer grained
3191 locking to improve SMP performance of the JBD layer.
3195 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3200 Description: merge and cleanup kernel patches.
3203 -----------------------------------------------------------------------------
3205 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3207 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3208 versions of Lustre. Please update all servers and clients to
3209 version 1.4.2 or later at the same time. You must also run
3210 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3211 to update the configuration logs.
3213 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3214 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3215 - swab llog records of type '0' so we get proper header size/idx (5861)
3216 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3217 - fix rename of one directory over another leaking an inode (5953)
3218 - avoid SetPageDirty on 2.6 (5981)
3219 - don't re-add just-being-destroyed locks to the waiting list (5653)
3220 - when creating new directories, inherit the parent's custom
3221 striping settings if present parent (3048)
3222 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3223 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3224 - don't hold client locks on temporary worklist from l_lru (5666)
3225 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3226 - automatically evict dead clients (5921)
3227 - Update file size properly in create+truncate+fstat case (6196)
3228 - Do not unhash mountpoint dentries, do not allow removal of
3230 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3231 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3232 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3233 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3234 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3235 - Correctly handle reads of files with no objects (6243)
3236 - lctl recover will also mark a device active if deactivate used (5933)
3238 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3239 - allow --write-conf on an MDS with different nettype than client (5619)
3240 - don't write config llogs to MDS for mounts not from that MDS (5617)
3241 - lconf should create multiple TCP connections from a client (5201)
3242 - init scripts are now turned off by default; run chkconfig --on
3243 lustre and chkconfig --on lustrefs to use them
3244 - upcalls are no longer needed for clients to recover to failover
3246 - add --abort-recovery option to lconf to abort recovery on device
3248 - add support for an arbitrary number of OSTs (3026)
3249 - Quota support protocol changes.
3250 - forward compatibility changes to wire structs (6007)
3251 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3252 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3253 - improved "lustre status" script
3254 - initialize blocksize for non-regular files (6062)
3255 - added --disable-server and --disable-client configure options (5782)
3256 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3257 - Vanilla 2.4.29 support
3258 - increase maximum number of obd devices to 520 (6242)
3259 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3260 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3261 SLES 9 2.6.5 kernels
3262 - limit stripes per file to 160 (the maximum EA size) (6093)
3264 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3267 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3268 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3269 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3270 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3271 - small ext3 extents cleanups and fixes (5733)
3272 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3273 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3274 - avoid needless client->OST connect, fix handle mismatch (5317)
3275 - fix DLM error path that led to out-of-sync client, long delays (5779)
3276 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3277 - fix several locking issues related to i_size (5492,5624,5654,5672)
3278 - don't move pending lock onto export if it is already evicted (5683)
3279 - fix kernel oops when creating .foo in unlinked directory (5548)
3280 - fix deadlock in obdfilter statistics vs. object create (5811)
3281 - use time_{before,after} to avoid timer jiffies wrap (5882)
3282 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3283 - Do not leak request if server was not able to process it (5154)
3284 - If mds_open unable to find parent dir, make that negative lookup(5154)
3285 - don't create new directories with extent-mapping (5909, 5936)
3287 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3288 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3289 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3290 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3291 - support for RHEL4 kernel on i686 (5773)
3292 - provide error messages when incompatible logs are encountered (5898)
3294 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3295 * version 1.4.0.10 (1.4.1 release candidate 1)
3297 - don't keep a lock reference when lock is not granted (4238)
3298 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3299 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3300 - drop import inflight refcount on signal_completed_replay error (5255)
3301 - unlock page after async write error during send (3677)
3302 - handle missing objects in filter_preprw_read properly (5265)
3303 - no transno return for symlink open, don't save no-trasno open (3440)
3304 - don't try to complete elan receive that already failed (4012)
3305 - free RPC server reply state on error (5406)
3306 - clean up thread from ptlrpc_start_thread() on error (5160)
3307 - readahead could read extra page into cache that wasn't ejected (5388)
3308 - prevent races in class_attach/setup/cleanup/detach (5260)
3309 - don't dereference de->d_inode after l_dput of de (5458)
3310 - use "int" for stripe value returned from lock_to_stripe (5544)
3311 - mballoc allocation and error-checking fixes in 2.6 (5504)
3312 - block device patches to fix I/O request sizes in 2.6 (5482)
3313 - look up hostnames for IB nals (5602)
3314 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3315 - don't start multiple acceptors for the same port (5277)
3316 - fix incorrect LASSERT in mds_getattr_name (5635)
3317 - export a proc file for general "ping" checking (5628)
3318 - fix "lfs check" to not block when the MDS is down (5628)
3320 - service request history (4965)
3321 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3322 - create an "evict_client" /proc entry on OSTs, like the MDS has
3323 - fix mount usage message, return errors per mount(8) (5168)
3324 - change grep [] to grep "[]" in tests so they work in more UMLs
3325 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3326 - remove ancient LOV_MAGIC_V0 EA support (5047)
3327 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3328 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3329 - allow readahead allocations to fail when low on memory (5383)
3330 - mmap locking landed again, after considerable improvement (2828)
3331 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3333 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3336 - send OST transaction number in read/write reply to free req (4966)
3337 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3338 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3339 - the watchdog thread now runs as interruptible (5246)
3340 - flock/lockf fixes (but it's still disabled, pending 5135)
3341 - don't use EXT3 constants in llite code (5094)
3342 - memory shortage at startup could cause assertion (5176)
3344 - reorganization of lov code
3345 - single portals codebase
3347 - add extents/mballoc support (5025)
3348 - direct I/O reads in the obdfilter (4048)
3349 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3351 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3354 - send OST transaction number in read/write reply to free req (4966)
3355 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3356 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3357 - the watchdog thread now runs as interruptible (5246)
3358 - handle missing objects in filter_preprw_read properly (5265)
3359 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3360 - drop import inflight refcount on signal_completed_replay error (5255)
3361 - unlock page after async write error during send (3677)
3362 - return original error code on reconstructed replies (3761)
3363 - no transno return for symlink open, don't save no-trasno open (3440)
3365 - add pid to ldlm debugging output (4922)
3366 - bump the watchdog timeouts -- we can't handle 30sec yet
3367 - extra debugging for orphan dentry/inode bug (5259)
3369 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3372 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3373 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3374 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3375 - fix NULL dereference in /proc/sys/portals/routes (4827)
3376 - allow failed mdc_close() operations to be interrupted (4561)
3377 - stop precreate on OST before MDS would time out on it (4778)
3378 - don't send partial-page writes before EOF from client (4410)
3379 - discard client grant for sub-page writes on large-page clients (4520)
3380 - don't free dentries not owned by NFS code, check generation (4806)
3381 - fix lsm leak if mds_create_objects() fails (4801)
3382 - limit debug_daemon file size, always print CERROR messages (4789)
3383 - use transno after validating reply (3892)
3384 - process timed out requests if import state changes (3754)
3385 - update mtime on OST during writes, return in glimpse (4829)
3386 - add mkfsoptions to LDAP (4679)
3387 - use ->max_readahead method instead of zapping global ra (5039)
3388 - don't interrupt __l_wait_event() during strace
3390 - add software watchdogs to catch hung threads quickly (4941)
3391 - make lustrefs init script start after nfs is mounted
3392 - fix CWARN/ERROR duplication (4930)
3393 - return async write errors to application if possible (2248)
3394 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3395 - print NAL number in %x format (4645)
3396 - update barely-supported suse-2.4.21-171 series (4842)
3397 - support for sles 9 %post scripts
3398 - support for building 2.6 kernel-source packages
3399 - support for sles km_* packages
3401 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3404 - ignore -ENOENT errors in osc_destroy (3639)
3405 - notify osc create thread that OSC is being cleaned up (4600)
3406 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3407 - reconstruct ost_handle() like mds_handle() (4657)
3408 - create a new thread to do import eviction to avoid deadlock (3969)
3409 - let lconf resolve symlinked-to devices (4629)
3410 - don't unlink "objects" from directory with default EA (4554)
3411 - hold socknal file ref over connect in case target is down (4394)
3412 - allow more than 32000 subdirectories in a single directory (3244)
3413 - fix blocks count for O_DIRECT writes (3751)
3414 - OST returns ENOSPC from object create when no space left (4539)
3415 - don't send truncate RPC if file size isn't changing (4410)
3416 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3417 - bind to privileged port in socknal and tcpnal (3689)
3419 - rate limit CERROR/CWARN console message to avoid overload (4519)
3420 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3421 - basic mmap support (3918)
3422 - kernel patch series update from b1_4 (4711)
3424 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3427 - avoid crash during MDS cleanup with OST shut down (2775)
3428 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3429 - don't use bad inodes on the MDS (3744)
3430 - dynamic object preallocation to improve recovery speed (4236)
3431 - don't hold spinlock over lock dumping or change debug flags (4401)
3432 - don't zero obd_dev when it is force cleaned (3651)
3433 - print grants to console if they go negative (4431)
3434 - "lctl deactivate" will stop automatic recovery attempts (3406)
3435 - look for existing locks in ldlm_handle_enqueue() (3764)
3436 - don't resolve lock handle twice in recovery avoiding race (4401)
3437 - revalidate should check working dir is a directory (4134)
3439 - don't always mark "slow" obdfilter messages as errors (4418)
3441 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3444 - don't close LustreDB during write_conf until it is done (3860)
3445 - fix typo in lconf for_each_profile (3821)
3446 - allow dumping logs from multiple threads at one time (3820)
3447 - don't allow multiple threads in OSC recovery (3812)
3448 - fix debug_size parameters (3864)
3449 - fix mds_postrecov to initialize import for llog ctxt (3121)
3450 - replace config semaphore with spinlock (3306)
3451 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3452 - don't allow enqueue to complete on a destroyed export (3822)
3453 - down write_lock before checking llog header bitmap (3825)
3454 - recover from lock replay timeout (3764)
3455 - up llog sem before sending rpc (3652)
3456 - reduce ns lock hold times when setting kms (3267)
3457 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3458 - fix NULL deref and obd_dev leak on setup error (3312)
3459 - replace some LBUG about llog ops with error handling (3841)
3460 - don't match INVALID dentries from d_lookup and spin (3784)
3461 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3462 - fix invalid assertion in ptlrpc_set_wait (3880)
3464 - add libwrap support for the TCP acceptor (3996)
3465 - add /proc/sys/portals/routes for non-root route listing (3994)
3466 - allow setting MDS UUID in .xml (2580)
3467 - print the stack of a process that LBUGs (4228)
3469 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3472 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3473 - make sure to unset replay flag from failed open requests (3440)
3474 - if default stripe count is 0, use OST count for inode size (3636)
3475 - update parent mtime/ctime on client for create/unlink (2611)
3476 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3477 - free recovery state on server during a forced cleanup (3571)
3478 - unregister_reply for resent reqs (3063)
3479 - loop back devices mounting and status check on 2.6 (3563)
3480 - fix resource-creation race that can provoke i_size == 0 (3513)
3481 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3482 - more debugging for page-accounting assertion (3746)
3483 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3484 - avoid deadlock after precreation failure (3758)
3485 - fix race and lock order deadlock in orphan handling (3450, 3750)
3486 - add validity checks when grabbing inodes from l_ast_data (3599)
3488 - add /proc/.../recovery_status to obdfilter (3428)
3489 - lightweight CDEBUG infrastructure, debug daemon (3668)
3490 - change default OSC RPC parameters to be better on small clusters
3491 - turn off OST read cache for files smaller than 32MB
3492 - install man pages and include them in rpms (3100)
3493 - add new init script for (un)mounting lustre filesystems (2593)
3494 - run chkconfig in %post for init scripts (3701)
3495 - drop scimac NAL (unmaintained)
3497 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3500 - clean kiobufs before and after use (3485)
3501 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3502 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3503 - change init script to fail more clearly if not run as root (1528)
3504 - allow clients to reconnect during replay (1742)
3505 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3506 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3507 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3508 - fix cond_resched (3554)
3509 - only evict unfinished clients after recovery (3515)
3510 - allow bulk resend, prevent data loss (3570)
3511 - dynamic ptlrpc request buffer allocation (2102)
3512 - don't allow unlinking open directory if it isn't empty (2904)
3513 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3514 - remove extraneous obd dereference causing LASSERT failure (3334)
3515 - don't use get_cycles() when creating temp. files on the mds (3156)
3516 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3517 - handle EEXIST for set-stripe, set proper directory name (3336)
3519 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3520 - fix ksocknal_fmb_callback() error messages (2918)
3522 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3525 - don't copy lvb into (possibly NULL) reply on error (2983)
3526 - don't deref dentry after dput, don't free lvb on error (2922)
3527 - use the kms to determine writeback rpc length (2947)
3528 - increment oti_logcookies when osc is inactive (2948)
3529 - update client's i_blocks count via lvb messages (2543)
3530 - handle intent open/close of special files properly (1557)
3531 - mount MDS with errors=remount-ro, like obdfilter (2009)
3532 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3533 - don't use cancelling-locks' kms values (2947)
3534 - use highest lock extent for kms, not last one (2925)
3535 - don't dereference ERR_PTR() dentry in error handling path (3107)
3536 - fix thread race in portals_debug_dumplog() (3122)
3537 - create lprocfs device entries at setup instead of at attach (1519)
3538 - common AST error handler, don't evict client on completion race (3145)
3539 - zero nameidata in detach_mnt in 2.6 (3118)
3540 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3541 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3542 - initialize RPC timeout timer earlier for 2.6 (3219)
3543 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3544 - print nal/nid for unknown nid (3258)
3545 - additional checks for oscc recovery before doing precreate (3284)
3546 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3547 - don't crash in mdc_close for bad permissions on open (3285)
3548 - zero i_rdev for non-device files (3147)
3549 - clear page->private before handing to FS, better assertion (3119)
3550 - tune the read pipeline (3236)
3551 - fix incorrect decref of invalidated dentry (2350)
3552 - provide read-ahead stats and refine rpc in flight stats (3328)
3553 - don't hold journal transaction open across create RPC (3313)
3554 - update atime on MDS at close time (3265)
3555 - close LDAP connection when recovering to avoid server load (3315)
3556 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3557 - don't leak open file on MDS after open resend (3325)
3558 - serialize filter_precreate and filter_destroy_precreated (3329)
3559 - loop device shouldn't call sync_dev() for nul device (3092)
3560 - clear page cache after eviction (2766)
3561 - resynchronize MDS->OST in background (2824)
3562 - refuse to mount the same filesystem twice on same mountpoint (3394)
3563 - allow llmount to create routes for mounting behind routers (3320)
3564 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3565 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3566 - fix rare problem with rename on htree directories (3417)
3568 - allow default OST striping configuration per directory (1414)
3569 - fix compilation for qswnal for 2.6 kernels (3125)
3570 - increase maximum number of MDS request buffers for large systems
3571 - change liblustreapi to be useful for external progs like lfsck (3098)
3572 - increase local configuration timeout for slow disks (3353)
3573 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3575 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3578 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3579 - don't overwrite extent policy data in reply if lock was blocked (2901)
3580 - drop filter export grants atomically with removal from device (2663)
3581 - del obd_self_export from work_list in class_disconnect_exports (2908)
3582 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3583 - swab reply message in mdc_close, other PPC fixes (2464)
3584 - fix destroying of named logs (2325)
3585 - overwrite old logs when running lconf --write_conf (2264)
3586 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3587 - fix race in target_handle_connect (2898)
3588 - mds_reint_create() should take same inode create lock (2926)
3589 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3590 - don't close files for self_export to avoid uninitialized obd (2936)
3591 - allow MDS with the same name as client node (2939)
3592 - hold dentry reference for closed log files for unlink (2325)
3593 - reserve space for all logs during transactions (2059)
3594 - don't evict page beyond end of stripe extent (2925)
3595 - don't oops on a deleted current working directory (2399)
3596 - handle hard links to targets without a parent properly (2517)
3597 - don't dereference NULL lock when racing during eviction (2867)
3598 - don't grow lock extents when lots of conflicting locks (2919)
3600 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3603 - account for cache space usage on clients to avoid data loss (974)
3604 - lfsck support in lustre kernel code (2349)
3605 - reduce journal credits needed for BRW writes (2370)
3606 - orphan handling to avoid losing space on client/server crashes
3607 - ptlrpcd can be blocked, stopping ALL progress (2477)
3608 - use lock value blocks to assist in proper KMS, faster stat (1021)
3609 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3610 - recovery for initial connections (2355)
3611 - fixes for mds_cleanup_orphans (1934)
3612 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3613 - block all file creations until orphan recovery completes (1901)
3614 - client remove rq_connection from request struct (2423)
3615 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3616 - recovery timer race (2670)
3617 - mdc_close recovey bug (2532)
3618 - ptlrpc cleanup bug (2710)
3619 - mds timeout on local locks (2588)
3620 - namespace lock held during RPCs (2431)
3621 - handle interrupted sync write properly (2503)
3622 - don't try to handle a message that hasn't been replied to (2699)
3623 - client assert failure during cleanup after abort recovery (2701)
3624 - leak mdc device after failed mount (2712)
3625 - ptlrpc_check_set allows timedout requests to complete (2714)
3626 - wait for inflight reqs when ptlrpcd finishes (2710)
3627 - make sure unregistered services are removed from the srv_list
3628 - reset bulk XID's when resending them (caught by 1138 test)
3629 - unregister_bulk after timeout
3630 - fix lconf error (2694)
3631 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3632 - readahead locks pages, leaves pending causing memory pressure (2673)
3633 - increase OST request buffers to 4096 on large machines (2729)
3634 - fix up permission of existing directories in simple_mkdir (2661)
3635 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3636 - don't assign transno to errored transactions (2742)
3637 - don't delete objects on OST if given a bogus objid from MDS (2751)
3638 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3639 - if rq_no_resend, then timeout request after recovery (2432)
3640 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3641 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3642 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3643 - MDS recovery shouldn't delete objects that it has given out (2730)
3644 - if enqueue arrives after completion, don't clobber LVB (2819)
3645 - don't unlock pages twice when trigger_group_io returns error (2814)
3646 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3647 - don't write pages to disk if there was an error (1450)
3648 - don't ping imports that have recovery disabled (2676)
3649 - take buffered bytes into account when balancing socknal conn (2817)
3650 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3651 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3652 - remove little-endian swabbing of llog records (1987)
3653 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3654 - retry reposting request buffers if they fail (1191)
3655 - grow extent at grant time to avoid granting a revoked lock (2809)
3656 - lock revoke doesn't evict page if covered by a second lock (2765)
3657 - disable VM readahead to avoid reading outside lock extents (2805)
3659 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3660 - updated kernel patches for hp-2.4.20 kernel (2681)
3662 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3665 - fix truncated write corruption (2366)
3666 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3668 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3669 - ensure len(uuid) < 37 in lmc (1171)
3670 - fix ia64 OOPS in llog_test (2255)
3671 - zero end of page at obdfilter for partial page writes (2648)
3672 - don't leave stale dentries around after renames (bug 2428)
3673 - fix timeouts when evicting a client with a single lock held (2642)
3674 - set deadline for the initial HELLO message to drain (2634)
3675 - print out dotted-quad IP addresses in the socknal (2302)
3677 - additional debugging for MDS client eviction problem (2443)
3678 - fix mkfsoptions support for osts (2603, 2604)
3680 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3683 - add series for the vanilla 2.6.0 kernel
3684 - add series for the vanilla 2.4.24 kernel
3685 - add series for a cray x86/64 UL kernel drop
3686 - fix xattr patches for the vanilla 2.4.19 series
3688 - generate true UUIDs in lmc (1171)
3689 - have portals stack dumping break in UML (2466)
3690 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3691 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3692 - wake callback waiting threads on client eviction (2460)
3693 - Add --ptldebug and --subsystem to lmc (1719)
3694 - update assertion to allow safe interrupt allocation
3695 - set rq_no_resend for cancel requests (2432)
3696 - recalculate ptlrpcd timeout after resend (2494)
3697 - call vfs_rmdir when removing pending directories (2368)
3698 - fix renaming a file to itself (2429)
3699 - lmc creates a default one-stripe lov (2454)
3700 - expand procfs space to handle large clusters (2326)
3701 - increase UML stack to avoid overflow
3702 - update lconf's list of debug and subsystem masks
3703 - fix lfs find --obd (2510)
3704 - /proc tunable for disabling filter read caching (2591)
3705 - stop rpm packages from altering slapd.conf (2301)
3706 - disable nagle in the socknal under 0conf (2578)
3707 - choose mds inode size based on stripe count (2572)
3708 - fix kernel-source rpm problems (2516)
3710 - add --disable-doc to avoid pdf generation (2421)
3711 - update documentation, tests, type-os, comments
3712 - avoid format warnings on ia64
3713 - remove the TOE NAL
3714 - tiny code cleanups by removing unused fields
3716 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3719 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3720 - remove the most blatant lies from BUILDING file (2371)
3721 - change default debug level to reasonable production setting
3722 - reduce client side cache size to reduce cache flush time
3723 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3724 - make TCP zerocopy and pinger support enabled by default (2476)
3725 - sync writes completed after process exits caused crashes (2319)
3726 - maintain correct mount count on the MDS (2356)
3727 - backout 1557, because 2316 wasn't really fixed
3728 - better file I/O statistics gathering in /proc
3729 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3730 - another kernel patch to fix zero-copy TCP function export
3731 - don't take duplicate lock when processing re-sent getattr (2420)
3732 - lctl uses obd_self_export instead of creating new conn (2353)
3733 - MDS/OST recovery case which requires object creation asserted (2425)
3734 - move lfs from /usr/sbin to /usr/bin in packages
3735 - fix race between mds_client_add and mds_client_free (2417)
3736 - use kmalloc instead of slabs in portals (2430)
3737 - don't create duplicate records when a failover MDS is present (2442)
3738 - remove unnecessary mount age check (2332)
3739 - don't remove directory inodes from locks prematurely (2451)
3740 - don't break if MDS service name is the same as hostname (2103)
3741 - fix races in client write RPC generation when cache full (2482)
3743 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3746 - remove now-unused request->rq_obd (278)
3747 - if an allocation fails, print out how much memory we've used (1933)
3748 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3749 - add the "configurable stack size" patch to most series files (1256)
3750 - ability to write large log records, for 100+ OST configs (2306)
3751 - fix NULL deref when filter_prep fails (2314)
3752 - fix operator precedence error in filter_sync
3753 - dynamic allocation of socknal TX descriptors (2315)
3754 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3755 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3756 - ensure that CWARN messages in Portals always get to the syslog
3757 - __init/__exit are not for prototype decls (ldlm_init/exit)
3758 - x86-64 compile warning fixes
3759 - fix gateway LMC keyword conflict (2318)
3760 - fix MDS lock inversions in getattr/reint paths (1844)
3761 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3762 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3763 - DLM race condition prevented some lock evictions (2328)
3764 - ENOMEM detection and retry on socknal sends (2230)
3765 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3766 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3767 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3768 - better error messages when a client is rejected during recovery (1505)
3769 - avoid cancelling locks which were never granted, after failure (2330)
3770 - fix i_sem/journal inversion in mds_client_add (2333)
3771 - fix truncate/getattr lock cycle deadlock (2334)
3772 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3773 - fix two rare exit paths which could leak an l_lock() ref (2321)
3774 - fencepost error in MDS/OST orphan recovery (2226)
3775 - make log record alignment 8 bytes (1988)
3776 - lstripe now fails when requested offset > ost_count (2237)
3777 - ensure that all kernel series have a complete list.h (1607)
3778 - fix crashes in special-file operations (2316)
3779 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3781 - allow configurable automake binary, for testing new versions
3782 - small update to the lfs documentation
3784 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3786 * fix negative export reference count in fsfilt_sync (2312)
3788 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3789 * release candidate 0.9.1
3791 - orphans are moved into the PENDING directory for possible recovery
3792 - replayed opens now open by fid for orphan/rename safety (1042)
3793 - last close of an orphan inode generates a transno (683)
3794 - chdir() and mount() now pin the directory entry (1020)
3795 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3796 - discard very old requests without processing them (1502)
3797 - remove some common, well-understood CERRORs (1505)
3798 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3799 - clear "grant" flags in OST replies until OST grant code lands (1644)
3800 - fix read performance by not clobbering i_blksize on client (1598)
3801 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3802 - make LRU size a /proc tunable, clears locks when reduced (707)
3803 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3804 - server locks take references on exports now (1558)
3805 - build fixes for 2.4.20-rh trees (1663)
3806 - return an error from lov_create if all OSCs are inactive (1751)
3807 - fix import levels when a reconnect happens without a timeout (1597)
3808 - exit early from mds_open if we get a lookup error (1749)
3809 - partial page read at EOF wouldn't wait for disk before sending (1642)
3810 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3811 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3812 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3813 - write/truncate lock inversion (1639)
3814 - Don't auto-load obdclass, portals modules during cleanup (1495)
3815 - fix timestamps from jumping to "now" (1763)
3816 - extra journal assertions (1648)
3817 - add an extra multiunlink test (1771)
3818 - fix read_record/write_record API (1776)
3819 - fix leak of offset_extent, possible incorrect i_size later (1772)
3820 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3821 - Debugging for the kqswnal_get_idle_tx problems (1820)
3822 - Allow recovery to be attempted multiple times (1536)
3823 - Write out MDS last_rcvd file after it is first created (1600)
3824 - Fix tx_descriptor leak in failed transmit situations (1827)
3825 - ext3 journaling fixes for assertion failure after IO error (1871)
3826 - class_export_put() on freed export after completion AST error (1896)
3827 - Fix revalidate looping in VFS (1322)
3828 - Don't access a freed export during MDS_REINT timeout (1521)
3829 - Add open-unlink recovery support on the MDS (1673,1764)
3830 - Return an error if no MDS data was read from last_rcvd (1946)
3831 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3832 - Don't setup a disk device that doesn't match exported UUID (317)
3833 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3834 - avoid committing NULL handle in force close
3835 - local.sh is now a one-stripe LOV configuration
3836 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3837 - fix ext3 htree duplicate directory entry corruption (1516)
3838 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3839 - update to new LOV EA format (2097)
3840 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3842 2003-06-15 Phil Schwan <phil@clusterfs.com>
3845 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3846 - per-import/export recovery handling (958, 931, 959)
3847 - multiple last-rcvd slots, for serving multiple FSes (949)
3848 - connections are again shared between multiple imp/exports (963, 964)
3849 - "umount -f" would hang if any requests needed to be sent (393, 978)
3850 - avoid pinning large req buffer by copying for queued messages (989)
3851 - add "uuid" to "lctl device" command to help upcalls (991)
3852 - "open" RPCs with transnos would confuse recovery counters (1037)
3853 - do proper endian conversion of all wire messages (288, 340, 891)
3854 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3855 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3856 - fix ptlrpc_connection leak in target_handle_connect (1174)
3857 - fix import refcounting bug in OST and MDS cleanup (1134)
3858 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3859 - fix very unlikely obd_types race condition (501)
3860 - remove osc_open hack for echo_client (1187)
3861 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3862 - a failure in read_inode2 leads to deadlock (1139)
3863 - cancel ack-locks as soon as transaction is committed (1072)
3864 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3865 - make sure to commitrw after any preprw to avoid deadlock (1162)
3866 - failing to execute a file in a lustre FS would lock inode (1203)
3867 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3868 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3869 - obd_uuid2tgt would walk off the end of the list (1255)
3870 - on IA64 the osc would give portals incorrect bulk size (1258)
3871 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3872 - fix lock inversion caused by new llite matching code (1282)
3873 - limit the number of dirty pages on a client to 10MB (1286)
3874 - timed out locks were not being corrected cancelled (1289)
3875 - fix O_DIRECT above 4GB on IA-32 (1292)
3876 * major user-visible changes
3877 - fail out/fail over policy now controlled by the upcall (993)
3879 - add OBD_PING to check server availability and failure (954)
3880 - lustre messages are now sent in sending host order (288, 340, 891)
3881 - add eadatalen to MDS getattr reply (340)
3882 - OST read replies may contain second buffer, with per-page status (593)
3884 2003-03-11 Phil Schwan <phil@clusterfs.com>
3887 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3888 - failed open()s could cause deadlock; fixed (867, 869)
3889 - stop cancelling OST locks when files are closed (481)
3890 - overlapping XID spaces caused network corruption (851, 853)
3891 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3892 - setattr_raw added, to do single-RPC, server-side setattrs
3893 - lmc/lconf syntax change for OST UUIDs
3894 - fix crashy race condition between ptlrpc_free_req and osc_close
3895 - don't use request in mdc_enqueue if we hit a timeout (889)
3896 - don't set the inode i_size for regular files from the MDS (896)
3897 - handle out of order completion AST (842)
3898 - don't LBUG if a lock request times out after receiving AST (913)
3899 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3900 - if a bad lock AST arrives, send an error instead of dropping entirely
3901 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3902 - fix leak in bulk IO when only partially completed (899, 900, 926)
3903 - fix O_DIRECT for ia64 (55)
3904 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3905 - C-z after timeout could hang a process forever; fixed (977)
3907 - client-side I/O cache (678, 924, 929, 941, 970)
3909 - READPAGE and SETATTRs which don't take server-side locks get
3912 2003-02-11 Phil Schwan <phil@clusterfs.com>
3915 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3916 - Fix fsfilt_extN_readpage() to read a full page of directory
3917 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3918 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3919 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3920 - Most intent operations were removed, in favour of a new RPC mode
3921 that does a single RPC to the server and bypasses most of the VFS
3922 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3923 - Aggressively cancel local locks on DLM servers
3924 - mds_reint_unlink sends EA to the client if it's the last nlink.
3925 client uses that EA to unlink OST objects.
3926 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3927 - recursive symlinks were fixed (439)
3928 - fixed NULL deref in DEBUG_REQ
3929 - filter_update_lastobjid no longer calls sync, which annoyed extN
3930 - fixed multi-client small-writes to a single file problem (445)
3931 - fixed mtime updates during file writes (607)
3932 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3933 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3934 - fixed Linux OST size reporting problem (444, 656)
3935 - OST now updates object mtime with writes or setattr (607, 619)
3936 - client verifies file size before zeroing page past EOF (445)
3937 - OST now writes last allocated objid to disk with allocation (108)
3938 - LOV on echo now works (409)
3940 - mds_reint_unlink sends a new buffer, with the EA included. this
3941 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3942 set if a regular file was being unlinked, and it was the last link
3943 - use PtlGet from the target for bulk writes (315)
3944 - OST now updates object mtime with writes or setattr (607, 619)
3945 - LDLM now has a grant-time callback to revalidate locked items, if
3947 - Many MDS operations were reorganized to combat race conditions
3949 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3950 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3952 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3955 - Fully reactivate OST imports after reconnection (512, others)
3956 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3957 - More graceful error handling for truncating on dead OST (515)
3958 - Don't error out unless we're actually accessing dead stripes (474)
3959 - Fix garbage sizes when stripes are missing (410)
3960 - LRU counters were broken, causing constant lock purge (433, 432)
3961 - garbage on read from stripes with failed OSTs (441)
3962 - mark OSCs as active before reconnecting during recovery (438)
3963 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3964 - lfind did not preserve OST order in output (443)
3965 - symlinks cause hung clients, incorrect data (439)
3966 - stop dereferencing request after dropping refcount (457)
3967 - don't LASSERT(spin_is_locked) on non-SMP (455)
3968 - fixes for many rename() bugs
3969 - fstat didn't correctly synchronize attributes (399)
3970 - server must handle lock cancellation during blocking AST prep (487)
3971 - bulk descriptors were free()d too soon (511)
3972 - fix paths in lconf, which would load incorrect modules (451, 507)
3973 - fix confusing lconf 'host not found' error message (386)
3974 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3975 - fix race condition in mdc_blocking_ast() for inode access (526)
3976 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3977 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3978 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3979 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3980 - fix lstripe in several ways (kernel side) (527)
3981 - fix request leak in ldlm_cli_enqueue (262)
3982 - incorrect OSC was marked inactive after OST failure
3983 - call mds_fs_cleanup before unmounting filesystem (524)
3984 - fix races between taking ns_lock and ldlm_lock_change_resource
3985 - fix races updating LOV export open file list
3986 - fix lov_enqueue error path, avoid decref-ing bad lock handle
3987 - fix recovery NULL deref in ldlm_cli_cancel_unused
3988 - fix some DLM races by using new hash table for lock handles (419)
3989 - permit the client to specify desired inodes, at replay
3990 - duplicate requests when we queue them for replay reintegration
3991 - fix last_rcvd offset calculation
3992 - sync after each recovered transaction, so we always make progress
3993 - never, not always, ERESTART requests without transnos
3994 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
3996 - skip replay if the MDS says that the client is already connected
3997 - don't check for a recovery-enabled export to match lctl's UUID
3998 - don't INC_USE_COUNT for phantom exports
3999 - don't crash when cleaning up phantom exports (567)
4000 - don't double-finish or set replay data for errored mdc_open requests
4001 - abort requests when they time out, so we don't get old replies
4002 - send/receive replies for AST messages again
4003 - if the client says that it doesn't have the lock, cancel it on the
4005 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4006 mark it as destroyed, it will all work out when decref is called
4007 - fix module use counts (22, 581)
4009 - ASTs now expect a reply (server cancels lock on error reply)
4011 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4014 - fix many simultaneous client startup (392)
4015 - fix dentry->d_it clobbering
4016 - credentials weren't being shipped for readdir/getattr operations
4017 - remove invalid assertions triggered during some concurrent MD
4019 - proper Lustre versions added (336, 389)
4020 - fix memory leak for create error case (398)
4021 - fix LOV locking bug that would get cli/srv out of sync
4022 - fix echo client over LOV (409)
4023 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4024 - fix double-O_EXCL intent crash (424)
4025 - avoid sending multiple lock CANCELs (352)
4027 - MDS can do multi-client recovery (modulo bugs in new code)
4029 - many updates, edits, cleanups
4031 2002-11-18 Phil Schwan <phil@clusterfs.com>
4034 - fix null d_it dereference (346)
4035 - fix full OST/dbench hang (333)
4036 - fix permission problem with file removal (286)
4037 - fix removal of OSCs from LOV when they fail
4038 - fix NULL deref during bulk timeout (214)
4039 - fix problems related to multiple filesystems on one MDS (241)
4040 - fixed serious subtle metadata locking bugs
4041 - free locks on clients when inodes are removed due to memory
4043 - fix inode pointer in lock data (285)
4044 - partial support for multiple MDS on a single host (241)
4045 - data locks weren't cancelled at clear_inode time (290, 311)
4046 - intent locks could lead to unbounded lock growth (205)
4047 - added a maximum lock count, an LRU list, and a flusher
4048 - fix multiple rename (365)
4049 - properly abstracted the echo client
4050 - OSC locked 1 byte too many; fixed
4051 - rewrote brw callback code:
4052 - fixed recovery bugs related to LOVs (306)
4053 - fixed too-many-pages-in-one-write crash (191)
4054 - fixed (again) crash in sync_io_timeout (214)
4055 - probably fixed callback-related race (385)
4057 - Add capability to MDS protocol
4058 - LDLM cancellations and callbacks on different portals
4060 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4063 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4064 - defer open object destruction to close time (601981, 138)
4065 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4066 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4067 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4068 - don't call obd_setattr in ll_file_release for destroyed objects
4069 * protocol change to lustre_msg: move |version| and add |flags|
4070 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4071 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4072 * added replay of create, unlink, link and rename operations during
4073 MDS failover; recovery should be much more robust now
4074 * remove failed OSCs from LOVs (only lov_create uses this so far)
4075 * the lustre-HOWTO was brought (more) up to date (582544)
4077 2002-10-23 Phil Schwan <phil@clusterfs.com>
4080 - in-use dentries weren't being reused properly (617851)
4081 - prevent multiple LDLM setup (599178)
4082 - fix LOV size calculations for truncate (617853)
4083 - fix client handling of MDS intent errors (POSIX)
4084 - fix permission bug in lovstripe.c test (624321)
4085 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4086 - truncate past end of file could corrupt data
4087 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4088 - a race in recovery could return ETIMEDOUT to apps (623947)
4089 - building outside the source directory was fixed
4090 * the lustre-HOWTO was brought (more) up to date (582544)
4091 * major progress was made on recovery functionality
4093 2002-10-10 Phil Schwan <phil@clusterfs.com>
4096 - recovery deadlock fix
4097 - rm -rf causes LBUG fix (617817)
4098 - file open by multiple tasks fix (618962)
4099 - directory permissions bugs (602707 and 620007)
4100 - journal_stop fixed with locking (611313)
4101 - O_APPEND failures resolved (618273, perhaps 614459)
4102 - lconf PATH fix (619770)
4103 - IA64 build fix (621450)
4104 - RPC buffer sizes scale with amount of memory
4106 2002-10-01 Phil Schwan <phil@clusterfs.com>
4109 - locks would be cancelled without throwing away data pages,
4110 resulting in inconsistent data (605627)
4111 - inode attributes were not always being refreshed (605627, 612449)
4112 - lconf now continues to cleanup after lctl reports an error
4113 - MDS now enforces user permissions (602707)
4114 - lprocfs cleanup fixed, but not yet enabled (614157)
4115 - fixed infinite server hang, should a client not respond to an AST
4116 - avoid going into recovery if user calls readlink() with a buffer
4117 that's too small (613941)
4118 - AST RPCs no longer require replies (614867) -- this may be changed
4119 - don't crash server if client sends an IOV that's too big (611336)
4120 - fixed lock conversion deadlock (611892)
4121 - fixed the following of symlinks (614622)
4122 * recovery: the server can remove locks from a client that dies, other
4123 clients can make progress
4124 * more extN patch fixes
4125 * compile-time configurable ptlrpc buffer allocations
4127 - collaborative read cache document
4128 - Lustre Lite Performance CDR document-in-progress
4130 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4133 - fix typo in patch-2.4.18
4135 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4138 - clear ptlrpc request each time in handle_incoming_request()
4139 - unlink of files now destroys the object on the OST
4141 2002-09-19 Peter Braam <braam@clusterfs.com>
4143 * add hard link support
4144 * change obdfile creation method
4145 * kernel patch changed
4147 2002-09-19 Peter Braam <braam@clusterfs.com>
4150 - stack overflow bug in extN fixed
4152 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4154 * documentation updates
4155 - add man pages for config tools
4156 - update tests/README to describe testing with new config tools
4157 - finish metadata API descriptions
4158 * bug fixes and cleanups
4159 - statfs workaround for 16TB limit
4160 - LOV stripe allocation improved, can stripe on subset of OSTs
4161 - LOV file size/IO offset was wrong for files > 4GB in size
4162 - object EA data was being dropped, caused files to be unreadable
4163 - memory overflow with non-LOV OST caused memory corruption
4164 - fixed regression tests to work with new config tools, obdfilter
4165 - fixed bug when directory size became larger than 1 block
4166 - fixed bug (for single client case) when PWD was deleted
4167 - invalidate local directory pages when doing intent-based ops
4168 - avoid LDLM oops when lock callback contained bad data
4170 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4172 * documentation updates
4173 * bug fixes and cleanups
4174 - configuration tools
4177 - 64-bit compile warnings
4178 - 64-bit internal statfs data
4180 * test_brw on persistent OST devices
4182 * lprocfs (disabled)
4184 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4186 * documentation updates
4187 * bug fixes and cleanups
4188 * configuration tools
4190 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4193 * many small fixes to 0_5_4
4194 * io/network handling
4195 * thinkos in MDS operations
4197 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4200 * crucial basic fixes to 0.5.3
4201 * IOR, Iozone work over Elan
4204 2002-08-07 Phil Schwan <phil@clusterfs.com>
4205 * version 0_5_3, our first alpha
4206 * we use the new Portals iovs
4207 * documentation updates
4208 * bug fixes and cleanups
4209 * small changes in the DLM wire protocol
4211 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4212 * version 0_5_1 with some initial stability,
4213 * locking on MD and file I/O.
4214 * documentation updates
4215 * several bug fixes since 0.5.0
4216 * small changes in wire protocol
4218 2002-07-18 Phil Schwan <phil@clusterfs.com>
4220 * delivered as Lustre Light Alpha
4221 * fixed a crash after handling invalid MDS requests
4222 * fixed directory pages for architectures with non-4k pages sizes
4224 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4225 * release version v0_4_4
4226 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4228 * Fixes a number of interruption problems with OST operations.
4229 * Update documentation for portals header changes
4230 * Move all wire protocol structs/defines to lustre_idl.h
4231 * Fixes symlink length bug.
4232 * Add tcpdump to repository.
4234 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4235 * release version v0_4_3
4236 * Fixes statfs for inodes on extN.
4237 * Fixes bug in runtests which would delete /etc/hosts.
4238 * Use 64-bit object IDs wherever possible (not into VFS though)
4239 Remove ost_get_info, which is unused by lustre, and out of date.
4241 2002-07-03 Peter Braam <braam@clusterfs.com>
4242 * release version v0_4_2 Fixes a lookup error (type not passed)
4243 * move forward to head of Portals
4244 * move forward to latest Lustre kernel
4246 2002-06-25 Peter Braam <braam@clusterfs.com>
4247 * release version v0_4_1. Hopefully stable on single node use.